iccv iccv2013 iccv2013-322 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Karteek Alahari, Guillaume Seguin, Josef Sivic, Ivan Laptev
Abstract: We seek to obtain a pixel-wise segmentation and pose estimation of multiple people in a stereoscopic video. This involves challenges such as dealing with unconstrained stereoscopic video, non-stationary cameras, and complex indoor and outdoor dynamic scenes. The contributions of our work are two-fold: First, we develop a segmentation model incorporating person detection, pose estimation, as well as colour, motion, and disparity cues. Our new model explicitly represents depth ordering and occlusion. Second, we introduce a stereoscopic dataset with frames extracted from feature-length movies “StreetDance 3D ” and “Pina ”. The dataset contains 2727 realistic stereo pairs and includes annotation of human poses, person bounding boxes, and pixel-wise segmentations for hundreds of people. The dataset is composed of indoor and outdoor scenes depicting multiple people with frequent occlusions. We demonstrate results on our new challenging dataset, as well as on the H2view dataset from (Sheasby et al. ACCV 2012).
Reference: text
sentIndex sentText sentNum sentScore
1 Pose Estimation and Segmentation of People in 3D Movies Karteek Alahari1,∗ Guillaume Seguin2,∗ Josef Sivic1,∗ 1Inria 2E´cole Normale Sup e´rieure Ivan Laptev1,∗ Abstract We seek to obtain a pixel-wise segmentation and pose estimation of multiple people in a stereoscopic video. [sent-1, score-0.736]
2 The contributions of our work are two-fold: First, we develop a segmentation model incorporating person detection, pose estimation, as well as colour, motion, and disparity cues. [sent-3, score-1.228]
3 The dataset contains 2727 realistic stereo pairs and includes annotation of human poses, person bounding boxes, and pixel-wise segmentations for hundreds of people. [sent-6, score-0.554]
4 First and foremost, we wish to develop a mid-level representation of stereoscopic videos suitable for subsequent video understanding tasks such as recognition of actions and interactions of people [42]. [sent-15, score-0.45]
5 Second, disparity cues available from stereoscopic movies are expected to improve results of person segmentation and pose estimation. [sent-19, score-1.571]
6 Such results, in turn, can be used as a (noisy) supervisory signal for learning person segmentation and pose estimation in monocular videos or still images [1, 17, 26, 40]. [sent-20, score-0.701]
7 segmentation of people will also support interactive annotation, editing, and navigation in stereo videos [16, 22], which are important tasks in post-production and home video applications. [sent-22, score-0.513]
8 Given the recent success of analyzing people in range data from active sensors, such as Microsoft Kinect [27, 3 1], and a plethora of methods to estimate pixel-wise depth from stereo pairs [2], the task at hand may appear solved. [sent-23, score-0.394]
9 We cast the problem as a discrete labelling task involving multiple person labels, devise a suitable cost function (Section 2), and optimize it efficiently (Section 4). [sent-28, score-0.396]
10 We compute the disparity map (b) from the stereo pair. [sent-40, score-0.789]
11 Occlusion-aware unary costs based on disparity and articulated pose mask are computed for all the people detected in the scene. [sent-41, score-1.457]
12 In (c) we show the unary cost for the person labelled 1. [sent-42, score-0.399]
13 The estimated articulated pose for person 1 is shown in (e). [sent-45, score-0.633]
14 (f) shows the final segmentation result, where each colour represents a unique person, and the numbers denote the front (0) to back (4) ordering of people. [sent-46, score-0.361]
15 In addition to the significant progress in human pose estimation in still images and videos [15, 18, 28, 41], there has been some work in joint pose estimation and segmentation [20, 24, 30, 36]. [sent-50, score-0.677]
16 The proposed method not only computes a segmentation of people and their poses, but also estimates their depth ordering and occlusion. [sent-55, score-0.429]
17 The model incorporates person detections and learnt articulated pose-specific segmentation masks, as well as colour, motion, and stereo disparity cues. [sent-66, score-1.424]
18 As a second contribution, we introduce a new annotated dataset with more than 400 pixel-wise segmentations of people in frames extracted from a stereoscopic movie. [sent-68, score-0.495]
19 Here we consider a stereo pair (only the left image is shown in the figure), estimate the disparity for every pixel, and use it together with person detections, colour and motion features, and pose estimates, to segment individual people, as shown in Figure 1(f). [sent-73, score-1.525]
20 We initialize our model using automatically obtained person detections and assign every detection to a person, i. [sent-74, score-0.383]
21 1 The cost of assigning a person (or background) label, from the set L, to every pixel i, E(x; Θ, τ), igsr given by: 1With a slight abuse of terminology we refer to image regions that correspond to other objects, which may lie in front of or behind people, as background. [sent-87, score-0.45]
22 , Θs cLh−a1r r}a atnerdi disparity parameters Θ{τ =0, τ {1Θ, . [sent-104, score-0.649]
23 , τL−1}, whe}re aΘndl a dnidsp τarli represent tehres pose a {nτd disparity parameters rfoer a person label l respectively. [sent-107, score-1.18]
24 Note that the pose and disparity parameters vary across time. [sent-108, score-0.869]
25 However, in our problem, we also aim to optimize over the set of pose and disparity parameters. [sent-113, score-0.844]
26 Each pixel itakes one of the person or background labels from the label set L. [sent-121, score-0.391]
27 Given a person detection and its corresponding pose estimate Θl, the problem of computing the label likelihood βl can be viewed as that of segmenting an image into person vs. [sent-127, score-0.925]
28 However, we do not make a binary decision of assigning pixels to either the person or the background label. [sent-129, score-0.364]
29 This computation is more akin to generating a soft likelihood map for each pixel taking a particular person label. [sent-130, score-0.374]
30 We define this using disparity and pose cues as: βil = αlψp(Θl) + (1 − αl)ψd(τl), where ψp(Θl) is an articulated pose m)a +sk ( d1e −scri αbed in Section 3, ψd(τl) is a disparity likelihood, and αl is a mixing parameter that controls the relative influence of pose and disparity. [sent-131, score-2.069]
31 The disparity potential is given by: ψd(di;τl,σl) = exp? [sent-132, score-0.624]
32 ,(5) where di is the disparity value computed at pixel i. [sent-134, score-0.675]
33 The disparity potential is a Gaussian characterized by mean τl and standard deviation σl, which together with the pose parameter Θl determines the model for person l. [sent-135, score-1.124]
34 In some cases, the disparity cue used for computing the unary costs may not be very strong or may “leak” into the background (see example in Figure 5). [sent-139, score-0.801]
35 2The order is determined by the disparity parameters τ as discussed in Section 4. [sent-140, score-0.649]
36 Given a pose estimate (a), we compute a pose-specific mask (b) using per- mixture part masks learnt from manually segmented training data. [sent-143, score-0.57]
37 In (b,c), the cost for a pixel to take a person label is denoted by the red (low) blue (high) spectrum of colours. [sent-146, score-0.419]
38 The function (di − dj)2 measures the difference in disparity between pixels i− −an dd j. [sent-151, score-0.653]
39 Thus far we have discussed the model given person detections, their pose and disparity parameters. [sent-155, score-1.124]
40 Estimating an Articulated Pose Mask The aim here is to obtain an articulated pose segmentation mask for each person in the image, which can act as a strong cue to guide the pixel-wise labelling. [sent-159, score-0.955]
41 First, we incorporate disparity as input to take advantage of the available stereo signal. [sent-162, score-0.789]
42 Second, we augment the output to provide an articulated pose-specific soft-segmentation mask learnt from manually annotated training data. [sent-163, score-0.383]
43 We found this to perform empirically better than using the articulated pose estimator [41] for detecting people. [sent-167, score-0.389]
44 To benefit from the stereo signal, we trained a joint appearance and disparity model by concatenating appearance and disparity features into one representation. [sent-168, score-1.495]
45 The disparity features are obtained by comput- ing HOG on disparity maps. [sent-170, score-1.248]
46 Our HOG feature representation for disparity maps is similar to that used in [32, 35] for person/pedestrian detection. [sent-171, score-0.624]
47 We track the person detections computed in each frame of the video, and interpolate to fillin any missing detections, similar to [13]. [sent-172, score-0.395]
48 We estimate the pose of the person within each person detection bounding box. [sent-175, score-0.841]
49 We restrict our pose estimation models to upper body poses, which are more commonly found in movie data. [sent-176, score-0.368]
50 Again, to benefit from the stereo video, we extract both appearance and disparity features in the frame. [sent-177, score-0.818]
51 , between people in similar clothing, can be more pronounced in the disparity map. [sent-180, score-0.794]
52 The mixture components for an elbow part, for example, can be interpreted as capturing different appearances of the elbow as the pose varies, including occlusions by other limbs or people, that are explicitly labelled in the training data. [sent-188, score-0.452]
53 The output of the pose estimator is the location of the individual parts in the frame as shown in Figure 3(a). [sent-191, score-0.361]
54 To obtain a pose-specific mask we learn an average mask for each mixture component for each part. [sent-192, score-0.414]
55 At test time, given an estimated pose with an instantiated mixture component c∗ for a part k, the likelihood for the person, ψp(Θ, i) at pixel i, is obtained by laying out and composing the articulated masks mkc∗ for all the parts. [sent-197, score-0.584]
56 We found that taking the max was beneficial for person segmentation targeted in this pa- per as it suppresses internal edges between body parts, such as a hand positioned in front of the torso. [sent-199, score-0.461]
57 An illustration of the articulated pose masks for various examples is shown in Figure 3. [sent-200, score-0.456]
58 Inference In the previous section we have outlined how we compute the pose parameters Θl and the corresponding articulated pose mask for each person l. [sent-203, score-1.068]
59 The aim is to compute the optimal disparity parameters τ∗ and pixel labels x∗ given the pose parameters Θ, as described by the minimization problem (2). [sent-205, score-0.945]
60 It is well known that minimizing multilabel functions such as E(x; Θ, τ), which corresponds to the segmentation problem, given the pose and disparity parameters, is in itself NP-hard (for the type of smoothness cost we use) [6]. [sent-206, score-1.071]
61 The additional complexity of optimizing over disparity parameters τ further adds to the challenge. [sent-207, score-0.649]
62 We propose a two-step strategy, where we first: (i) estimate the optimal disparity parameters τ∗ using an approximation to (2), without the pairwise terms; and then (ii) obtain the pixel labels x∗ with the estimated (and now fixed) parameters τ∗ by minimizing the full cost (1). [sent-208, score-0.782]
63 The estimation of the set of disparity parameters τ for all the people in a frame can be succinctly written as: τ∗= argm{τi}nE˜( x˜;Θ,τ), (7) where we further approximate the original cost function (1) by only using u? [sent-211, score-0.96]
64 Further, the disparity parameter τ is inversely related to depth, and determines the front-to-back order of people in a frame. [sent-217, score-0.794]
65 rTehleat siveet oorfd possible disparity parameter vpatilmueisz nfogr oeavechr person can still be large, and exploring the exponentially many combinations for all the people in the frame may not be feasible. [sent-219, score-1.122]
66 4 sNmoatell lt sheatt tohfe ( disparity parameters are e,s ftoimr eaatcehd jointly f lo. [sent-221, score-0.649]
67 With the estimated disparity (and pose) parameters, we compute the unary and smoothness costs, and use the efficient α-expansion algorithm [8] to optimize (1). [sent-224, score-0.752]
68 This assigns every pixel a person or background label from the set L. [sent-225, score-0.391]
69 Experiments In this section we detail our method for extracting disparity maps from stereo videos, and report results for person detection, pose estimation, and segmentation. [sent-227, score-1.289]
70 A joint estimation of motion and disparity from video is also possible [38]. [sent-233, score-0.751]
71 4Using a thresholded pose mask, we compute mean disparity μl of all the pixels within, and set {τl} = {μl, μl σl}. [sent-237, score-0.873]
72 Precision-recall curves for person detection using appearance (HOG) and disparity (HOGdisp) based detectors, as well as the jointly trained appearance & disparity based detector (HOGcomb). [sent-240, score-1.646]
73 Note that the detectors using disparity cues have an almost perfect precision until around 35% recall. [sent-241, score-0.652]
74 Furthermore, the ability to handle occlusions explicitly resulted in better disparity maps than other methods, such as [25]. [sent-251, score-0.691]
75 We use the horizontal component of the estimated disparity field in our formulation. [sent-252, score-0.624]
76 We follow [35] and work with disparity values directly rather than depth to avoid problems with infinite depth, and amplifying errors at small disparities. [sent-253, score-0.683]
77 Estimating the dense disparity field for a single stereo pair of 960 540 pixels takes approximately 30 secsotnedres on a rm oofd 9e6r0n ×GP 5U4 using ltsh eta implementation flryo 3m0 [ s5e]c. [sent-254, score-0.818]
78 We trained our person detection and pose estimation methods on an annotated dataset from the featurelength movie “StreetDance”. [sent-256, score-0.704]
79 The training set is composed of 520 annotated person bounding boxes and poses from 265 frames. [sent-259, score-0.435]
80 The test set for evaluating person detection has 638 person bounding boxes in 193 frames, among which a few do not contain any people. [sent-261, score-0.65]
81 Given the cost of annotating poses and pixel-wise segmentation, we evaluated them on a smaller subset of 180 frames, containing 464 annotated person segmentations and poses. [sent-262, score-0.496]
82 We report person detection and pose estimation results for models trained using: (i) standard HOG extracted from grayscale images (HOG), (ii) HOG extracted from disparity maps (HOGdisp), and (iii) joint appearance and disparity based model, using the concatenation ofthe two features (HOGcomb). [sent-264, score-1.873]
83 Furthermore, combining appearance and disparity cues improves the lower arm localization by about 4%. [sent-275, score-0.681]
84 In other words, this method uses disparity features, but not the pose information. [sent-289, score-0.844]
85 We use this as theperson likelihood ψp, and combine it with disparity likelihood ψd, as explained in Section 2. [sent-291, score-0.71]
86 The third variant (Proposed + pose mask) incorporates the articulated pose mask, described in Section 3. [sent-292, score-0.601]
87 Our complete model (Proposed + pose mask + temporal) introduces temporal smoothness across frames. [sent-293, score-0.525]
88 For the “Colour only” baseline, we used a colour-based model for the unary costs without the disparity potential. [sent-294, score-0.744]
89 The combination of appearance and disparity features (HOGcomb) outperforms the individual estimators (HOG, HOGdisp). [sent-316, score-0.683]
90 We used the result obtained by segmenting in the disparity space, i. [sent-319, score-0.659]
91 The background histogram was computed with bounding boxes harvested from regions with no person detections. [sent-323, score-0.395]
92 From Table 2, the method “Proposed + pose mask + temporal” performs better than the others. [sent-329, score-0.41]
93 ec Ot nan ad 9 6tra0c ×k people, m8se t hoe ee mstiemthaoted the pose of each person, and 30s per frame to perform the segmentation with our non-optimized Matlab implementation. [sent-338, score-0.372]
94 The most prominent failure modes of our method are: (i) challenging poses very different from training data; and (ii) cases where the disparity signal is noisy for people far away from the camera (e. [sent-343, score-0.884]
95 The H2view dataset [30] was acquired using a static stereo rig, in combination with a Kinect active Method Proposed + no mask Proposed + uni mask Proposed + pose mask Proposed + pose mask + temporal Baselines: Colour only Eichner et al. [sent-347, score-1.44]
96 Discussion We have developed a model for segmentation of people in stereoscopic movies. [sent-370, score-0.48]
97 The model explicitly represents occlusions, incorporates person detections, pose estimates, and can recover the depth ordering of people in the scene. [sent-371, score-0.849]
98 The results suggest that disparity estimates from stereo video, while noisy, can serve as a strong cue for localizing and segmenting people. [sent-372, score-0.881]
99 For instance, in Row 1, the segmentation is leaking into background for persons 3 and 5, due to the weak disparity cue for these people far away from the camera. [sent-403, score-0.955]
100 2d articulated human pose estimation and retrieval in (almost) unconstrained still images. [sent-451, score-0.415]
wordName wordTfidf (topN-words)
[('disparity', 0.624), ('person', 0.28), ('pose', 0.22), ('stereoscopic', 0.206), ('mask', 0.19), ('people', 0.17), ('stereo', 0.165), ('streetdance', 0.164), ('colour', 0.154), ('articulated', 0.133), ('hogcomb', 0.123), ('movies', 0.109), ('segmentation', 0.104), ('masks', 0.103), ('hogdisp', 0.102), ('pina', 0.082), ('movie', 0.071), ('layered', 0.07), ('detections', 0.067), ('ordering', 0.067), ('smoothness', 0.066), ('poses', 0.064), ('unary', 0.062), ('mkc', 0.061), ('depth', 0.059), ('segmentations', 0.058), ('costs', 0.058), ('cost', 0.057), ('elbow', 0.054), ('motion', 0.052), ('pixel', 0.051), ('hog', 0.049), ('temporal', 0.049), ('frame', 0.048), ('itj', 0.047), ('xi', 0.044), ('likelihood', 0.043), ('occlusions', 0.042), ('pcp', 0.042), ('body', 0.041), ('sheasby', 0.041), ('video', 0.039), ('pb', 0.038), ('annotated', 0.037), ('detection', 0.036), ('estimation', 0.036), ('estimator', 0.036), ('ofpeople', 0.036), ('front', 0.036), ('videos', 0.035), ('indoor', 0.035), ('segmenting', 0.035), ('outdoor', 0.035), ('ayvaci', 0.034), ('mixture', 0.034), ('harvested', 0.032), ('hips', 0.032), ('labelling', 0.031), ('label', 0.031), ('torr', 0.03), ('rig', 0.03), ('torso', 0.03), ('individual', 0.03), ('pixels', 0.029), ('iand', 0.029), ('appearance', 0.029), ('wrist', 0.029), ('normale', 0.029), ('rieure', 0.029), ('estimates', 0.029), ('background', 0.029), ('boxes', 0.029), ('incorporates', 0.028), ('involving', 0.028), ('eichner', 0.028), ('cues', 0.028), ('cue', 0.028), ('il', 0.027), ('cole', 0.027), ('parts', 0.027), ('signal', 0.026), ('assigning', 0.026), ('head', 0.026), ('human', 0.026), ('buffy', 0.026), ('uni', 0.026), ('parameters', 0.025), ('neighbouring', 0.025), ('explicitly', 0.025), ('bounding', 0.025), ('frames', 0.024), ('trained', 0.024), ('neighbourhood', 0.023), ('limbs', 0.023), ('minx', 0.023), ('sup', 0.023), ('xj', 0.023), ('learnt', 0.023), ('ij', 0.023), ('occlusion', 0.023)]
simIndex simValue paperId paperTitle
same-paper 1 0.99999815 322 iccv-2013-Pose Estimation and Segmentation of People in 3D Movies
Author: Karteek Alahari, Guillaume Seguin, Josef Sivic, Ivan Laptev
Abstract: We seek to obtain a pixel-wise segmentation and pose estimation of multiple people in a stereoscopic video. This involves challenges such as dealing with unconstrained stereoscopic video, non-stationary cameras, and complex indoor and outdoor dynamic scenes. The contributions of our work are two-fold: First, we develop a segmentation model incorporating person detection, pose estimation, as well as colour, motion, and disparity cues. Our new model explicitly represents depth ordering and occlusion. Second, we introduce a stereoscopic dataset with frames extracted from feature-length movies “StreetDance 3D ” and “Pina ”. The dataset contains 2727 realistic stereo pairs and includes annotation of human poses, person bounding boxes, and pixel-wise segmentations for hundreds of people. The dataset is composed of indoor and outdoor scenes depicting multiple people with frequent occlusions. We demonstrate results on our new challenging dataset, as well as on the H2view dataset from (Sheasby et al. ACCV 2012).
2 0.36624134 304 iccv-2013-PM-Huber: PatchMatch with Huber Regularization for Stereo Matching
Author: Philipp Heise, Sebastian Klose, Brian Jensen, Alois Knoll
Abstract: Most stereo correspondence algorithms match support windows at integer-valued disparities and assume a constant disparity value within the support window. The recently proposed PatchMatch stereo algorithm [7] overcomes this limitation of previous algorithms by directly estimating planes. This work presents a method that integrates the PatchMatch stereo algorithm into a variational smoothing formulation using quadratic relaxation. The resulting algorithm allows the explicit regularization of the disparity and normal gradients using the estimated plane parameters. Evaluation of our method in the Middlebury benchmark shows that our method outperforms the traditional integer-valued disparity strategy as well as the original algorithm and its variants in sub-pixel accurate disparity estimation.
3 0.34986365 382 iccv-2013-Semi-dense Visual Odometry for a Monocular Camera
Author: Jakob Engel, Jürgen Sturm, Daniel Cremers
Abstract: We propose a fundamentally novel approach to real-time visual odometry for a monocular camera. It allows to benefit from the simplicity and accuracy of dense tracking which does not depend on visual features while running in real-time on a CPU. The key idea is to continuously estimate a semi-dense inverse depth map for the current frame, which in turn is used to track the motion of the camera using dense image alignment. More specifically, we estimate the depth of all pixels which have a non-negligible image gradient. Each estimate is represented as a Gaussian probability distribution over the inverse depth. We propagate this information over time, and update it with new measurements as new images arrive. In terms of tracking accuracy and computational speed, the proposed method compares favorably to both state-of-the-art dense and feature-based visual odometry and SLAM algorithms. As our method runs in real-time on a CPU, it is oflargepractical valuefor robotics and augmented reality applications. – – 1. Towards Dense Monocular Visual Odometry Tracking a hand-held camera and recovering the threedimensional structure of the environment in real-time is among the most prominent challenges in computer vision. In the last years, dense approaches to these challenges have become increasingly popular: Instead of operating solely on visual feature positions, they reconstruct and track on the whole image using a surface-based map and thereby are fundamentally different from feature-based approaches. Yet, these methods are to date either not real-time capable on standard CPUs [11, 15, 17] or require direct depth measurements from the sensor [7], making them unsuitable for many practical applications. In this paper, we propose a novel semi-dense visual odometry approach for a monocular camera, which combines the accuracy and robustness of dense approaches with the efficiency of feature-based methods. Further, it computes highly accurate semi-dense depth maps from the monocular images, providing rich information about the 3D ∗ This work was supported by the ERC Starting Grant ConvexVision and the DFG project Mapping on Demand Figure1.Semi-Dens MoncularVisualOdometry:Oucfrloas rpe- proach works on a semi-dense inverse depth map and combines the accuracy and robustness of dense visual SLAM methods with the efficiency of feature-based techniques. Left: video frame, Right: color-coded semi-dense depth map, which consists of depth estimates in all image regions with sufficient structure. structure of the environment. We use the term visual odometry as supposed to SLAM, as for simplicity we deliberately maintain only information about the currently visible scene, instead of building a global world-model. – – 1.1. Related Work Feature-based monocular SLAM. In all feature-based methods (such as [4, 8]), tracking and mapping consists of two separate steps: First, discrete feature observations (i.e., their locations in the image) are extracted and matched to each other. Second, the camera and the full feature poses are calculated from a set of such observations disregarding the images themselves. While this preliminary abstrac– tion step greatly reduces the complexity of the overall problem and allows it to be tackled in real time, it inherently comes with two significant drawbacks: First, only image information conforming to the respective feature type and parametrization typically image corners and blobs [6] or line segments [9] is utilized. Second, features have to be matched to each other, which often requires the costly computation of scale- and rotation-invariant descriptors and robust outlier estimation methods like RANSAC. – – Dense monocular SLAM. To overcome these limitations and to better exploit the available image information, dense monocular SLAM methods [11, 17] have recently been proposed. The fundamental difference to keypoint-based approaches is that these methods directly work on the images 11444499 instead of a set of extracted features, for both mapping and tracking: The world is modeled as dense surface while in turn new frames are tracked using whole-image alignment. This concept removes the need for discrete features, and allows to exploit all information present in the image, increasing tracking accuracy and robustness. To date however, doing this in real-time is only possible using modern, powerful GPU processors. Similar methods are broadly used in combination with RGB-D cameras [7], which directly measure the depth of each pixel, or stereo camera rigs [3] greatly reducing the – complexity of the problem. Dense multi-view stereo. Significant prior work exists on multi-view dense reconstruction, both in a real-time setting [13, 11, 15], as well as off-line [5, 14]. In particular for offline reconstruction, there is a long history of using different baselines to steer the stereo-inherent trade-off between accuracy and precision [12]. Most similar to our approach is the early work of Matthies et al., who proposed probabilistic depth map fusion and propagation for image sequences [10], however only for structure from motion, i.e., not coupled with subsequent dense tracking. 1.2. Contributions In this paper, we propose a novel semi-dense approach to monocular visual odometry, which does not require feature points. The key concepts are • a probabilistic depth map representation, • tracking based on whole-image alignment, • the reduction on image-regions which carry informattihoen (esdeumctii-odenn osen), i manadg • the full incorporation of stereo measurement uncertainty. To the best of our knowledge, this is the first featureless, real-time monocular visual odometry approach, which runs in real-time on a CPU. 1.3. Method Outline Our approach is partially motivated by the basic principle that for most real-time applications, video information is abundant and cheap to come by. Therefore, the computational budget should be spent such that the expected information gain is maximized. Instead of reducing the images to a sparse set of feature observations however, our method continuously estimates a semi-dense inverse depth map for the current frame, i.e., a dense depth map covering all image regions with non-negligible gradient (see Fig. 2). It is comprised of one inverse depth hypothesis per pixel modeled by a Gaussian probability distribution. This representation still allows to use whole-image alignment [7] to track new orignalimagesemi-densedepthmap(ours)clfoasre keypointdepthmap[8]densedepthmap[1 ]RGB-Dcamera[16] Figure 2. Semi-Dense Approach: Our approach reconstructs and tracks on a semi-dense inverse depth map, which is dense in all image regions carrying information (top-right). For comparison, the bottom row shows the respective result from a keypoint-based approach, a fully dense approach and the ground truth from an RGB-D camera. frames, while at the same time greatly reducing computational complexity compared to volumetric methods. The estimated depth map is propagated from frame to frame, and updated with variable-baseline stereo comparisons. We explicitly use prior knowledge about a pixel’s depth to select a suitable reference frame on a per-pixel basis, and to limit the disparity search range. The remainder of this paper is organized as follows: Section 2 describes the semi-dense mapping part of the proposed method, including the derivation of the observation accuracy as well as the probabilistic data fusion, propagation and regularization steps. Section 3 describes how new frames are tracked using whole-image alignment, and Sec. 4 summarizes the complete visual odometry method. A qualitative as well as a quantitative evaluation is presented in Sec. 5. We then give a brief conclusion in Sec. 6. 2. Semi-Dense Depth Map Estimation One of the key ideas proposed in this paper is to estimate a semi-dense inverse depth map for the current camera image, which in turn can be used for estimating the camera pose of the next frame. This depth map is continuously propagated from frame to frame, and refined with new stereo depth measurements, which are obtained by performing per-pixel, adaptive-baseline stereo comparisons. This allows us to accurately estimate the depth both of close-by and far-away image regions. In contrast to previous work that accumulates the photometric cost over a sequence of several frames [11, 15], we keep exactly one inverse depth hypothesis per pixel that we represent as Gaussian probability distribution. This section is comprised of three main parts: Sec11445500 reference small baseline medium baseline large baseline tcso0120 .050.10.150.20.2sl5m areagdleiulm0.3 inverse depth d Figure 3. Variable Baseline Stereo: Reference image (left), three stereo images at different baselines (right), and the respective matching cost functions. While a small baseline (black) gives a unique, but imprecise minimum, a large baseline (red) allows for a very precise estimate, but has many false minima. tion 2. 1 describes the stereo method used to extract new depth measurements from previous frames, and how they are incorporated into the prior depth map. In Sec. 2.2, we describe how the depth map is propagated from frame to frame. In Sec. 2.3, we detail how we partially regularize the obtained depth map in each iteration, and how outliers are handled. Throughout this section, d denotes the inverse depth of a pixel. 2.1. Stereo-Based Depth Map Update It is well known [12] that for stereo, there is a trade-off between precision and accuracy (see Fig. 3). While many multiple-baseline stereo approaches resolve this by accumulating the respective cost functions over many frames [5, 13], we propose a probabilistic approach which explicitly takes advantage of the fact that in a video, smallbaseline frames are available before large-baseline frames. The full depth map update (performed once for each new frame) consists of the following steps: First, a subset of pixels is selected for which the accuracy of a disparity search is sufficiently large. For this we use three intuitive and very efficiently computable criteria, which will be derived in Sec. 2. 1.3. For each selected pixel, we then individually select a suitable reference frame, and perform a onedimensional disparity search. Propagated prior knowledge is used to reduce the disparity search range when possible, decreasing computational cost and eliminating false minima. The obtained inverse depth estimate is then fused into the depth map. 2.1.1 Reference Frame Selection Ideally, the reference frame is chosen such that it maximizes the stereo accuracy, while keeping the disparity search range as well as the observation angle sufficiently cur ent framepixel’s “age” -4.8 s -3.9 s -3.1 s -2.2 s -1.2 s -0.8 s -0.5 s -0.4 s Figure 4. Adaptive Baseline Selection: For each pixel in the new frame (top left), a different stereo-reference frame is selected, based on how long the pixel was visible (top right: the more yellow, the older the pixel.). Some of the reference frames are displayed below, the red regions were used for stereo comparisons. small. As the stereo accuracy depends on many factors and because this selection is done for each pixel independently, we employ the following heuristic: We use the oldest frame the pixel was observed in, where the disparity search range and the observation angle do not exceed a certain threshold (see Fig. 4). If a disparity search is unsuccessful (i.e., no good match is found), the pixel’s “age” is increased, such that subsequent disparity searches use newer frames where the pixel is likely to be still visible. 2.1.2 Stereo Matching Method We perform an exhaustive search for the pixel’s intensity along the epipolar line in the selected reference frame, and then perform a sub-pixel accurate localization of the matching disparity. If a prior inverse depth hypothesis is available, the search interval is limited by d 2σd, where d and σd de,e nthoete s etharec mean avnadl ssta lnimdaiterdd d beyv dia ±tion 2σ σof the prior hypothesis. Otherwise, the full disparity range is searched. In our implementation, we use the SSD error over five equidistant points on the epipolar line: While this significantly increases robustness in high-frequent image regions, it does not change the purely one-dimensional nature of this search. Furthermore, it is computationally efficient, as 4 out ± of 5 interpolated image values can be re-used for each SSD evaluation. 2.1.3 Uncertainty Estimation In this section, we use uncertainty propagation to derive an expression for the error variance σd2 on the inverse depth d. 11445511 In general this can be done by expressing the optimal inverse depth d∗ as a function of the noisy inputs here we consider the images I0, I1 themselves, their relative orientation ξ and the camera calibration in terms of a projection function π1 – d∗ = d(I0, I1, ξ, π) . The error-variance of d∗ is then given by σd2 = JdΣJdT, (1) (2) where Jd is the Jacobian of d, and Σ the covariance of the input-error. For more details on covariance propagation, including the derivation of this formula, we refer to [2]. For simplicity, the following analysis is performed for patchfree stereo, i.e., we consider only a point-wise search for a single intensity value along the epipolar line. For this analysis, we split the computation into three steps: First, the epipolar line in the reference frame is computed. Second, the best matching position λ∗ ∈ R along it (i.e., the disparity) is determined. Third, the i∈nv eRrse al depth d∗ is computed from the disparity λ∗ . The first two steps involve two independent error sources: the geometric error, which originates from noise on ξ and π and affects the first step, and the photometric error, which originates from noise in the images I0, I1 and affects the second step. The third step scales these errors by a factor, which depends on the baseline. Geometric disparity error. The geometric error is the error ?λ on the disparity λ∗ caused by noise on ξ and π. While it would be possible to model, propagate, and estimate the complete covariance on ξ and π, we found that the gain in accuracy does not justify the increase in computational complexity. We therefore use an intuitive approximation: Let the considered epipolar line segment L ⊂ R2 be deLfineted th by L := ?l0 + λ?llyx? |λ ∈ S? , (3) where λ is the disparity with search interval S, (lx , ly)T the normalized epipolar line direction and l0 the point corresponding to infinite depth. We now assume that only the absolute position of this line segment, i.e., l0 is subject to isotropic Gaussian noise ?l . As in practice we keep the searched epipolar line segments short, the influence of rotational error is small, making this a good approximation. Intuitively, a positioning error ?l on the epipolar line causes a small disparity error ?λ if the epipolar line is parallel to the image gradient, and a large one otherwise (see Fig. 5). This can be mathematically derived as follows: The image constrains the optimal disparity λ∗ to lie on a certain isocurve, i.e. a curve of equal intensity. We approximate 1In the linear case, this is the camera matrix K – in practice however, nonlinear distortion and other (unmodeled) effects also play a role. FiguLre5.Geo?l mλetricDigs,palrityEroL?rl:Influe?nλceofgasmla posi- tioning error ?l of the epipolar line on the disparity error ?λ . The dashed line represents the isocurve on which the matching point has to lie. ?λ is small if the epipolar line is parallel to the image gradient (left), and a large otherwise (right). this isocurve to be locally linear, i.e. the gradient direction to be locally constant. This gives l0 + λ∗ ?llxy? =! + γ?−gxgy?, g0 γ ∈ R (4) where g := (gx , gy) ?is the image gradient and g0 a point on the isoline. The influence of noise on the image values will be derived in the next paragraph, hence at this point g and g0 are assumed noise-free. Solving for λ gives the optimal disparity λ∗ in terms of the noisy input l0: λ∗(l0) =?g,g?g0,−l? l0? (5) Analogously to (2), the variance of the geometric disparity error can then be expressed as σλ2(ξ,π)= Jλ∗(l0)?σ0l2 σ0l2?JλT∗(l0)=?gσ,l 2?2, (6) where g is the normalized image gradient, lthe normalized epipolar line direction and σl2 the variance of ?l. Note that this error term solely originates from noise on the relative camera orientation and the camera calibration π, i.e., it is independent of image intensity noise. ξ Photometric disparity error. Intuitively, this error encodes that small image intensity errors have a large effect on the estimated disparity if the image gradient is small, and a small effect otherwise (see Fig. 6). Mathematically, this relation can be derived as follows. We seek the disparity λ∗ that minimizes the difference in intensities, i.e., λ∗ = mλin (iref − Ip(λ))2, (7) where iref is the reference intensity, and Ip(λ) the image intensity on the epipolar line at disparity λ. We assume a good initialization λ0 to be available from the exhaustive search. Using a first-order Taylor approximation for Ip gives λ∗(I) = λ0 + (iref − Ip(λ0)) g−p1, (8) where gp is the gradient of Ip, that is image gradient along the epipolar line. For clarity we only consider noise on iref and Ip(λ0) ; equivalent results are obtained in the general case when taking into account noise on the image values involved in the computation of gp. The variance of the pho11445522 ?i Ip?λ ?iiIp?λλ Figure 6. Photometric Disparity Error: Noise ?i on the image intensity values causes a small disparity error ?λ if the image gradient along the epipolar line is large (left). If the gradient is small, the disparity error is magnified (right). tometric disparity error is given by σλ2(I) = Jλ∗(I)?σ0i2 σ0i2?Jλ∗(I) =2gσ2pi2, (9) where σi2 is the variance of the image intensity noise. The respective error originates solely from noisy image intensity values, and hence is independent of the geometric disparity error. Pixel to inverse depth conversion. Using that, for small camera rotation, the inverse depth d is approximately proportional to the disparity λ, the observation variance of the inverse depth σd2,obs can be calculated using σd2,obs = α2 ?σ2λ(ξ,π) + σλ2(I)? , (10) where the proportionality ?constant α in th?e general, nonrectified case – is different for each pixel, and can be calculated from – α :=δδdλ, (11) where δd is the length of the searched inverse depth interval, and δλ the length of the searched epipolar line segment. While α is inversely linear in the length of the camera translation, it also depends on the translation direction and the pixel’s location in the image. When using an SSD error over multiple points along the epipolar line – as our implementation does – a good upper bound for the matching uncertainty is then given by ?min{σ2λ(ξ,π)} + min{σλ2(I)}? σd2,obs-SSD ≤ α2 , (12) where the min goes over all points included in the? SSD error. 2.1.4 Depth Observation Fusion After a depth observation for a pixel in the current image has been obtained, we integrate it into the depth map as follows: If no prior hypothesis for a pixel exists, we initialize it directly with the observation. Otherwise, the new observation is incorporated into the prior, i.e., the two distribu- tions are multiplied (corresponding to the update step in a Knoailsmya onb fsieltrvera)t:io Gniv Nen(do a, pσrio2o)r, d thiest priobsutetiroionr N is( gdipv,eσnp2 b)y and a N?σ2pdσo2p++ σ σo2o2dp,σ2σpp2+σo2 σo2?. 2.1.5 (13) Summary of Uncertainty-Aware Stereo New stereo observations are obtained on a per-pixel basis, adaptively selecting for each pixel a suitable reference frame and performing a one-dimensional search along the epipolar line. We identified the three major factors which determine the accuracy of such a stereo observation, i.e., • the photometric disparity error σλ2(ξ,π), depending on tphheo magnitude sofp trhiet image gradient along the epipolar line, • the geometric disparity error σλ2(I) ,depending on the athnegl gee bometewtereinc dthisep image gradient and the epipolar line (independent of the gradient magnitude), and • the pixel to inverse depth ratio α, depending on the camera etlra tons ilantvioenrs, eth dee pfothcal r length ,a dndep tehned pixel’s position. These three simple-to-compute and purely local criteria are used to determine for which pixel a stereo update is worth the computational cost. Further, the computed observation variance is then used to integrate the new measurements into the existing depth map. 2.2. Depth Map Propagation We continuously propagate the estimated inverse depth map from frame to frame, once the camera position of the next frame has been estimated. Based on the inverse depth estimate d0 for a pixel, the corresponding 3D point is calculated and projected into the new frame, providing an inverse depth estimate d1 in the new frame. The hypothesis is then assigned to the closest integer pixel position to eliminate discretization errors, the sub-pixel accurate image location of the projected point is kept, and re-used for the next propagation step. For propagating the inverse depth variance, we assume the camera rotation to be small. The new inverse depth d1 can then be approximated by – d1(d0) = (d0−1 − tz)−1, (14) where tz is the camera translation along the optical axis. The variance of d1 is hence given by σd21= Jd1σd20JTd1+ σp2=?dd01?4σd20+ σp2, (15) where σp2 is the prediction uncertainty, which directly corresponds to the prediction step in an extended Kalman filter. It can also be interpreted as keeping the variance on 11445533 in the top right shows the new frame I2 (x) without depth information. Middle: Intermediate steps while minimizing E(ξ) on different pyramid levels. The top row shows the back-warped new frame I2 (w(x, d, ξ)), the bottom row shows the respective residual image I2 (w(x, di,ξ)) − I1 (x) . The bottom right image shows the final pixel-weights (black = small weight). Small weights mainly correspond to newly oc,cξl)ud)e −d or disoccluded pixel. tWhe z fo-cuonodrtd hina t uesi onfg a sm poailnlt v failxue ds, fo i.re. σ,p2 sedteticnrgea σsez2s0 d=rift σ,z2 a1s. it causes the estimated geometry to gradually ”lock” into place. Collision handling. At all times, we allow at most one inverse depth hypothesis per pixel: If two inverse depth hypothesis are propagated to the same pixel in the new frame, we distinguish between two cases: 1. if they are statistically similar, i.e., lie within 2σ bounds, they are treated as two independent observations of the pixel’s depth and fused according to (13). 2. otherwise, the point that is further away from the camera is assumed to be occluded, and is removed. 2.3. Depth Map Regularization For each frame – after all observations have been incorporated – we perform one regularization iteration by assign- ing each inverse depth value the average of the surrounding inverse depths, weighted by their respective inverse variance. To preserve sharp edges, if two adjacent inverse depth values are statistically different, i.e., are further away than 2σ, they do not contribute to one another. Note that the respective variances are not changed during regularization to account for the high correlation between neighboring hypotheses. Instead we use the minimal variance of all neighboring pixel when defining the stereo search range, and as a weighting factor for tracking (see Sec. 3). Outlier removal. To handle outliers, we continuously keep track of the validity of each inverse depth hypothesis in terms of the probability that it is an outlier, or has become invalid (e.g., due to occlusion or a moving object). For each successful stereo observation, this probability is decreased. It is increased for each failed stereo search, if the respective intensity changes significantly on propagation, or when the absolute image gradient falls below a given threshold. If, during regularization, the probability that all contributing neighbors are outliers i.e., the product of their individual outlier-probabilities rises above a given threshold, the hypothesis is removed. Equally, if for an “empty” pixel this product drops below a given threshold, a new hypothesis is created from the neighbors. This fills holes arising from the forward-warping nature of the propagation step, and dilates the semi-dense depth map to a small neighborhood around sharp image intensity edges, which signifi– – × cantly increases tracking and mapping robustness. 3. Dense Tracking Based on the inverse depth map of the previous frame, we estimate the camera pose of the current frame using dense image alignment. Such methods have previously been applied successfully (in real-time on a CPU) for tracking RGB-D cameras [7], which directly provide dense depth measurements along with the color image. It is based on the direct minimization of the photometric error ri (ξ) := (I2 (w(xi, di , ξ)) − I1 , (16) where the warp function w : Ω1 R R6 → Ω2 maps each point xi ∈ Ω1 in the reference× image RI1 →to Ωthe respective point w(x∈i, Ωdi, ξ) ∈ Ω2 in the new image I2. As input it requires the 3D,ξ pose Ωof the camera ξ ∈ R6 and uses the reestqiumiraetesd t hienv 3erDse p depth fd it ∈e cRa mfore rthae ξ pixel in I1. Note that no depth information with respect t toh Ie2 p i sx required. To increase robustness to self-occlusion and moving objects, we apply a weighting scheme as proposed in [7]. Further, we add the variance of the inverse depth σd2i as an additional weighting term, making the tracking resistant to recently initialized and still inaccurate depth estimates from 11445544 (xi))2 Figure 8. Examples: Top: Camera images overlaid with the respective stimated semi-dense inverse depth map. Bot om: 3D view of tracked scene. Note the versatility of our approach: It accurately reconstructs and tracks through (outside) scenes with a large depth- variance, including far-away objects like clouds , as well as (indoor) scenes with little structure and close to no image corners / keypoints. More examples are shown in the attached video. the mapping process. The final energy that is minimized is hence given by E(ξ) :=?iα(rσid2(iξ))ri(ξ), (17) where α : R → R defines the weight for a given residual. Minimizing t h→is error can b thee interpreted as computing uthale. maximum likelihood estimator for ξ, assuming independent noise on the image intensity values. The resulting weighted least-squares problem is solved efficiently using an iteratively reweighted Gauss-Newton algorithm coupled with a coarse-to-fine approach, using four pyramid levels. Figure 7 shows an example of the tracking process. For further details on the minimization we refer to [1]. 4. System Overview Tracking and depth estimation is split into two separate threads: One continuously propagates the inverse depth map to the most recent tracked frame, updates it with stereocomparisons and partially regularizes it. The other simultaneously tracks each incoming frame on the most recent available depth map. While tracking is performed in real- time at 30Hz, one complete mapping iteration takes longer and is hence done at roughly 15Hz if the map is heavily populated, we adaptively reduce the number of stereo comparisons to maintain a constant frame-rate. For stereo observations, a buffer of up to 100 past frames is kept, automatically removing those that are used least. We use a standard, keypoint-based method to obtain the relative camera pose between two initial frames, which are then used to initialize the inverse depth map needed for tracking successive frames. From this point onward, our method is entirely self-contained. In preliminary experiments, we found that in most cases our approach is even able to recover from random or extremely inaccurate initial depth maps, indicating that the keypoint-based initialization might become superfluous in the future. Table 1. Results on RGB-D Benchmark position drift (cm/s) rotation drift (deg/s) ours [7] [8] ours [7] [8] – fr2/xyz fr2/desk 0.6 2.1 0.6 2.0 8.2 - 0.33 0.65 0.34 0.70 3.27 - 5. Results We have tested our approach on both publicly available benchmark sequences, as well as live, using a hand-held camera. Some examples are shown in Fig. 8. Note that our method does not attempt to build a global map, i.e., once a point leaves the field of view of the camera or becomes occluded, the respective depth value is deleted. All experiments are performed on a standard consumer laptop with Intel i7 quad-core CPU. In a preprocessing step, we rectify all images such that a pinhole camera-model can be applied. 5.1. RGB-D Benchmark Sequences As basis for a quantitative evaluation and to facilitate reproducibility and easy comparison with other methods, we use the TUM RGB-D benchmark [16]. For tracking and mapping we only use the gray-scale images; for the very first frame however the provided depth image is used as initialization. Our method (like any monocular visual odometry method) fails in case of pure camera rotation, as the depth of new regions cannot be determined. The achieved tracking accuracy for two feasible sequences that is, sequences which do not contain strong camera rotation without simultaneous translation is given in Table 1. For comparison we also list the accuracy from (1) a state-of-the-art, dense RGB-D odometry [7], and (2) a state-of-the-art, keypointbased monocular SLAM system (PTAM, [8]). We initialize PTAM using the built-in stereo initializer, and perform a 7DoF (rigid body plus scale) alignment to the ground truth trajectory. Figure 9 shows the tracked camera trajectory for fr2/desk. We found that our method achieves similar accu– – 11445555 era the the the trajectory (black), the depth map of the first frame (blue), and estimated depth map (gray-scale) after a complete loop around table. Note how well certain details such as the keyboard and monitor align. racy as [7] which uses the same dense tracking algorithm but relies on the Kinect depth images. The keypoint-based approach [8] proves to be significantly less accurate and robust; it consistently failed after a few seconds for the second sequence. 5.2. Additional Test Sequences To analyze our approach in more detail, we recorded additional challenging sequences with the corresponding ground truth trajectory in a motion capture studio. Figure 10 shows an extract from the video, as well as the tracked and the ground-truth camera position over time. As can be seen from the figure, our approach is able to maintain a reasonably dense depth map at all times and the estimated camera trajectory matches closely the ground truth. 6. Conclusion In this paper we proposed a novel visual odometry method for a monocular camera, which does not require discrete features. In contrast to previous work on dense tracking and mapping, our approach is based on probabilistic depth map estimation and fusion over time. Depth measurements are obtained from patch-free stereo matching in different reference frames at a suitable baseline, which are selected on a per-pixel basis. To our knowledge, this is the first featureless monocular visual odometry method which runs in real-time on a CPU. In our experiments, we showed that the tracking performance of our approach is comparable to that of fully dense methods without requiring a depth sensor. References [1] S. Baker and I. Matthews. Lucas-Kanade 20 years on: A unifying framework. Technical report, Carnegie Mellon Univ., 2002. 7 [2] A. Clifford. Multivariate Error Analysis. John Wiley & Sons, 1973. 4 sionpito[m ]− 024 2 0 s1xzy0s20s30s40s50s60s Figure 10. Additional Sequence: Estimated camera trajectory and ground truth (dashed) for a long and challenging sequence. The complete sequence is shown in the attached video. [3] A. Comport, E. Malis, and P. Rives. Accurate quadri-focal tracking for robust 3d visual odometry. In ICRA, 2007. 2 [4] A. Davison, I. Reid, N. Molton, and O. Stasse. MonoSLAM: Real-time single camera SLAM. Trans. on Pattern Analysis and Machine Intelligence (TPAMI), 29, 2007. 1 [5] D. Gallup, J. Frahm, P. Mordohai, and M. Pollefeys. Variable baseline/resolution stereo. In CVPR, 2008. 2, 3 [6] C. Harris and M. Stephens. A combined corner and edge detector. In Alvey Vision Conference, 1988. 1 [7] C. Kerl, J. Sturm, and D. Cremers. Robust odometry estimation for RGB-D cameras. In ICRA, 2013. 1, 2, 6, 7, 8 [8] G. Klein and D. Murray. Parallel tracking and mapping for small AR workspaces. In Mixed and Augmented Reality (ISMAR), 2007. 1, 2, 7, 8 [9] G. Klein and D. Murray. Improving the agility of keyframebased SLAM. In ECCV, 2008. 1 [10] M. Pollefes et al. Detailed real-time urban 3d reconstruction from video. IJCV, 78(2-3): 143–167, 2008. 2, 3 [11] L. Matthies, R. Szeliski, and T. Kanade. Incremental estimation of dense depth maps from image image sequences. In CVPR, 1988. 2 [12] R. Newcombe, S. Lovegrove, and A. Davison. DTAM: Dense tracking and mapping in real-time. In ICCV, 2011. 1, 2 [13] M. Okutomi and T. Kanade. A multiple-baseline stereo. Trans. on Pattern Analysis and Machine Intelligence (TPAMI), 15(4):353–363, 1993. 2, 3 [14] T. Sato, M. Kanbara, N. Yokoya, and H. Takemura. Dense 3-d reconstruction of an outdoor scene by hundreds-baseline stereo using a hand-held camera. IJCV, 47: 1–3, 2002. 2 [15] J. Stuehmer, S. Gumhold, and D. Cremers. Real-time dense geometry from a handheld camera. In Pattern Recognition (DAGM), 2010. 1, 2 [16] J. Sturm, N. Engelhard, F. Endres, W. Burgard, and D. Cremers. A benchmark for the evaluation of RGB-D SLAM systems. In Intelligent Robot Systems (IROS), 2012. 2, 7 [17] A. Wendel, M. Maurer, G. Graber, T. Pock, and H. Bischof. Dense reconstruction on-the-fly. In ECCV, 2012. 1 11445566
4 0.22048801 252 iccv-2013-Line Assisted Light Field Triangulation and Stereo Matching
Author: Zhan Yu, Xinqing Guo, Haibing Lin, Andrew Lumsdaine, Jingyi Yu
Abstract: Light fields are image-based representations that use densely sampled rays as a scene description. In this paper, we explore geometric structures of 3D lines in ray space for improving light field triangulation and stereo matching. The triangulation problem aims to fill in the ray space with continuous and non-overlapping simplices anchored at sampled points (rays). Such a triangulation provides a piecewise-linear interpolant useful for light field superresolution. We show that the light field space is largely bilinear due to 3D line segments in the scene, and direct triangulation of these bilinear subspaces leads to large errors. We instead present a simple but effective algorithm to first map bilinear subspaces to line constraints and then apply Constrained Delaunay Triangulation (CDT). Based on our analysis, we further develop a novel line-assisted graphcut (LAGC) algorithm that effectively encodes 3D line constraints into light field stereo matching. Experiments on synthetic and real data show that both our triangulation and LAGC algorithms outperform state-of-the-art solutions in accuracy and visual quality.
5 0.21168929 226 iccv-2013-Joint Subspace Stabilization for Stereoscopic Video
Author: Feng Liu, Yuzhen Niu, Hailin Jin
Abstract: Shaky stereoscopic video is not only unpleasant to watch but may also cause 3D fatigue. Stabilizing the left and right view of a stereoscopic video separately using a monocular stabilization method tends to both introduce undesirable vertical disparities and damage horizontal disparities, which may destroy the stereoscopic viewing experience. In this paper, we present a joint subspace stabilization method for stereoscopic video. We prove that the low-rank subspace constraint for monocular video [10] also holds for stereoscopic video. Particularly, the feature trajectories from the left and right video share the same subspace. Based on this proof, we develop a stereo subspace stabilization method that jointly computes a common subspace from the left and right video and uses it to stabilize the two videos simultaneously. Our method meets the stereoscopic constraints without 3D reconstruction or explicit left-right correspondence. We test our method on a variety of stereoscopic videos with different scene content and camera motion. The experiments show that our method achieves high-quality stabilization for stereoscopic video in a robust and efficient way.
6 0.19849543 28 iccv-2013-A Rotational Stereo Model Based on XSlit Imaging
7 0.19640414 225 iccv-2013-Joint Segmentation and Pose Tracking of Human in Natural Videos
8 0.19421275 317 iccv-2013-Piecewise Rigid Scene Flow
9 0.158163 403 iccv-2013-Strong Appearance and Expressive Spatial Models for Human Pose Estimation
10 0.15443258 242 iccv-2013-Learning People Detectors for Tracking in Crowded Scenes
11 0.1461481 143 iccv-2013-Estimating Human Pose with Flowing Puppets
12 0.14351034 111 iccv-2013-Detecting Dynamic Objects with Multi-view Background Subtraction
13 0.13972755 316 iccv-2013-Pictorial Human Spaces: How Well Do Humans Perceive a 3D Articulated Pose?
14 0.13733463 166 iccv-2013-Finding Actors and Actions in Movies
15 0.13660681 88 iccv-2013-Constant Time Weighted Median Filtering for Stereo Matching and Beyond
16 0.13560943 273 iccv-2013-Monocular Image 3D Human Pose Estimation under Self-Occlusion
17 0.12600961 218 iccv-2013-Interactive Markerless Articulated Hand Motion Tracking Using RGB and Depth Data
18 0.12594281 65 iccv-2013-Breaking the Chain: Liberation from the Temporal Markov Assumption for Tracking Human Poses
19 0.1243014 379 iccv-2013-Semantic Segmentation without Annotating Segments
20 0.12206274 377 iccv-2013-Segmentation Driven Object Detection with Fisher Vectors
topicId topicWeight
[(0, 0.258), (1, -0.143), (2, 0.014), (3, 0.109), (4, 0.112), (5, -0.071), (6, -0.026), (7, -0.006), (8, -0.04), (9, 0.101), (10, 0.047), (11, 0.051), (12, -0.042), (13, -0.088), (14, -0.107), (15, -0.012), (16, -0.142), (17, -0.129), (18, 0.055), (19, 0.191), (20, 0.098), (21, 0.048), (22, -0.152), (23, 0.055), (24, 0.157), (25, 0.07), (26, -0.059), (27, -0.19), (28, 0.003), (29, 0.117), (30, -0.075), (31, 0.007), (32, -0.05), (33, -0.184), (34, 0.24), (35, 0.02), (36, -0.116), (37, -0.046), (38, 0.016), (39, 0.037), (40, -0.014), (41, 0.111), (42, -0.123), (43, -0.078), (44, 0.044), (45, -0.014), (46, 0.006), (47, -0.119), (48, -0.002), (49, -0.042)]
simIndex simValue paperId paperTitle
same-paper 1 0.92855525 322 iccv-2013-Pose Estimation and Segmentation of People in 3D Movies
Author: Karteek Alahari, Guillaume Seguin, Josef Sivic, Ivan Laptev
Abstract: We seek to obtain a pixel-wise segmentation and pose estimation of multiple people in a stereoscopic video. This involves challenges such as dealing with unconstrained stereoscopic video, non-stationary cameras, and complex indoor and outdoor dynamic scenes. The contributions of our work are two-fold: First, we develop a segmentation model incorporating person detection, pose estimation, as well as colour, motion, and disparity cues. Our new model explicitly represents depth ordering and occlusion. Second, we introduce a stereoscopic dataset with frames extracted from feature-length movies “StreetDance 3D ” and “Pina ”. The dataset contains 2727 realistic stereo pairs and includes annotation of human poses, person bounding boxes, and pixel-wise segmentations for hundreds of people. The dataset is composed of indoor and outdoor scenes depicting multiple people with frequent occlusions. We demonstrate results on our new challenging dataset, as well as on the H2view dataset from (Sheasby et al. ACCV 2012).
2 0.79062569 304 iccv-2013-PM-Huber: PatchMatch with Huber Regularization for Stereo Matching
Author: Philipp Heise, Sebastian Klose, Brian Jensen, Alois Knoll
Abstract: Most stereo correspondence algorithms match support windows at integer-valued disparities and assume a constant disparity value within the support window. The recently proposed PatchMatch stereo algorithm [7] overcomes this limitation of previous algorithms by directly estimating planes. This work presents a method that integrates the PatchMatch stereo algorithm into a variational smoothing formulation using quadratic relaxation. The resulting algorithm allows the explicit regularization of the disparity and normal gradients using the estimated plane parameters. Evaluation of our method in the Middlebury benchmark shows that our method outperforms the traditional integer-valued disparity strategy as well as the original algorithm and its variants in sub-pixel accurate disparity estimation.
3 0.75068617 28 iccv-2013-A Rotational Stereo Model Based on XSlit Imaging
Author: Jinwei Ye, Yu Ji, Jingyi Yu
Abstract: Traditional stereo matching assumes perspective viewing cameras under a translational motion: the second camera is translated away from the first one to create parallax. In this paper, we investigate a different, rotational stereo model on a special multi-perspective camera, the XSlit camera [9, 24]. We show that rotational XSlit (R-XSlit) stereo can be effectively created by fixing the sensor and slit locations but switching the two slits’ directions. We first derive the epipolar geometry of R-XSlit in the 4D light field ray space. Our derivation leads to a simple but effective scheme for locating corresponding epipolar “curves ”. To conduct stereo matching, we further derive a new disparity term in our model and develop a patch-based graph-cut solution. To validate our theory, we assemble an XSlit lens by using a pair of cylindrical lenses coupled with slit-shaped apertures. The XSlit lens can be mounted on commodity cameras where the slit directions are adjustable to form desirable R-XSlit pairs. We show through experiments that R-XSlitprovides apotentially advantageous imaging system for conducting fixed-location, dynamic baseline stereo.
4 0.67061138 88 iccv-2013-Constant Time Weighted Median Filtering for Stereo Matching and Beyond
Author: Ziyang Ma, Kaiming He, Yichen Wei, Jian Sun, Enhua Wu
Abstract: Despite the continuous advances in local stereo matching for years, most efforts are on developing robust cost computation and aggregation methods. Little attention has been seriously paid to the disparity refinement. In this work, we study weighted median filtering for disparity refinement. We discover that with this refinement, even the simple box filter aggregation achieves comparable accuracy with various sophisticated aggregation methods (with the same refinement). This is due to the nice weighted median filtering properties of removing outlier error while respecting edges/structures. This reveals that the previously overlooked refinement can be at least as crucial as aggregation. We also develop the first constant time algorithmfor the previously time-consuming weighted median filter. This makes the simple combination “box aggregation + weighted median ” an attractive solution in practice for both speed and accuracy. As a byproduct, the fast weighted median filtering unleashes its potential in other applications that were hampered by high complexities. We show its superiority in various applications such as depth upsampling, clip-art JPEG artifact removal, and image stylization.
5 0.65927678 252 iccv-2013-Line Assisted Light Field Triangulation and Stereo Matching
Author: Zhan Yu, Xinqing Guo, Haibing Lin, Andrew Lumsdaine, Jingyi Yu
Abstract: Light fields are image-based representations that use densely sampled rays as a scene description. In this paper, we explore geometric structures of 3D lines in ray space for improving light field triangulation and stereo matching. The triangulation problem aims to fill in the ray space with continuous and non-overlapping simplices anchored at sampled points (rays). Such a triangulation provides a piecewise-linear interpolant useful for light field superresolution. We show that the light field space is largely bilinear due to 3D line segments in the scene, and direct triangulation of these bilinear subspaces leads to large errors. We instead present a simple but effective algorithm to first map bilinear subspaces to line constraints and then apply Constrained Delaunay Triangulation (CDT). Based on our analysis, we further develop a novel line-assisted graphcut (LAGC) algorithm that effectively encodes 3D line constraints into light field stereo matching. Experiments on synthetic and real data show that both our triangulation and LAGC algorithms outperform state-of-the-art solutions in accuracy and visual quality.
6 0.5789904 382 iccv-2013-Semi-dense Visual Odometry for a Monocular Camera
7 0.49982432 225 iccv-2013-Joint Segmentation and Pose Tracking of Human in Natural Videos
8 0.48198906 316 iccv-2013-Pictorial Human Spaces: How Well Do Humans Perceive a 3D Articulated Pose?
9 0.47921818 46 iccv-2013-Allocentric Pose Estimation
10 0.46495023 118 iccv-2013-Discovering Object Functionality
11 0.46265283 143 iccv-2013-Estimating Human Pose with Flowing Puppets
12 0.45552945 226 iccv-2013-Joint Subspace Stabilization for Stereoscopic Video
13 0.44711816 273 iccv-2013-Monocular Image 3D Human Pose Estimation under Self-Occlusion
14 0.44371074 403 iccv-2013-Strong Appearance and Expressive Spatial Models for Human Pose Estimation
15 0.43299669 255 iccv-2013-Local Signal Equalization for Correspondence Matching
16 0.42990679 308 iccv-2013-Parsing IKEA Objects: Fine Pose Estimation
17 0.42702997 344 iccv-2013-Recognising Human-Object Interaction via Exemplar Based Modelling
18 0.42687073 130 iccv-2013-Dynamic Structured Model Selection
19 0.42509156 111 iccv-2013-Detecting Dynamic Objects with Multi-view Background Subtraction
20 0.42270735 218 iccv-2013-Interactive Markerless Articulated Hand Motion Tracking Using RGB and Depth Data
topicId topicWeight
[(2, 0.325), (4, 0.015), (7, 0.015), (26, 0.087), (31, 0.038), (35, 0.022), (40, 0.012), (42, 0.082), (64, 0.08), (73, 0.029), (89, 0.179), (95, 0.011)]
simIndex simValue paperId paperTitle
1 0.96888715 294 iccv-2013-Offline Mobile Instance Retrieval with a Small Memory Footprint
Author: Jayaguru Panda, Michael S. Brown, C.V. Jawahar
Abstract: Existing mobile image instance retrieval applications assume a network-based usage where image features are sent to a server to query an online visual database. In this scenario, there are no restrictions on the size of the visual database. This paper, however, examines how to perform this same task offline, where the entire visual index must reside on the mobile device itself within a small memory footprint. Such solutions have applications on location recognition and product recognition. Mobile instance retrieval requires a significant reduction in the visual index size. To achieve this, we describe a set of strategies that can reduce the visual index up to 60-80 compared to a scatannd raerddu iens tthaen vceis rueatrli ienvdaelx xim upple tom 6en0t-8at0io ×n found on ddte osk atops or servers. While our proposed reduction steps affect the overall mean Average Precision (mAP), they are able to maintain a good Precision for the top K results (PK). We argue that for such offline application, maintaining a good PK is sufficient. The effectiveness of this approach is demonstrated on several standard databases. A working application designed for a remote historical site is also presented. This application is able to reduce an 50,000 image index structure to 25 MBs while providing a precision of 97% for P10 and 100% for P1.
2 0.96740812 13 iccv-2013-A General Two-Step Approach to Learning-Based Hashing
Author: Guosheng Lin, Chunhua Shen, David Suter, Anton van_den_Hengel
Abstract: Most existing approaches to hashing apply a single form of hash function, and an optimization process which is typically deeply coupled to this specific form. This tight coupling restricts the flexibility of the method to respond to the data, and can result in complex optimization problems that are difficult to solve. Here we propose a flexible yet simple framework that is able to accommodate different types of loss functions and hash functions. This framework allows a number of existing approaches to hashing to be placed in context, and simplifies the development of new problemspecific hashing methods. Our framework decomposes the hashing learning problem into two steps: hash bit learning and hash function learning based on the learned bits. The first step can typically be formulated as binary quadratic problems, and the second step can be accomplished by training standard binary classifiers. Both problems have been extensively studied in the literature. Our extensive experiments demonstrate that the proposed framework is effective, flexible and outperforms the state-of-the-art.
3 0.95491552 446 iccv-2013-Visual Semantic Complex Network for Web Images
Author: Shi Qiu, Xiaogang Wang, Xiaoou Tang
Abstract: This paper proposes modeling the complex web image collections with an automatically generated graph structure called visual semantic complex network (VSCN). The nodes on this complex network are clusters of images with both visual and semantic consistency, called semantic concepts. These nodes are connected based on the visual and semantic correlations. Our VSCN with 33, 240 concepts is generated from a collection of 10 million web images. 1 A great deal of valuable information on the structures of the web image collections can be revealed by exploring the VSCN, such as the small-world behavior, concept community, indegree distribution, hubs, and isolated concepts. It not only helps us better understand the web image collections at a macroscopic level, but also has many important practical applications. This paper presents two application examples: content-based image retrieval and image browsing. Experimental results show that the VSCN leads to significant improvement on both the precision of image retrieval (over 200%) and user experience for image browsing.
4 0.95095968 191 iccv-2013-Handling Uncertain Tags in Visual Recognition
Author: Arash Vahdat, Greg Mori
Abstract: Gathering accurate training data for recognizing a set of attributes or tags on images or videos is a challenge. Obtaining labels via manual effort or from weakly-supervised data typically results in noisy training labels. We develop the FlipSVM, a novel algorithm for handling these noisy, structured labels. The FlipSVM models label noise by “flipping ” labels on training examples. We show empirically that the FlipSVM is effective on images-and-attributes and video tagging datasets.
5 0.95036054 352 iccv-2013-Revisiting Example Dependent Cost-Sensitive Learning with Decision Trees
Author: Oisin Mac Aodha, Gabriel J. Brostow
Abstract: Typical approaches to classification treat class labels as disjoint. For each training example, it is assumed that there is only one class label that correctly describes it, and that all other labels are equally bad. We know however, that good and bad labels are too simplistic in many scenarios, hurting accuracy. In the realm of example dependent costsensitive learning, each label is instead a vector representing a data point’s affinity for each of the classes. At test time, our goal is not to minimize the misclassification rate, but to maximize that affinity. We propose a novel example dependent cost-sensitive impurity measure for decision trees. Our experiments show that this new impurity measure improves test performance while still retaining the fast test times of standard classification trees. We compare our approach to classification trees and other cost-sensitive methods on three computer vision problems, tracking, descriptor matching, and optical flow, and show improvements in all three domains.
6 0.94833422 244 iccv-2013-Learning View-Invariant Sparse Representations for Cross-View Action Recognition
7 0.93462515 214 iccv-2013-Improving Graph Matching via Density Maximization
8 0.92732143 374 iccv-2013-Salient Region Detection by UFO: Uniqueness, Focusness and Objectness
9 0.88787103 239 iccv-2013-Learning Hash Codes with Listwise Supervision
same-paper 10 0.88379419 322 iccv-2013-Pose Estimation and Segmentation of People in 3D Movies
11 0.88316423 153 iccv-2013-Face Recognition Using Face Patch Networks
12 0.8645314 229 iccv-2013-Large-Scale Video Hashing via Structure Learning
13 0.86299157 409 iccv-2013-Supervised Binary Hash Code Learning with Jensen Shannon Divergence
14 0.86164784 83 iccv-2013-Complementary Projection Hashing
15 0.83521688 313 iccv-2013-Person Re-identification by Salience Matching
16 0.83354449 248 iccv-2013-Learning to Rank Using Privileged Information
17 0.83127898 73 iccv-2013-Class-Specific Simplex-Latent Dirichlet Allocation for Image Classification
18 0.82121313 406 iccv-2013-Style-Aware Mid-level Representation for Discovering Visual Connections in Space and Time
19 0.81979591 443 iccv-2013-Video Synopsis by Heterogeneous Multi-source Correlation
20 0.81842238 448 iccv-2013-Weakly Supervised Learning of Image Partitioning Using Decision Trees with Structured Split Criteria