cvpr cvpr2013 cvpr2013-226 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Tony Tung, Takashi Matsuyama
Abstract: This paper presents a novel approach to characterize deformable surface using intrinsic property dynamics. 3D dynamic surfaces representing humans in motion can be obtained using multiple view stereo reconstruction methods or depth cameras. Nowadays these technologies have become capable to capture surface variations in real-time, and give details such as clothing wrinkles and deformations. Assuming repetitive patterns in the deformations, we propose to model complex surface variations using sets of linear dynamical systems (LDS) where observations across time are given by surface intrinsic properties such as local curvatures. We introduce an approach based on bags of dynamical systems, where each surface feature to be represented in the codebook is modeled by a set of LDS equipped with timing structure. Experiments are performed on datasets of real-world dynamical surfaces and show compelling results for description, classification and segmentation.
Reference: text
sentIndex sentText sentNum sentScore
1 on Abstract This paper presents a novel approach to characterize deformable surface using intrinsic property dynamics. [sent-2, score-0.518]
2 3D dynamic surfaces representing humans in motion can be obtained using multiple view stereo reconstruction methods or depth cameras. [sent-3, score-0.381]
3 Nowadays these technologies have become capable to capture surface variations in real-time, and give details such as clothing wrinkles and deformations. [sent-4, score-0.675]
4 Assuming repetitive patterns in the deformations, we propose to model complex surface variations using sets of linear dynamical systems (LDS) where observations across time are given by surface intrinsic properties such as local curvatures. [sent-5, score-1.177]
5 We introduce an approach based on bags of dynamical systems, where each surface feature to be represented in the codebook is modeled by a set of LDS equipped with timing structure. [sent-6, score-0.827]
6 Experiments are performed on datasets of real-world dynamical surfaces and show compelling results for description, classification and segmentation. [sent-7, score-0.353]
7 Nowadays, advances in visual sensing systems (for color and depth) allow us to capture smaller variations and details on object surfaces in real-time (i. [sent-13, score-0.317]
8 For example, techniques such as performance capture or 3D video [27, 17, 34, 11, 20] can return complete and accurate 3D dynamic surface models, reconstructed by multiview stereo (MVS) methods or fusion of depth maps. [sent-16, score-0.72]
9 Here, we propose to characterize, classify and segment dynamic deformable surfaces using surface intrinsic property dynamics (see Fig. [sent-19, score-0.93]
10 Dynamic surface characterization from intrinsic property extraction, tracking, and dynamics modeling across time. [sent-28, score-0.673]
11 Here, we show local curvatures estimated at each surface point, and curvature maps obtained after mapping on square domain. [sent-29, score-0.646]
12 , colored squares) can be classified using curvature dynamics modeled by sets of dynamical systems {LDS}. [sent-32, score-0.451]
13 For example, clothing made of soft fabrics worn by a human in motion usually exhibit more surface variations that bare skin. [sent-35, score-0.606]
14 Unfortunately, to capture those complex variations one cannot (only) rely on visual appearance-based methods [13, 7, 30], as surface texture of complete 3D surface models can be poor (e. [sent-36, score-0.938]
15 Hence, we propose to characterize dynamic surfaces using a 222333333 geometry-dynamics-based approach that relies on intrinsic surface properties as follows: (1) surfaces are first aligned in order to locate and track surface feature (e. [sent-43, score-1.408]
16 Section 3 presents the extraction dynamic surface intrinsic feature. [sent-49, score-0.676]
17 Section 5 describes surface dynamics modeling using bags of dynamical systems. [sent-51, score-0.754]
18 Related work Complete reconstruction of dynamic surfaces is an active research area due to the numerous potential applications: medicine, sports, entertainment, digital archiving, etc. [sent-55, score-0.381]
19 The resulting performance capture or 3D video consists of a stream of textured surface mesh models undergoing free-form deformation. [sent-65, score-0.562]
20 However, recent efforts have been done to produce consistent sequences by 3D scene flow estimation, surface matching or tracking [39, 33, 12, 38, 41, 4, 28, 35, 16, 3]. [sent-67, score-0.511]
21 Nevertheless, photometric feature matching approaches require surface models with good texture and color consistency between the multiple capture viewpoints, and across time. [sent-68, score-0.486]
22 Hence, most appearance-based methods are not able to accurately track true deformations of low-frequency surface details (e. [sent-69, score-0.44]
23 We propose to model complex surface variations using linear dynamical systems (LDS). [sent-73, score-0.699]
24 Dynamic surfaces reconstructed from multiview stereo methods (MVS) [2]: surface and processed surface (curvatures). [sent-76, score-0.982]
25 lar, dynamical models have been applied in computer vision for dynamic texture modeling [32, 13], recognition[32, 30], and segmentation[14, 7, 40]. [sent-77, score-0.439]
26 In [30], the authors propose to tackle challenging scenarios and model dynamic textures with a collection of LDS, by following the bag of features (BoF) approach, where a LDS is associated to a spatiotemporal volume obtained by tracking a feature point. [sent-79, score-0.325]
27 However, in the context of dynamic surfaces from human performance, the nature of deformations can be heterogeneous in time, and therefore requires several LDS for modeling. [sent-80, score-0.423]
28 Thus, dynamic surfaces can be segmented into patches that are classified into regions corresponding to each body parts (e. [sent-82, score-0.397]
29 Dynamic surface feature extraction This section presents surface intrinsic feature extraction from surface points which are tracked across time. [sent-88, score-1.282]
30 In particular, we estimate local curvatures as features using a continuous surface shape index. [sent-89, score-0.585]
31 Surface intrinsic characterization To perform surface intrinsic characterization, we propose to represent local curvatures by computing the Koenderink shape index for each surface point, as it is known to be more stable for natural scenes than a classification by Gaussian and mean curvatures [25]. [sent-92, score-1.34]
32 The differential structure of a surface can be captured by the local Hessian matrix H, which is computed using surtfahece l noocarml Haless: H =⎛⎝−−? [sent-94, score-0.402]
33 Local curvature variation across time: a) from surface alignment [4], b) from MVS ground truth [34] and c) from alignment and correction. [sent-103, score-0.497]
34 where n is a surface normal, and which eigenvalues are the principal curvatures κ1 and κ2 (κ1 ≥ κ2). [sent-105, score-0.551]
35 For all surface point, the shape index σ describes loc≥al κ κsurface topology in terms of the principal curvatures: σ = 2 π arctan κ2+ κ1. [sent-106, score-0.443]
36 Local curvatures computed on different surface meshes are shown in Fig. [sent-108, score-0.591]
37 Intrinsic feature tracking One challenge to overcome when characterizing surface dynamics is surface alignment for surface point tracking. [sent-113, score-1.351]
38 2, methods involving color information cannot be used for that purpose as surfaces from performance capture (or 3D video data) usually suffer from color inconsistency or poorly textured regions. [sent-115, score-0.232]
39 As well, methods which are too sensitive to surface deformation or topology change can produce inaccurate results. [sent-116, score-0.443]
40 Here, we propose to use [4] to perform surface alignments independently from color information and topology change. [sent-117, score-0.443]
41 Nevertheless, while the global surface geometry is correctly deformed and aligned across time, the patch-based approach does not preserve intrinsic information such as local curvatures. [sent-118, score-0.478]
42 Hence, we propose to register original surface meshes (with computed local curvature information at full resolution) to sequences aligned as in [4], and correct local curvature with exact values for each mesh vertex on the latter ones by assigning the nearest neighbor values. [sent-119, score-0.778]
43 Actually, 3D video sequences obtained from MVS usually contain surface noise. [sent-120, score-0.511]
44 However, as the reconstruction is performed frame-by-frame they can still be a good approximation of ground truth surface as no noise is propagated through the sequence, as opposed to spatiotemporal reconstruction. [sent-121, score-0.489]
45 Figure 3 shows local curvatures computed on surface mesh models across time. [sent-122, score-0.624]
46 Curvature maps obtained after surface alignment and mapping on square parametrization domain [3 1] are given for visualization purpose. [sent-123, score-0.402]
47 Note that recently in [36], the authors have proposed an invariant surface descriptor that could potentially be used for surface alignment and surface point tracking. [sent-124, score-1.206]
48 Dynamic surface modeling using LDS When representing dynamic surfaces as curvature maps (see Fig. [sent-126, score-0.84]
49 1 and 3), analogy can be made with dynamic textures [32]. [sent-127, score-0.274]
50 However, surfaces from performance capture can exhibit heterogeneous deformations in time (see Sect. [sent-128, score-0.318]
51 Hence we model surface dynamics using hybrid linear dynamical systems (hybrid LDS) that can describe both continuous and discrete events. [sent-130, score-0.877]
52 Hybrid linear dynamical system Assuming a temporal Y = {y(t)}t≥0, y(t) ∈ {x(t)}t≥0, }x(t) y∈( R) n∈ space, a lin,ea xr( dynamical ? [sent-139, score-0.449]
53 , for dynamic texture segmentation [7], and facial movement recognition [24]). [sent-149, score-0.231]
54 a) Observed signals from surface patch #274 in Bouncing sequence [4] (torso region). [sent-169, score-0.505]
55 Dynamic surface characterization Bag-of-features (BoF) have been successfully applied to various visual classification tasks thanks to their ability to capture invariance aspects of local features [26, 21, 30]. [sent-176, score-0.539]
56 In [30], the authors introduce the bags of dynamical systems (BoS) for dynamic texture recognition and outperforms [32]. [sent-177, score-0.513]
57 Here, we propose to apply the BoS framework to characterize dynamic surfaces. [sent-178, score-0.238]
58 Moreover, each surface patch is modeled by a set of N LDS (as opposed to only one per video feature in prior work). [sent-179, score-0.477]
59 As well, we introduce timing structure information given by the hybrid LDS model in the codebook formation of BoS. [sent-180, score-0.267]
60 Here, our features are sets of LDS parameters (extracted from surface patches) belonging to a non-Euclidean space: (Ai, C) ∈ GL(n) ST(m, n) (see Sect. [sent-184, score-0.402]
61 In our framework, as each surface region is represented by a limited number of LDS features {(Ai, C)}, T remains relatively esdm nalulm. [sent-211, score-0.402]
62 We propose treop use esontfts- twheeig ohbtjiencgt as igt i,s f loers sc lsaesnsisifiticvaeto noise [21, 30], and we introduce timing structure information given by the hybrid LDS modeling into the weighting scheme using the term frequency ρj : wk=? [sent-264, score-0.231]
63 If β = 0, then we lose the timing structure characterizing the duration of each state of the LDS in the model. [sent-276, score-0.226]
64 Every mesh has reasonable resolution and quality which allow us to capture local surface variations across 222333777 Figure 5. [sent-292, score-0.576]
65 Our method using BoS with SW and timing structure and SVM returns the best performance compare to the state-of-the-art techniques used for dynamic texture recognition. [sent-294, score-0.405]
66 However, using intrinsic characterization, we could observe that reconstructed surfaces from [34] (such as Free) are still very noisy at small scale due to drawbacks from MVS, despite being visually very compelling. [sent-296, score-0.254]
67 On the other hand, surfaces from [11] and [2] (such as Samba and Bouncing respectively) contain drawbacks for spatiotemporal reconstruction, and therefore reconstructed surfaces of clothing might eventually be more rigid and less prone to wrinkle. [sent-297, score-0.494]
68 However all of these dynamic surfaces could exhibit temporal statistics in different regions or body parts. [sent-298, score-0.418]
69 In what follows, we particularly focus on rigid and non-rigid regions as for the time being, it is still difficult to characterize different surface materials based only on surface dynamics. [sent-299, score-0.886]
70 However, the top of the dress is tight and its surface does not exhibit that much variations. [sent-301, score-0.485]
71 However, it is challenging to distinguish the T-shirt from the pants using surface dynamics as variations are similar (and may have been kept rigid and smooth by the reconstruction process). [sent-303, score-0.673]
72 4, dynamic surfaces can be treated as dynamic textures although surface variations from performance capture can be unpredictable or heterogeneous in time. [sent-306, score-1.128]
73 As no prior work on surface dynamics characterization as been proposed in the computer vision literature, we propose to use state-of-theart approaches related to dynamic texture recognition as baseline for comparison. [sent-307, score-0.828]
74 In [8, 32], the authors use a single LDS to model a video sequence (of dynamic textures), the Martin distance is used to calculate distances between LDS, and NN and SVM are used for classification. [sent-308, score-0.267]
75 As we deal with dynamic surfaces representing continuous human performances, we expect to characterize repetitive patterns that can be found in surface deformations. [sent-313, score-0.819]
76 Challenges come from surface noise and irregular (repetition of) patterns, as subjects repeat or perform various actions in a same sequence. [sent-314, score-0.402]
77 We could observe that even rigid surface regions such as bare skin or faces exhibit some variations. [sent-315, score-0.547]
78 First, we propose to classify dynamic surface patches from different sequences into rigid and non-rigid classes. [sent-316, score-0.769]
79 Furthermore, to obtain ground truth classification, we manually labeled each patch and assigned them to a surface region (i. [sent-318, score-0.441]
80 Although the performances are different, the surface dynamics of rigid and non-rigid regions show same characteristics. [sent-337, score-0.587]
81 Here, all surface patches from all the test datasets from the Free sequence were tested for classification using patches from the Samba sequence as training data. [sent-340, score-0.576]
82 This is primarily because dynamic surfaces for performance capture can exhibit various behaviors in time that cannot be well modeled using a single LDS per feature. [sent-342, score-0.436]
83 Surface region characterization allows body part segmentation, as skin and clothing can potentially be identified using surface patch dynamics. [sent-352, score-0.636]
84 Conclusion As 3D reconstruction technologies have become capable to capture surface deformations in real-time and details such as clothing wrinkles, dynamic surfaces representing human performance can now be characterized using local geometry information. [sent-355, score-1.001]
85 Moreover, assuming dynamic sur- faces as streams of temporally continuous and indefinitely varying data having certain temporal statistics, we can draw an analogy with the dynamic textures. [sent-356, score-0.536]
86 Hence in this paper, we present the following contributions: 1) no prior work has addressed dynamic surface characterization using surface intrinsic properties (such as local curvatures), 2) we propose to model surface dynamics using hybrid linear dynamical system models (i. [sent-357, score-1.968]
87 , with N LDS per surface feature) within the bag of dynamical systems (BoS) framework, and 3) we introduce LDS timing structure in the codebook formation of the BoS. [sent-359, score-0.831]
88 We show experimental results on datasets of real-world dynamical surfaces for description, classification and segmentation. [sent-360, score-0.353]
89 Appendix: Surface patches Figure 6 shows surface patches for the models from the sequences Free, Samba and Bouncing, computed as in [4]. [sent-369, score-0.583]
90 3, as it is a good representation to understand the analogy with dynamic textures. [sent-372, score-0.232]
91 However, we recall that curvature maps are used only for visualization and assessment of surface point tracking, and not for the tracking itself (as surfaces are cut and geometry is distorted). [sent-373, score-0.678]
92 Surface patches for the models from the sequences Free, Samba and Bouncing and projections on square domain given for visualization and assessment of surface point tracking. [sent-375, score-0.529]
93 Group action induced distances for averaging and clustering 222333999 [2] [3] [4] [5] [6] [7] [8] [9] [10] linear dynamical systems with applications to the analysis of dynamic scenes. [sent-381, score-0.478]
94 Histograms of oriented optical flow and binet-cauchy kernels on nonlinear dynamical systems for the recognition of human actions. [sent-434, score-0.247]
95 3d reconstruction of dynamic scenes with multiple handheld cameras. [sent-512, score-0.236]
96 Interval-based modeling of human communication dynamics via hybrid dynamical systems. [sent-540, score-0.402]
97 Real-time 3d shape reconstruction, dynamic 3d mesh deformation, and high fidelity visualization for 3d video. [sent-558, score-0.271]
98 View-invariant dynamic texture recognition using a bag of dynamical systems. [sent-579, score-0.439]
99 Planar parameterization [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] of triangulated surface meshes. [sent-585, score-0.402]
100 Dynamic surface matching by geodesic mapping for 3d animation transfer. [sent-609, score-0.402]
wordName wordTfidf (topN-words)
[('lds', 0.54), ('surface', 0.402), ('dynamical', 0.208), ('dynamic', 0.198), ('samba', 0.193), ('curvatures', 0.149), ('timing', 0.146), ('surfaces', 0.145), ('bouncing', 0.143), ('bos', 0.137), ('nlds', 0.135), ('dynamics', 0.109), ('sw', 0.101), ('curvature', 0.095), ('characterization', 0.086), ('hybrid', 0.085), ('clothing', 0.08), ('tung', 0.079), ('intrinsic', 0.076), ('free', 0.075), ('mesh', 0.073), ('sequences', 0.073), ('mvs', 0.068), ('xj', 0.067), ('boyer', 0.066), ('ki', 0.066), ('svm', 0.059), ('fsm', 0.058), ('swtf', 0.058), ('patches', 0.054), ('markerless', 0.051), ('capture', 0.051), ('codewords', 0.05), ('variations', 0.05), ('bof', 0.05), ('spatiotemporal', 0.049), ('technologies', 0.049), ('ravichandran', 0.045), ('duration', 0.044), ('dm', 0.043), ('chaudhry', 0.043), ('wrinkles', 0.043), ('textures', 0.042), ('exhibit', 0.042), ('rigid', 0.042), ('heterogeneous', 0.042), ('dress', 0.041), ('topology', 0.041), ('martin', 0.041), ('meshes', 0.04), ('characterize', 0.04), ('doretto', 0.04), ('systems', 0.039), ('patch', 0.039), ('indefinitely', 0.039), ('kawashima', 0.039), ('matsuyama', 0.039), ('medoid', 0.039), ('reconstruction', 0.038), ('deformations', 0.038), ('video', 0.036), ('nn', 0.036), ('tracking', 0.036), ('codebook', 0.036), ('state', 0.036), ('bags', 0.035), ('qi', 0.035), ('gl', 0.034), ('analogy', 0.034), ('performances', 0.034), ('continuous', 0.034), ('starck', 0.034), ('cgal', 0.034), ('medoids', 0.034), ('varanasi', 0.034), ('reconstructed', 0.033), ('sequence', 0.033), ('clustering', 0.033), ('temporal', 0.033), ('texture', 0.033), ('sensing', 0.032), ('rut', 0.032), ('pants', 0.032), ('bare', 0.032), ('confusions', 0.032), ('franco', 0.032), ('koenderink', 0.032), ('stoll', 0.032), ('zaharescu', 0.032), ('signals', 0.031), ('sk', 0.031), ('chan', 0.03), ('organs', 0.03), ('aguiar', 0.03), ('wears', 0.03), ('forearm', 0.03), ('skin', 0.029), ('kyoto', 0.028), ('returns', 0.028), ('torso', 0.028)]
simIndex simValue paperId paperTitle
same-paper 1 1.0000001 226 cvpr-2013-Intrinsic Characterization of Dynamic Surfaces
Author: Tony Tung, Takashi Matsuyama
Abstract: This paper presents a novel approach to characterize deformable surface using intrinsic property dynamics. 3D dynamic surfaces representing humans in motion can be obtained using multiple view stereo reconstruction methods or depth cameras. Nowadays these technologies have become capable to capture surface variations in real-time, and give details such as clothing wrinkles and deformations. Assuming repetitive patterns in the deformations, we propose to model complex surface variations using sets of linear dynamical systems (LDS) where observations across time are given by surface intrinsic properties such as local curvatures. We introduce an approach based on bags of dynamical systems, where each surface feature to be represented in the codebook is modeled by a set of LDS equipped with timing structure. Experiments are performed on datasets of real-world dynamical surfaces and show compelling results for description, classification and segmentation.
2 0.18386717 298 cvpr-2013-Multi-scale Curve Detection on Surfaces
Author: Michael Kolomenkin, Ilan Shimshoni, Ayellet Tal
Abstract: This paper extends to surfaces the multi-scale approach of edge detection on images. The common practice for detecting curves on surfaces requires the user to first select the scale of the features, apply an appropriate smoothing, and detect the edges on the smoothed surface. This approach suffers from two drawbacks. First, it relies on a hidden assumption that all the features on the surface are of the same scale. Second, manual user intervention is required. In this paper, we propose a general framework for automatically detecting the optimal scale for each point on the surface. We smooth the surface at each point according to this optimal scale and run the curve detection algorithm on the resulting surface. Our multi-scale algorithm solves the two disadvantages of the single-scale approach mentioned above. We demonstrate how to realize our approach on two commonly-used special cases: ridges & valleys and relief edges. In each case, the optimal scale is found in accordance with the mathematical definition of the curve.
3 0.16886249 423 cvpr-2013-Template-Based Isometric Deformable 3D Reconstruction with Sampling-Based Focal Length Self-Calibration
Author: Adrien Bartoli, Toby Collins
Abstract: It has been shown that a surface deforming isometrically can be reconstructed from a single image and a template 3D shape. Methods from the literature solve this problem efficiently. However, they all assume that the camera model is calibrated, which drastically limits their applicability. We propose (i) a general variational framework that applies to (calibrated and uncalibrated) general camera models and (ii) self-calibrating 3D reconstruction algorithms for the weak-perspective and full-perspective camera models. In the former case, our algorithm returns the normal field and camera ’s scale factor. In the latter case, our algorithm returns the normal field, depth and camera ’s focal length. Our algorithms are the first to achieve deformable 3D reconstruction including camera self-calibration. They apply to much more general setups than existing methods. Experimental results on simulated and real data show that our algorithms give results with the same level of accuracy as existing methods (which use the true focal length) on perspective images, and correctly find the normal field on affine images for which the existing methods fail.
Author: Won Hwa Kim, Moo K. Chung, Vikas Singh
Abstract: The analysis of 3-D shape meshes is a fundamental problem in computer vision, graphics, and medical imaging. Frequently, the needs of the application require that our analysis take a multi-resolution view of the shape ’s local and global topology, and that the solution is consistent across multiple scales. Unfortunately, the preferred mathematical construct which offers this behavior in classical image/signal processing, Wavelets, is no longer applicable in this general setting (data with non-uniform topology). In particular, the traditional definition does not allow writing out an expansion for graphs that do not correspond to the uniformly sampled lattice (e.g., images). In this paper, we adapt recent results in harmonic analysis, to derive NonEuclidean Wavelets based algorithms for a range of shape analysis problems in vision and medical imaging. We show how descriptors derived from the dual domain representation offer native multi-resolution behavior for characterizing local/global topology around vertices. With only minor modifications, the framework yields a method for extracting interest/key points from shapes, a surprisingly simple algorithm for 3-D shape segmentation (competitive with state of the art), and a method for surface alignment (without landmarks). We give an extensive set of comparison results on a large shape segmentation benchmark and derive a uniqueness theorem for the surface alignment problem.
5 0.14033861 289 cvpr-2013-Monocular Template-Based 3D Reconstruction of Extensible Surfaces with Local Linear Elasticity
Author: Abed Malti, Richard Hartley, Adrien Bartoli, Jae-Hak Kim
Abstract: We propose a new approach for template-based extensible surface reconstruction from a single view. We extend the method of isometric surface reconstruction and more recent work on conformal surface reconstruction. Our approach relies on the minimization of a proposed stretching energy formalized with respect to the Poisson ratio parameter of the surface. We derive a patch-based formulation of this stretching energy by assuming local linear elasticity. This formulation unifies geometrical and mechanical constraints in a single energy term. We prevent local scale ambiguities by imposing a set of fixed boundary 3D points. We experimentally prove the sufficiency of this set of boundary points and demonstrate the effectiveness of our approach on different developable and non-developable surfaces with a wide range of extensibility.
6 0.13449708 208 cvpr-2013-Hyperbolic Harmonic Mapping for Constrained Brain Surface Registration
7 0.13321914 465 cvpr-2013-What Object Motion Reveals about Shape with Unknown BRDF and Lighting
8 0.13062698 443 cvpr-2013-Uncalibrated Photometric Stereo for Unknown Isotropic Reflectances
9 0.11593992 286 cvpr-2013-Mirror Surface Reconstruction from a Single Image
10 0.11167229 196 cvpr-2013-HON4D: Histogram of Oriented 4D Normals for Activity Recognition from Depth Sequences
11 0.10770375 284 cvpr-2013-Mesh Based Semantic Modelling for Indoor and Outdoor Scenes
12 0.10659896 303 cvpr-2013-Multi-view Photometric Stereo with Spatially Varying Isotropic Materials
13 0.10604383 97 cvpr-2013-Correspondence-Less Non-rigid Registration of Triangular Surface Meshes
14 0.10518946 230 cvpr-2013-Joint 3D Scene Reconstruction and Class Segmentation
15 0.10149407 237 cvpr-2013-Kernel Learning for Extrinsic Classification of Manifold Features
16 0.098570041 348 cvpr-2013-Recognizing Activities via Bag of Words for Attribute Dynamics
17 0.098439321 327 cvpr-2013-Pattern-Driven Colorization of 3D Surfaces
18 0.096071079 44 cvpr-2013-Area Preserving Brain Mapping
19 0.093959987 331 cvpr-2013-Physically Plausible 3D Scene Tracking: The Single Actor Hypothesis
20 0.093613543 316 cvpr-2013-Optical Flow Estimation Using Laplacian Mesh Energy
topicId topicWeight
[(0, 0.19), (1, 0.112), (2, -0.011), (3, 0.002), (4, -0.046), (5, -0.103), (6, -0.086), (7, -0.016), (8, 0.009), (9, -0.021), (10, 0.012), (11, -0.054), (12, -0.125), (13, -0.053), (14, 0.1), (15, -0.028), (16, 0.071), (17, 0.042), (18, 0.036), (19, 0.056), (20, -0.088), (21, -0.011), (22, 0.059), (23, 0.031), (24, -0.194), (25, 0.037), (26, 0.008), (27, 0.002), (28, -0.03), (29, -0.071), (30, 0.047), (31, 0.007), (32, 0.054), (33, 0.041), (34, 0.015), (35, 0.008), (36, -0.017), (37, -0.103), (38, -0.031), (39, -0.04), (40, 0.086), (41, -0.011), (42, 0.047), (43, 0.084), (44, 0.007), (45, 0.061), (46, -0.013), (47, -0.009), (48, 0.072), (49, 0.006)]
simIndex simValue paperId paperTitle
same-paper 1 0.9673872 226 cvpr-2013-Intrinsic Characterization of Dynamic Surfaces
Author: Tony Tung, Takashi Matsuyama
Abstract: This paper presents a novel approach to characterize deformable surface using intrinsic property dynamics. 3D dynamic surfaces representing humans in motion can be obtained using multiple view stereo reconstruction methods or depth cameras. Nowadays these technologies have become capable to capture surface variations in real-time, and give details such as clothing wrinkles and deformations. Assuming repetitive patterns in the deformations, we propose to model complex surface variations using sets of linear dynamical systems (LDS) where observations across time are given by surface intrinsic properties such as local curvatures. We introduce an approach based on bags of dynamical systems, where each surface feature to be represented in the codebook is modeled by a set of LDS equipped with timing structure. Experiments are performed on datasets of real-world dynamical surfaces and show compelling results for description, classification and segmentation.
2 0.87479007 289 cvpr-2013-Monocular Template-Based 3D Reconstruction of Extensible Surfaces with Local Linear Elasticity
Author: Abed Malti, Richard Hartley, Adrien Bartoli, Jae-Hak Kim
Abstract: We propose a new approach for template-based extensible surface reconstruction from a single view. We extend the method of isometric surface reconstruction and more recent work on conformal surface reconstruction. Our approach relies on the minimization of a proposed stretching energy formalized with respect to the Poisson ratio parameter of the surface. We derive a patch-based formulation of this stretching energy by assuming local linear elasticity. This formulation unifies geometrical and mechanical constraints in a single energy term. We prevent local scale ambiguities by imposing a set of fixed boundary 3D points. We experimentally prove the sufficiency of this set of boundary points and demonstrate the effectiveness of our approach on different developable and non-developable surfaces with a wide range of extensibility.
3 0.83992124 44 cvpr-2013-Area Preserving Brain Mapping
Author: Zhengyu Su, Wei Zeng, Rui Shi, Yalin Wang, Jian Sun, Xianfeng Gu
Abstract: Brain mapping transforms the brain cortical surface to canonical planar domains, which plays a fundamental role in morphological study. Most existing brain mapping methods are based on angle preserving maps, which may introduce large area distortions. This work proposes an area preserving brain mapping method based on MongeBrenier theory. The brain mapping is intrinsic to the Riemannian metric, unique, and diffeomorphic. The computation is equivalent to convex energy minimization and power Voronoi diagram construction. Comparing to the existing approaches based on Monge-Kantorovich theory, the proposed one greatly reduces the complexity (from n2 unknowns to n ), and improves the simplicity and efficiency. Experimental results on caudate nucleus surface mapping and cortical surface mapping demonstrate the efficacy and efficiency of the proposed method. Conventional methods for caudate nucleus surface mapping may suffer from numerical instability; in contrast, current method produces diffeomorpic mappings stably. In the study of cortical sur- face classification for recognition of Alzheimer’s Disease, the proposed method outperforms some other morphometry features.
4 0.83590591 208 cvpr-2013-Hyperbolic Harmonic Mapping for Constrained Brain Surface Registration
Author: Rui Shi, Wei Zeng, Zhengyu Su, Hanna Damasio, Zhonglin Lu, Yalin Wang, Shing-Tung Yau, Xianfeng Gu
Abstract: Automatic computation of surface correspondence via harmonic map is an active research field in computer vision, computer graphics and computational geometry. It may help document and understand physical and biological phenomena and also has broad applications in biometrics, medical imaging and motion capture. Although numerous studies have been devoted to harmonic map research, limited progress has been made to compute a diffeomorphic harmonic map on general topology surfaces with landmark constraints. This work conquer this problem by changing the Riemannian metric on the target surface to a hyperbolic metric, so that the harmonic mapping is guaranteed to be a diffeomorphism under landmark constraints. The computational algorithms are based on the Ricci flow method and the method is general and robust. We apply our algorithm to study constrained human brain surface registration problem. Experimental results demonstrate that, by changing the Riemannian metric, the registrations are always diffeomorphic, and achieve relative high performance when evaluated with some popular cortical surface registration evaluation standards.
5 0.81486636 298 cvpr-2013-Multi-scale Curve Detection on Surfaces
Author: Michael Kolomenkin, Ilan Shimshoni, Ayellet Tal
Abstract: This paper extends to surfaces the multi-scale approach of edge detection on images. The common practice for detecting curves on surfaces requires the user to first select the scale of the features, apply an appropriate smoothing, and detect the edges on the smoothed surface. This approach suffers from two drawbacks. First, it relies on a hidden assumption that all the features on the surface are of the same scale. Second, manual user intervention is required. In this paper, we propose a general framework for automatically detecting the optimal scale for each point on the surface. We smooth the surface at each point according to this optimal scale and run the curve detection algorithm on the resulting surface. Our multi-scale algorithm solves the two disadvantages of the single-scale approach mentioned above. We demonstrate how to realize our approach on two commonly-used special cases: ridges & valleys and relief edges. In each case, the optimal scale is found in accordance with the mathematical definition of the curve.
6 0.78749746 435 cvpr-2013-Towards Contactless, Low-Cost and Accurate 3D Fingerprint Identification
8 0.70643669 97 cvpr-2013-Correspondence-Less Non-rigid Registration of Triangular Surface Meshes
9 0.69817209 141 cvpr-2013-Efficient Computation of Shortest Path-Concavity for 3D Meshes
10 0.63950139 423 cvpr-2013-Template-Based Isometric Deformable 3D Reconstruction with Sampling-Based Focal Length Self-Calibration
11 0.6348989 327 cvpr-2013-Pattern-Driven Colorization of 3D Surfaces
12 0.62472242 432 cvpr-2013-Three-Dimensional Bilateral Symmetry Plane Estimation in the Phase Domain
13 0.6107204 465 cvpr-2013-What Object Motion Reveals about Shape with Unknown BRDF and Lighting
14 0.5933075 52 cvpr-2013-Axially Symmetric 3D Pots Configuration System Using Axis of Symmetry and Break Curve
15 0.5847528 443 cvpr-2013-Uncalibrated Photometric Stereo for Unknown Isotropic Reflectances
16 0.56559074 286 cvpr-2013-Mirror Surface Reconstruction from a Single Image
17 0.53829682 316 cvpr-2013-Optical Flow Estimation Using Laplacian Mesh Energy
18 0.5381459 218 cvpr-2013-Improving the Visual Comprehension of Point Sets
19 0.53761256 391 cvpr-2013-Sensing and Recognizing Surface Textures Using a GelSight Sensor
20 0.5372085 303 cvpr-2013-Multi-view Photometric Stereo with Spatially Varying Isotropic Materials
topicId topicWeight
[(3, 0.234), (10, 0.092), (16, 0.03), (26, 0.047), (33, 0.256), (36, 0.019), (67, 0.045), (69, 0.057), (80, 0.018), (87, 0.112)]
simIndex simValue paperId paperTitle
1 0.89468884 289 cvpr-2013-Monocular Template-Based 3D Reconstruction of Extensible Surfaces with Local Linear Elasticity
Author: Abed Malti, Richard Hartley, Adrien Bartoli, Jae-Hak Kim
Abstract: We propose a new approach for template-based extensible surface reconstruction from a single view. We extend the method of isometric surface reconstruction and more recent work on conformal surface reconstruction. Our approach relies on the minimization of a proposed stretching energy formalized with respect to the Poisson ratio parameter of the surface. We derive a patch-based formulation of this stretching energy by assuming local linear elasticity. This formulation unifies geometrical and mechanical constraints in a single energy term. We prevent local scale ambiguities by imposing a set of fixed boundary 3D points. We experimentally prove the sufficiency of this set of boundary points and demonstrate the effectiveness of our approach on different developable and non-developable surfaces with a wide range of extensibility.
same-paper 2 0.83834738 226 cvpr-2013-Intrinsic Characterization of Dynamic Surfaces
Author: Tony Tung, Takashi Matsuyama
Abstract: This paper presents a novel approach to characterize deformable surface using intrinsic property dynamics. 3D dynamic surfaces representing humans in motion can be obtained using multiple view stereo reconstruction methods or depth cameras. Nowadays these technologies have become capable to capture surface variations in real-time, and give details such as clothing wrinkles and deformations. Assuming repetitive patterns in the deformations, we propose to model complex surface variations using sets of linear dynamical systems (LDS) where observations across time are given by surface intrinsic properties such as local curvatures. We introduce an approach based on bags of dynamical systems, where each surface feature to be represented in the codebook is modeled by a set of LDS equipped with timing structure. Experiments are performed on datasets of real-world dynamical surfaces and show compelling results for description, classification and segmentation.
3 0.83722645 9 cvpr-2013-A Fast Semidefinite Approach to Solving Binary Quadratic Problems
Author: Peng Wang, Chunhua Shen, Anton van_den_Hengel
Abstract: Many computer vision problems can be formulated as binary quadratic programs (BQPs). Two classic relaxation methods are widely used for solving BQPs, namely, spectral methods and semidefinite programming (SDP), each with their own advantages and disadvantages. Spectral relaxation is simple and easy to implement, but its bound is loose. Semidefinite relaxation has a tighter bound, but its computational complexity is high for large scale problems. We present a new SDP formulation for BQPs, with two desirable properties. First, it has a similar relaxation bound to conventional SDP formulations. Second, compared with conventional SDP methods, the new SDP formulation leads to a significantly more efficient and scalable dual optimization approach, which has the same degree of complexity as spectral methods. Extensive experiments on various applications including clustering, image segmentation, co-segmentation and registration demonstrate the usefulness of our SDP formulation for solving large-scale BQPs.
Author: Keyang Shi, Keze Wang, Jiangbo Lu, Liang Lin
Abstract: Driven by recent vision and graphics applications such as image segmentation and object recognition, assigning pixel-accurate saliency values to uniformly highlight foreground objects becomes increasingly critical. More often, such fine-grained saliency detection is also desired to have a fast runtime. Motivated by these, we propose a generic and fast computational framework called PISA Pixelwise Image Saliency Aggregating complementary saliency cues based on color and structure contrasts with spatial priors holistically. Overcoming the limitations of previous methods often using homogeneous superpixel-based and color contrast-only treatment, our PISA approach directly performs saliency modeling for each individual pixel and makes use of densely overlapping, feature-adaptive observations for saliency measure computation. We further impose a spatial prior term on each of the two contrast measures, which constrains pixels rendered salient to be compact and also centered in image domain. By fusing complementary contrast measures in such a pixelwise adaptive manner, the detection effectiveness is significantly boosted. Without requiring reliable region segmentation or post– relaxation, PISA exploits an efficient edge-aware image representation and filtering technique and produces spatially coherent yet detail-preserving saliency maps. Extensive experiments on three public datasets demonstrate PISA’s superior detection accuracy and competitive runtime speed over the state-of-the-arts approaches.
5 0.80051184 225 cvpr-2013-Integrating Grammar and Segmentation for Human Pose Estimation
Author: Brandon Rothrock, Seyoung Park, Song-Chun Zhu
Abstract: In this paper we present a compositional and-or graph grammar model for human pose estimation. Our model has three distinguishing features: (i) large appearance differences between people are handled compositionally by allowingparts or collections ofparts to be substituted with alternative variants, (ii) each variant is a sub-model that can define its own articulated geometry and context-sensitive compatibility with neighboring part variants, and (iii) background region segmentation is incorporated into the part appearance models to better estimate the contrast of a part region from its surroundings, and improve resilience to background clutter. The resulting integrated framework is trained discriminatively in a max-margin framework using an efficient and exact inference algorithm. We present experimental evaluation of our model on two popular datasets, and show performance improvements over the state-of-art on both benchmarks.
6 0.79807776 91 cvpr-2013-Consensus of k-NNs for Robust Neighborhood Selection on Graph-Based Manifolds
7 0.78633916 365 cvpr-2013-Robust Real-Time Tracking of Multiple Objects by Volumetric Mass Densities
8 0.78099924 44 cvpr-2013-Area Preserving Brain Mapping
10 0.7801652 71 cvpr-2013-Boundary Cues for 3D Object Shape Recovery
11 0.78000039 155 cvpr-2013-Exploiting the Power of Stereo Confidences
12 0.7796483 61 cvpr-2013-Beyond Point Clouds: Scene Understanding by Reasoning Geometry and Physics
13 0.77813882 147 cvpr-2013-Ensemble Learning for Confidence Measures in Stereo Vision
14 0.77732438 242 cvpr-2013-Label Propagation from ImageNet to 3D Point Clouds
15 0.77729815 373 cvpr-2013-SWIGS: A Swift Guided Sampling Method
16 0.77725369 244 cvpr-2013-Large Displacement Optical Flow from Nearest Neighbor Fields
17 0.77659082 19 cvpr-2013-A Minimum Error Vanishing Point Detection Approach for Uncalibrated Monocular Images of Man-Made Environments
18 0.77639014 298 cvpr-2013-Multi-scale Curve Detection on Surfaces
19 0.77562374 110 cvpr-2013-Dense Object Reconstruction with Semantic Priors
20 0.77533609 443 cvpr-2013-Uncalibrated Photometric Stereo for Unknown Isotropic Reflectances