210 likes | 237 Views
Explore robust stereo fitting using 3D morphable models (3DMMs) with occlusion handling for improved accuracy. Leveraging joint shape and texture stereo fitting, this approach enhances results under various occlusion scenarios. The methodology involves an optimization procedure and multiple experiments to demonstrate its efficacy.
E N D
Model-Based Stereo with Occlusions Fabiano Romeiro and Todd Zickler TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: AAAAAAAAA
Introduction Varying illumination Varying pose Occlusions Varying expressions
Introduction Past Work: Image-based For example: Eigenfaces [Turk and Pentland, 1991] Fisherfaces [Belhumeur et al, 1997]
3D Morphable Models (3DMMs) 2D+3D AAMs 2D AAMs [Cootes et al, 1998; Baker et al, 2004; Mathews et al, 2004; Gross et al, 2006] [Xiao et al, 2004] [Blanz et al, 1999; Blanz et al, 2003; Blanz et al, 2005; Smet et al, 2006] Introduction Past Work: Model-based
3D Morphable Models (3DMMs) Pros • - Self-occlusion handled by model itself • - Allows direct modeling of illumination Cons - Difficult and expensive fitting process [Blanz et al, 1999; Blanz et al, 2003; Blanz et al, 2005; Smet et al, 2006] Introduction Past Work: 3DMMs
- Stereo based cost [Fransens et al, 2005] • Stereo fitting with both shape and texture • →Robust to foreign-body occlusions Introduction Past Work: Stereo 3DMMs - Texture model not needed Our work • → Improved Accuracy
Outline • 3DMM Background • Joint Shape and Texture Stereo Fitting • Handling Occlusions • Conclusions
Basis for shape - Basis for texture Background 3DMMs Vectorization of laser scans: PCA performed: [Blanz and Vetter, 1999]
[Blanz and Vetter, 1999] Background 3DMMs Representation of face shape and texture: Prior probabilities on the coefficients:
Optimization Procedure Initial fit - Fit Shape, Pose to minimize reprojection error on selected feature points - Rough initial estimates of Shape and Pose Optimization procedure 4 experiments Stereo and texture Stereo Robust Stereo and texture Robust Stereo
Results First 2 experiments: Stereo and Texture vs. Stereo 480 recovered shape models (60 individuals, 8 poses) K.U. Leuven Stereo face database [Fransens et al, 2005]
Results Qualitative Results Stereo and Texture Stereo
Results Qualitative Results Stereo+texture Stereo
Results Quantitative Results
Results Under Occlusions Half-Occlusion Full Occlusion near Full-Occlusion far
Robust Stereo Robust S+T Results Under Occlusions Input Robust Stereo Robust S+T Shape Estimate Occlusion Map
Conclusions Robust stereo fitting of 3DMMs - Uses both stereo constraint, texture information • Increased accuracy of fit - Ability to handle occlusions Future Work - More sophisticated stereo matching term - Different feature spaces - Break model into segments respecting occlusion boundaries