An improved automatic lipreading system to enhance speech recognition
CHI '88 Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
Non-parametric local transforms for computing visual correspondence
ECCV '94 Proceedings of the third European conference on Computer Vision (Vol. II)
ACM Computing Surveys (CSUR)
Building 3-D Human Face Models from Two Photographs
Journal of VLSI Signal Processing Systems - Special issue on multimedia signal processing
A Taxonomy and Evaluation of Dense Two-Frame Stereo Correspondence Algorithms
International Journal of Computer Vision
Binocular Stereo by Maximizing the Likelihood Ratio Relative to a Random Terrain
RobVis '01 Proceedings of the International Workshop on Robot Vision
Multi-camera Scene Reconstruction via Graph Cuts
ECCV '02 Proceedings of the 7th European Conference on Computer Vision-Part III
Markov Random Fields with Efficient Approximations
CVPR '98 Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
A Maximum-Flow Formulation of the N-Camera Stereo Correspondence Problem
ICCV '98 Proceedings of the Sixth International Conference on Computer Vision
Robustness to Noise of Stereo Matching
ICIAP '03 Proceedings of the 12th International Conference on Image Analysis and Processing
Real-time combined 2D+3D active appearance models
CVPR'04 Proceedings of the 2004 IEEE computer society conference on Computer vision and pattern recognition
An interactive 3D video system for human facial reconstruction and expression modeling
Journal of Visual Communication and Image Representation
Hi-index | 0.00 |
This paper compares the efficiency of several stereo matching algorithms in reconstructing 3D faces from both real and synthetic stereo pairs. The stereo image acquisition system setup and the creation of a face disparity map benchmark image are detailed. Ground truth is build by visual matching of corresponding nodes of a dense colour grid projected onto the faces. This experiment was also performed on a human face model created using OpenGL with mapped texture to create as perfect as possible a set for evaluation, instead of real human faces like our previous experiments. Performance of the algorithms is measured by deviations of the reconstructed surfaces from a ground truth prototype. This experiment shows that contrary to expectations, there is seemingly very little difference between the currently most known stereo algorithms in the case of the human face reconstruction. It is shown that by combining the most efficient but slow graph-cut algorithm with fast dynamic programming, more accurate reconstruction results can be obtained.