A correlation-based model prior for stereo

All non-trivial stereo problems need model priors to deal with ambiguities and noise perturbations. To meet requirements of increasingly demanding tasks such as modeling for rendering, a proper model prior should impose preference on the true scene structure, while avoiding artificial bias such as fronto-parallel. We introduce a geometric model prior based on a novel technique we call kernel correlation. Maximizing kernel correlation is shown to be equal to distance minimization in the M-estimator sense. As a model prior, kernel correlation is demonstrated to have good properties that can result in renderable, very smooth and accurate depth map. The results are evaluated both qualitatively by view synthesis and quantitatively by error analysis.

[1]  Richard Szeliski,et al.  A Taxonomy and Evaluation of Dense Two-Frame Stereo Correspondence Algorithms , 2001, International Journal of Computer Vision.

[2]  Richard Szeliski,et al.  Hierarchical spline-based image registration , 1994, 1994 Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.

[3]  P. Anandan,et al.  A computational framework and an algorithm for the measurement of visual motion , 1987, International Journal of Computer Vision.

[4]  Pascal Fua,et al.  Reconstructing complex surfaces from multiple stereo views , 1995, Proceedings of IEEE International Conference on Computer Vision.

[5]  Richard Szeliski,et al.  Layered depth images , 1998, SIGGRAPH.

[6]  E. Parzen.  On Estimation of a Probability Density Function and Mode , 1962 .

[7]  Olga Veksler,et al.  Fast Approximate Energy Minimization via Graph Cuts , 2001, IEEE Trans. Pattern Anal. Mach. Intell..

[8]  A. Rényi.  On Measures of Entropy and Information , 1961 .

[9]  Carlo Tomasi,et al.  Surfaces with occlusions from layered stereo , 2004, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[10]  Hai Tao,et al.  A global matching framework for stereo computation , 2001, Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001.

[11]  Carlo Tomasi,et al.  Multiway cut for stereo and motion with slanted surfaces , 1999, Proceedings of the Seventh IEEE International Conference on Computer Vision.

[12]  Demetri Terzopoulos,et al.  Snakes: Active contour models , 2004, International Journal of Computer Vision.

[13]  Richard Szeliski,et al.  Symmetric Sub-Pixel Stereo Matching , 2002, ECCV.

[14]  S. Birchfiled.  A Pixel Dissimilarity Measure That Is Insensitive to Image Sampling , 1998 .

[15]  Yuichi Ohta,et al.  Occlusion detectable stereo-occlusion patterns in camera matrix , 1996, Proceedings CVPR IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[16]  S. Ullman,et al.  The interpretation of visual motion , 1977 .

[17]  Takeo Kanade,et al.  A Stereo Matching Algorithm with an Adaptive Window: Theory and Experiment , 1994, IEEE Trans. Pattern Anal. Mach. Intell..

[18]  Darius Burschka,et al.  Advances in Computational Stereo , 2003, IEEE Trans. Pattern Anal. Mach. Intell..

[19]  H. C. Longuet-Higgins,et al.  An algorithm for associating the features of two images , 1991, Proceedings of the Royal Society of London. Series B: Biological Sciences.

[20]  Kim Bong-Gyum,et al.  Stereo Matching using Belief Propagation with Line Grouping , 2005 .

[21]  R. Cipolla,et al.  A probabilistic framework for space carving , 2001, Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001.

[22]  Robert Tibshirani,et al.  The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd Edition , 2001, Springer Series in Statistics.

[23]  Richard Szeliski,et al.  Motion Estimation with Quadtree Splines , 1996, IEEE Trans. Pattern Anal. Mach. Intell..

[24]  T. Kanade,et al.  Kernel correlation as an affinity measure in point-sampled vision problems , 2003 .