Segmentation of non-rigid bodies in affine motion: a new framework

In this paper a new technique for partitioning a non-rigid body (human body), when it is in affine motion is presented. The technique is based on finding the contour in the "modified difference frame". These contours can then be used in identifying the regions of interest. Results show that the technique eliminates a lot of stationary regions and thus can reduce the amount of computational time required for the processing of the image. It reduces significantly the total bit rate required for transmission. It also provides an algorithm for contour tracing using extended boundary concept and contour compression by contour merging.

[1]  Yuh-Tay Liow A contour tracing algorithm that preserves common boundaries between regions , 1991, CVGIP Image Underst..

[2]  Yee-Hong Yang,et al.  Human body motion segmentation in a complex scene , 1987, Pattern Recognit..

[4]  Mubarak Shah,et al.  Establishing motion correspondence , 1991, CVGIP Image Underst..

[5]  Kanti V. Mardia,et al.  Deformable Template Recognition of Multiple Occluded Objects , 1997, IEEE Trans. Pattern Anal. Mach. Intell..

[6]  Jake K. Aggarwal,et al.  Estimation of motion from a pair of range images: A review , 1991, CVGIP Image Underst..

[7]  Alfred M. Bruckstein,et al.  A new method for image segmentation , 1988, [1988 Proceedings] 9th International Conference on Pattern Recognition.

[8]  Demin Wang,et al.  Segmentation-based motion-compensated video coding using morphological filters , 1997, IEEE Trans. Circuits Syst. Video Technol..

[9]  Alex Pentland,et al.  Pfinder: Real-Time Tracking of the Human Body , 1997, IEEE Trans. Pattern Anal. Mach. Intell..

[10]  Sun-Yuan Kung,et al.  Rate optimization by true motion estimation , 1997, Proceedings of First Signal Processing Society Workshop on Multimedia Signal Processing.