07_Vu_CVPR2008 - Shape Prior Segmentation of Multiple...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Shape Prior Segmentation of Multiple Objects with Graph Cuts Nhat Vu and B.S. Manjunath Department of Electrical and Computer Engineering University of California, Santa Barbara, CA 93106-9560 { nhat,manj } @ece.ucsb.edu Abstract We present a new shape prior segmentation method us- ing graph cuts capable of segmenting multiple objects. The shape prior energy is based on a shape distance popular with level set approaches. We also present a multiphase graph cut framework to simultaneously segment multiple, possibly overlapping objects. The multiphase formulation differs from multiway cuts in that the former can account for object overlaps by allowing a pixel to have multiple la- bels. We then extend the shape prior energy to encompass multiple shape priors. Unlike variational methods, a major advantage of our approach is that the segmentation energy is minimized directly without having to compute its gradi- ent, which can be a cumbersome task and often relies on ap- proximations. Experiments demonstrate that our algorithm can cope with image noise and clutter, as well as partial occlusions and affine transformations of the shape. 1. Introduction Segmentation methods based solely on image informa- tion [ 6 , 8 , 15 , 22 ] often perform poorly in the presence of noise, background clutter, and object occlusions. The ad- dition of shape prior information has shown to significantly improve segmentation results and is popular among varia- tional approaches [ 7 , 9 , 20 , 24 , 25 , 28 ]. Recently, there has been an increased interest in graph based segmentation al- gorithms [ 1 , 2 , 5 ], and subsequently the addition of prior shape information into their formulations. However, many continuous shape distances or dissimilarity measures can be difficult, if not impossible, to formulate as discrete energies for graph optimization. This is especially true for graph cut methods. The graph methods of Felzenszwalb [ 11 ] and Schoen- emann and Cremers [ 26 ] can segment objects under elastic deformations without needing any initialization and guaran- tee globally optimal solutions. In [ 11 ], nonserial dynamic programming is used to find the optimal matching between a deformable template represented by triangulated polygons and the image pixels. In [ 26 ], the segmentation is found by computing the minimal ratio cycle in a product graph of the image and a shape template parameterized by arc length. Both of these methods can be slow in practice, with run- times of up to several minutes on typical CPUs. Moreover, the triangulated polygon representations and arc length pa- rameterizations limit the topological flexibility of the tem- plate shapes and may not easily extend to the 3D case....
View Full Document

This note was uploaded on 12/28/2011 for the course BIO 100 taught by Professor Gomez during the Fall '11 term at UCSB.

Page1 / 8

07_Vu_CVPR2008 - Shape Prior Segmentation of Multiple...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online