cvpr01_segmentation - Object Based Segmentation of Video...

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

View Full Document Right Arrow Icon
Object Based Segmentation of Video Using Color, Motion and Spatial Information Sohaib Khan Mubarak Shah Computer Vision Laboratory School of EECS, University of Central Florida Orlando, FL 32816 khan,shah Abstract Video segmentation is different from segmentation of a sin- gle image. While several correct solutions may exist for segmenting a single image, there needs to be a consistency among segmentations of each frame for video segmenta- tion. Previous approaches of video segmentation concen- trate on motion, or combine motion and color information in a batch fashion. We propose a maximum a posteriori probability (MAP) framework that uses multiple cues, like spatial location, color and motion, for segmentation. We as- sign weights to color and motion terms, which are adjusted at every pixel, based on a confidence measure of each fea- ture. We also discuss the appropriate modeling of pdfs of each feature of a region. The correct modeling of the spa- tial pdf imposes temporal consistency among segments in consecutive frames. This approach unifies the strengths of both color segmentation and motion segmentation in one framework, and shows good results on videos that are not suited for either of these approaches. 1 Introduction Motion information has been used for video compression for a long time. Codecs based on MPEG 1-2 and H.26x se- ries of standards compute the motion of image blocks in a series of images and transmit only this motion information and the error in reconstructed image. If the estimates of mo- tion are reasonably correct, the entropy of the error image is much lower than the original image, thus achieving com- pression. Generally, some predetermined and fixed block size is used. Object-based segmentation, on the other hand, uses re- gions based on real world objects as compression primi- tives, rather than, say 8x8 blocks. Object based coding often creates primitives that are more homogeneous in texture and thus results in more compression. More importantly, how- ever, it allows the use of layers in coding. Real world scenes may be considered as a rendering of views of multiple ob- jects placed at appropriate locations. These objects may be in motion with respect to each other. If individual objects are segmented out during the coding phase, then they may be transmitted only once and the relative displacement of each layer may be transmitted in successive frames. The key bottleneck in object-based compression is re- liable segmentation of objects in an image. Image seg- mentation is a well studied but ill posed problem. Given a single image, there can be several ’correct’ segmentations of it. Moreover, there can be several levels of segmenta- tion. Looking at an outdoor scene, a clump of trees might be considered one segment, individual trees might be con- sidered different segments, or individuals features of a tree (like branches, trunk, fruit) might be segmented out. It is easy to visualize cases where, by zooming the camera into
Background image of page 1

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

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 02/09/2010 for the course MATH CS715 taught by Professor Zhulie during the Spring '10 term at University of Louisiana at Lafayette.

Page1 / 6

cvpr01_segmentation - Object Based Segmentation of Video...

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