Self-Organized Figure-Ground Segmentation by Multiple-Cue Integration

X. Tang, D. Garrett (USA), and C. von der Malsburg (Germany)


Image Processing and Applications, Image Segmentation, Cue Integration, Self-Organization.


This paper presents an efficient and reliable approach for figure-ground segmentation from video sequences that are taken with a stationary camera. Following the princi ples of “Democratic Integration”, the segmentation system dynamically integrates multiple visual cues, and aims to achieve synergy by sharing information from independent sources. A probabilistic cue integration framework is for mulated using Bayes’ rule. Each pixel in each frame can decide its own layer assignment distribution by deriving the posterior probabilities from the likelihood models of back ground subtraction, color and texture cues. These likeli hood models provide independent and complementary ob servations that are subsequently trained by self-adaptation toward the segmentation consensus. The amount of the contribution from an individual cue under different situa tions is adjusted by measuring that cue’s quality based on its similarity with the overall segmentation agreement. Al lowing cooperation and competition among the cues at the same time, the system maintains and improves its segmen tation results in a self-organized manner. We demonstrate the performance of the proposed system on several differ ent video sequences, where no specific parameter tuning is required for a particular sequence.

Important Links:

Go Back