Object Co-segmentation based on Shortest Path Algorithm and Saliency Model

Segmenting common objects that have variations in color, texture and shape is a challenging problem. In this paper, we propose a new model that efficiently segments common objects from multiple images. We first segment each original image into a number of local regions. Then, we construct a digraph based on local region similarities and saliency maps. Finally, we formulate the co-segmentation problem as the shortest path problem, and we use the dynamic programming method to solve the problem. The experimental results demonstrate that the proposed model can efficiently segment the common objects from a group of images with generally lower error rate than many existing and conventional co-segmentation methods.


Fanman Meng, Hongliang Li, Guanghui Liu, and King Ngi Ngan, "Object Co-segmentation based on Shortest Path Algorithm and Saliency Model,"IEEE Transactions on Multimedia, vol.14, no. 5, pp. 1429 - 1441, 2012.  [PDF][Source code]


The framework of the proposed method.

The segmentation results of [16], [21], [20] and the proposed method by considering color. These classes are goose2, FC player, hot balloons, flowers2, bear and goose1. The rows 1, 6, 11: Original images. The rows 2, 7, 12: The results for the method in [16]. The rows 3, 8, 13: The results for the method in [21]. The rows 4, 9, 14: The results for the method in [20]. The rows 5, 10, 15: The results by the proposed method.
The co-segmentation results by considering shape. For each class, the first row shows original images. The following three rows display the experimental results of the methods in [16], [21], [20] and ours, respectively.
Source Code (MATLAB)
   Source code has been released! [Source code]
This work was partially supported by NSFC (No.60972109 and 61101091), the Program for New Century Excellent Talents in University (NCET-08-0090), the Ph.D. Programs Foundation of Ministry of Education of China (No. 20110185110002), and the Fundamental Research Funds for the Central Universities (E022050205).