Abstract:

Stereo image segmentation is a crucial and difficult issue in the field of object-based stereo image processing.By improving the Grabcut algorithm and exploiting the inter-view correlations,a novel stereo image segmentation scheme was proposed.The original left image was firstly transformed into a super-pixel image by improving the Slic algorithm.Then the super-pixel image was segmented and the foreground object in the left image is extracted based on the framework of Grabcut,where the energy function was rebuilt.Finally,by exploiting the inter-view correlations between the left and the right images,the foreground object in the right image was obtained based on contour correspondence by fusing color and texture features.The experimental results show that the proposed algorithm is more efficient and can achieve better segmentation results than the existing methods.

Key words: stereo image segmentation, improved Grabcut algorithm, inter-view correlation, contour correspondence