Saliency driven region-edge-based top down level set evolution reveals the asynchronous focus in image segmentation. (August 2018)

• Record Type: Journal Article
• Authors: Zhi, Xu-Hao; Shen, Hong-Bin
• Is Part Of: Pattern recognition. Volume 80(2018:Aug.)
• Page Start: 241
• Page End: 255
• Publication Date: 2018-08
• Subjects: Image segmentation -- Level set evolution -- Saliency map -- Edge -- SDREL; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2018.03.010
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: A new level set energy function has been designed by incorporating both the region and edge information. Saliency knowledge has been modeled into the level set evolution. A hierarchical evolution approach has revealed the asynchronous focus in image segmentation. Abstract: Level set method (LSM) is popular in image segmentation due to its intrinsic features for handling complex shapes and topological changes. Existing LSM-based segmentation models can be generally grouped into region- and edge-based models. The former often have problems to deal with images whose objects have similar color intensity to that of the background when the region descriptor is insufficient. The latter usually suffer to boundary leakage problem when the images' edges are weak. To overcome these problems, we present a novel hierarchical level set evolution protocol (SDREL), wherein we propose to use both saliency map and color intensity as region external energy to motivate an initial evolution of level set function (LSF), followed by the LSF and further smoothed by an internal energy (regulation term) to recognize a more precise boundary positioning. Our results show that the newly introduced saliency map term improves extracting objects from complex background and the asynchronous evolution of a single LSF results in a better segmentation. The new hierarchical SDREL model has been evaluated extensively and the results indicate that it has the merits of flexible initialization, robust evolution, and fast convergence. SDREL is available at: www.csbio.sjtu.edu.cn/bioinf/SDREL/.