Saliency Map Based Image Segmentation Using Level Set

摘要：

A major task in image segmentation is to extract salient regions or objects for further Content-Based Image Retrieval (CBIR). Those regions or objects are also named as the region of interest (ROI) and the detection of ROI is important for later image compression, transmission and storage. In this paper, a coarse-to-fine segmentation is applied using both visual conspicuity map and level set. Inspired by the visual saliency map proposed by Laurent Itti, several kinds of conspicuity maps are constructed and used in coarse segmentation. Corresponding to the map used above, different component of image is picked up and employed in the finer segmentation. All segmentations are achieved by Chan-Vese active model that is a well-known image segmentation technique implemented by level set method. Finally, our method is tested on one gray level image and two color images. Segmentation results show that our method could extract different salient regions and objects commendably.

关键词： Visual saliency image segmentation level set

作者: Chen Wang Honghang Wang Yanbo Zhang

作者单位: School of Electronic and Information Engineering, Xian Jiaotong University, Xian, Shaanxi, China School of Mechanical Engineering, Xian Jiaotong University, Xian, Shaanxi, China

会议类型: 国际会议

会议名称: 2010 International Conference on Image Analysis and Signal Processing(2010 图像分析与信号处理国际会议 IASP 10)

会议地点: 厦门

会议语种:英文

页码: 90-93

在线出版日期: 2010-04-12（万方平台首次上网日期，不代表论文的发表时间）