Co‐segmentation of multiple similar images using saliency detection and region merging. Issue 3 (1st June 2014)

• Record Type: Journal Article
• Authors: Zhou, Chongbo; Liu, Chuancai
• Is Part Of: IET computer vision. Volume 8:Issue 3(2014)
• Page Start: 254
• Page End: 261
• Publication Date: 2014-06-01
• Subjects: image segmentation; multiple similar images cosegmentation -- saliency detection -- region merging -- maximal similarity measurement -- nearest similarity measurement; Computer vision -- Periodicals; Pattern recognition systems -- Periodicals; 006.37
• DOI: 10.1049/iet-cvi.2012.0266
• Languages: English
• ISSNs: 1751-9632

Abstract

Abstract : The aim of co‐segmentation is to simultaneously segment multiple images depicting an identical or similar object. In this study, a co‐segmentation method using saliency detection and region merging is proposed. The saliency detection results using different detection methods on different types of colour space are combined to produce seed regions for each image in the image group. The initial seed regions of all the images are refined by eliminating the dissimilar ones to ensure accurate seed regions for each images as possible. Region merging is performed on each image individually in order to allow our method to be applied to large image groups. The maximal similarity measurement and nearest similarity measurement are defined as merging rules. The deliberately designed merging strategy aims to merge two regions using the maximal similarity rule and label two regions as the same class but not merge them using the nearest similarity rule. The proposed method has been compared with some state‐of‐the‐art methods on three datasets, and the experimental results show its effectiveness.