Title: ROI extraction based on visual salience and visual evaluation

Authors: Zailiang Chen; Huajie Huang; Hailan Shen; Beiji Zou; Jiang Wang

Addresses: School of Information Science and Engineering, Central South University, 411 Computer Building, No. 932, South Lushan Road, Changsha City, Hunan Province, 410083, China ' School of Information Science and Engineering, Central South University, 411 Computer Building, No. 932, South Lushan Road, Changsha City, Hunan Province, 410083, China ' School of Information Science and Engineering, Central South University, 411 Computer Building, No. 932, South Lushan Road, Changsha City, Hunan Province, 410083, China ' School of Information Science and Engineering, Central South University, 411 Computer Building, No. 932, South Lushan Road, Changsha City, Hunan Province, 410083, China ' Jinchuan Group Engineering and Construction Co., Ltd., Jinchang City, Gansu Province, 737104, China

Abstract: With saliency map generated from visual attention model, this paper proposes two regions of interest (ROI) extraction algorithms respectively based on salient points and saliency regions. The former one adopts statistical and clustering techniques, selects cluster centre as seed points to fill outline map of image and finally makes mask operation between filled outline map and input image to implement ROI extraction. The latter one is based on salient regions, uses improved Grabcut image segmentation algorithm and saliency map generated from visual attention model to implement ROI extraction. To evaluate the performance of two proposed algorithms, this paper uses extracted ROI based on eye-movement data as evaluation criterion. The results show the algorithm based on salient points is applicable to extract simple images and has less runtime. The algorithm based on saliency regions is applicable to extract colourful and complex image. These two algorithms can be combined to get better performance.

Keywords: regions of interest; ROI extraction; visual evaluation; autonomous interaction; visual attention model; salient points; salient regions; visual salience; eye movement; image segmentation; saliency map; trimap; fixation; clustering.

DOI: 10.1504/IJAACS.2016.075392

International Journal of Autonomous and Adaptive Communications Systems, 2016 Vol.9 No.1/2, pp.57 - 70

Available online: 19 Mar 2016 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article