An image quality metric based on the Harris corner
Yazhou Yang, Guangquan Cheng, Dan Tu · 2012
Many image quality assessment (IQA) methods involve a two-stage structure: local distortion measurement followed by pooling. In this work, the optimal pooling strategy is investigated by using visual attention method. Under the hypothesis that humans tend to pay more visual attention to the regions with important information content, this paper explores the potential of a visual attention based pooling strategy, where more weights should be assigned to the regions subtending the Harris corner points. We combine the Harris corner based visual attention model with the structural similarity (SSIM) index, yielding a Harris-corner weighted SSIM (H-SSIM) approach. Experimental results on the LIVE image database show that the proposed H-SSIM approach achieves superior or comparable performance compared with a number of competitive IQA algorithms.