Image Retargeting Quality Assessment Based on Aspect Ratio Similarity of Salient Region

The shape distortions of the important regions in retargeted image attracts human perception, which are seldom considered in the existing image retargeting quality assessment (IRQA) algorithms. This would lower the performance of evaluation algorithms obviously. To address this issue, we propose an IRQA method which measures the aspect ratio similarity of the salient regions between the original and the retargeted images. To evaluate the distortions of the visual important contents, we first use a novel algorithm to extract the salient regions of retargeted images accurately, which makes full use of the original image to perform backward registration and index matching of salient region. Then the aspect ratio similarity of salient regions is proposed to measure the geometric distortion of silent regions in retargeted images. Experimental results show that the performance of the proposed IRQA method is superior to the other existing algorithms.