A Study of Perceptual Quality Assessment for Stereoscopic Image Retargeting

Subjective and objective perceptual quality assessment for stereoscopic retargeted images is a fundamentally important issue in stereoscopic image retargeting (SIR) which has not been deeply investigated. Here, a stereoscopic image retargeting quality assessment (SIRQA) database is proposed to study the perceptual quality of different stereoscopic retargeted images. To construct the database, we collect 720 stereoscopic retargeted images generated by eight representative SIR methods. The perceptual quality (mean opinion scores, MOS) of each stereoscopic retargeted image is subjectively rated by 30 viewers. For objective assessment, several publicly available quality evaluation metrics are tested on the database. Experimental results show that there is a large room for improving the accuracy of objective quality assessment in SIRQA by comprehensively considering geometric distortion, content loss and stereoscopic perceptual quality.