Quaternion Model of Fast Video Quality Assessment Based on Structural Similarity Normalization

Abstract

Video quality evaluation methods have been widely studied because of an increasing need in variety of video processing applications, such as compression, analysis, communication, enhancement and restoration. The quaternion models are also widely used to measure image or video quality. In this paper, we proposed a new quaternion model which mainly describes the contour feature, surface feature and temporal information of the video. We use structure similarity comparison to normalize four quaternion parts respectively, because each part of the quaternion use different metric. Structure similarity comparison is also used to measure the difference between reference videos and distortion videos. The results of experiments show that the new method has good correlation with perceived video quality when tested on the video quality experts group (VQEG) Phase I FR-TV test data set.