Reduced reference image quality assessment using Principal Component Analysis

Abstract

Listen

Translate article

With the tendency of converging services delivered on wired and wireless networks, the consumer expectancy includes live video streaming. In this context, the objective image and video quality assessment becomes an essential but a challenging requirement. With the rapid evolution of the wireless video applications, the continuation of the Quality of Service (QoS) is a key paradigm for the roll-out of these services, which demands for an efficient quality evaluator for the dynamic monitoring and parameter setting of the digital video system. In this paper, we propose a Reduced-Reference (RR) image quality assessment metric based on the Principal Component Analysis (PCA). In our metric, the transformed data set is obtained which represents the original data solely in terms of the eigenvectors we choose, giving us the most efficient expression of the data. The mean gradient values are calculated from the transformed data using edge detection methods based on the Sobel-operator. We define a Quality Index that measures the difference between our metric's values computed on the transmitted and received images, respectively. The experimental results show that our RR-PCA proposed metric correlates well with the subjective quality scores.