Using image signature for effective and efficient reduced-reference image quality assessment

Abstract

Reduced-reference (RR) image quality assessment (IQA) with only partial information of the reference image available, has aroused increasing research interests nowadays. Many efforts have been devoted to this area for years, and have introduced various kinds of effective models. However, those arithmetics are usually extremely complicated. Therefore, we in this paper propose a new RR Image-Signature (RRIS) induced IQA metric by estimating the similarity between the transformed images that are induced by the image signature. The main design principle of the proposed method is that the image signature can capture the main information of an image with very few features and computational cost. The proposed algorithm is tested and verified on four large-size image quality databases (TID2008, CSIQ, LIVE and CID2013). Experimental results demonstrate the effectiveness and efficiency of our algorithm over those competing full- and reduced-reference IQA methods.