Reduced reference image quality assessment using entropy of primitives

In this paper, we propose a new reduced reference image quality assessment algorithm based on the recent advances in sparse coding and representation, particularly, the entropy of primitives (EoP). The EoP is defined in terms of the distribution of the primitives, which form an overcomplete dictionary to represent the natural scene by linear combination. Constructively, we develop a reduced reference EoP based distortion metric (EoPM). EoPM has the property that it is nearly invariant to the geometry distortions, which hardly affect the visual quality but are often wrongly predicted by the existing image quality assessment metrics with severe distortion. Experimental results show that the accuracy of EoPM is highly competitive to the popular reduced reference image quality assessment algorithm on the public dataset.