Sparse Structural Similarity for Objective Image Quality Assessment

In this paper, a novel full-reference (FR) image quality assessment (IQA) metric based on sparse representation is proposed. Sparse representation has been widely applied in many applications such as image denoising and restoration. It is a high-efficiency way in representing sparse and redundant natural images. Also it has been shown to be highly related to the human visual perception, which is characterized by a set of responses of neurons in visual cortex. In this paper, the sparse representation is applied in decomposing natural images into multiple layers depending on the visual importance. Inspired by these observations, a novel IQA metric called sparse structural similarity is proposed by measuring the fidelity of the stimulation of visual cortices. Experimental results on public databases indicate that the proposed method is effective in predicting subjective evaluation and as compared to state-of-the-art FR-IQA methods.