Sparse feature fidelity for image quality assessment

A quality metric called sparse feature fidelity (SFF) is proposed for full-reference image quality assessment (IQA). It is inspired by the fact that images are transformed into sparse representations by the primary visual cortex which is the most important part of the human visual system (HVS). The proposed method is based on sparse features that are acquired from a set of feature detectors called simple cell matrix which is trained on samples of natural images by a sparse coding algorithm. Then the SFF scores are obtained by a similarity measurement between the features of reference and distorted images. Moreover, two strategies are designed to simulate the properties of the visual perception: visual attention and visual threshold. Experimental results on four image databases show that SFF is more consistent with the subjective evaluations than the leading IQA methods.