An Object IQA Framework Based on Saliency Map Similarity

Objective image quality assessment (IQA) aims to establish a mathematical model on computer to simulate human perception for image to evaluate the quality of it. This paper analyzes the image characteristics for IQA, from which gradient, spectrum, and color are selected as indexes to establish a novel IQA framework. The gradient similarity, spectral similarity and color similarity are fully utilized, and the similarity calculation results are weighted by the spectral saliency map to obtain the final evaluation result. The experiments were conducted on both TID2008 and TID2013 dataset respectively. Through comparison, it is found that the evaluation algorithm proposed in this article can not only ensure the robustness of the evaluation result, but also perform a lower algorithm complexity.