Study On No-reference Image Quality Assessment Method Based On Visual Attention Mechanism

No-reference image quality assessment(NR-IQA)can solve the evaluation problem of image distortion in the absence of reference image,and has attracted the attention and participation of many scholars.With the leap of computer vision from perception to cognition,how to use the visual attention mechanism of human visual system(HVS)to innovate the NR-IQA,establish the evaluation model based on the visual attention mechanism,and achieve the consistency of objective and subjective evaluation,has become the research focus of theory and technology of NR-IQA.Based on a comprehensive analysis of the visual attention mechanism and the theory of image quality assessment,a new NR-IQA based on visual attention mechanism was proposed in this paper.The proposed method accurately divided the region of interest,extracted the characteristics representing the image quality of the region of interest,and used machine learning to establish the NR-IQA model.The main innovations of this paper are as follows:(1)This paper analyzed the principle of visual attention mechanism and the shortcomings of Itti visual attention model.The improved Itti model was proposed by adding edge in feature division stage and optimizing weight relationship in feature integration stage.(2)Based on the improved Itti model,the BRISQUE feature was extracted from the region of interest to describe the image quality,and the support vector regression was used to establish the mapping relationship between BRISQUE feature and subjective score.(3)The effectiveness of the improved Itti model and NRIQA model was verified through the comparison experiment.The experimental results showed that the accuracy of the saliency map extracted by improved Itti model was significantly improved,and the scores obtained by the NR-IQA model had a higher correlation with the subjective results.