Research On Image Quality Evaluation Method Based On Visual Apparentness

With the rapid development of electronic information technology,computer application technology and multimedia communication technology,the higher quality image is required,So image processing technology has received more attention.The existing methods of evaluating image quality are numerous,but their accuracy is unsatisfactory and difficult to be consistent with the subjective feelings of the human visual system.The image quality is based on the human perception which is closely related to the human visual system.Therefore,It is hoped that the evaluation result of the objective image quality evaluation model should be consistent with the subjective evaluation of the human eye.Many existing objective evaluation models are designed to achieve this goal.The effect of visual saliency in the human visual system is similar to that in the image preprocessing.Through visual saliency,we can select the most interesting information from the visual scene.It is possible to get closer to subjective perception by integrating visual saliency into the image quality evaluation algorithm.Therefore,it is the main trend for current image quality evaluation algorithm to integrate multiple HVS features in order to obtain higher evaluation performance.Based on the research of the existing image quality evaluation algorithm and the significance algorithm,this paper analyzes the existing image quality evaluation algorithm which is based on visual saliency.According to the fact that the accuracy of the evaluation for existing visual saliency mode integrated into the image quality evaluation algorithm cannot be improved.An improved HFT algorithm is firstly proposed to improve the accuracy of the image quality evaluation model.The improved algorithm combined the edge significance of the image on the basis of the HFT method,and fused the final saliency map into the existing objective image quality evaluation algorithm.In addition,this paper has been experimentally simulated in the LIVE database.By comparing each image horizontally,it is found that not all evaluation result is improved after adding visual saliency into the original image.It is further discovered that if the salient region of the image are divergent,that is,the significant area is not concentrated in the whole image,but distributed in every corner of the image,the performance of the evaluation algorithm will not improve but will decline.Based on this,the convergence of the image is determined before the image fusion.The degree of convergence is significant by using the information entropy of the saliency map.When the image is judged to be convergent,the visual saliency is incorporated into the image quality evaluation algorithm;otherwise,no saliency map fusion is performed.Experiments show that the evaluation effect of this algorithm has been improved obviously compared with the existing image quality evaluation model,which is much closer to human perception.