Image Quality Assessment Method Based On Image Structure And HVS Characteristics Fusion

The visual signals are subject to various distortions in the process of acquisition,compression,storage,processing,transmission and reproduction,leading to perceived quality degradation.Therefore,image quality assessment plays a very important role in today's visual signal processing.Since the Human Visual System(HVS)is the recipient of visual information,the subjective image quality evaluation method is the most reliable way to evaluate the quality of digital images.However,it's time-consuming,laborious and high cost.So it is objective image quality evaluation methods which are widely used in practical applications.Methods like mean square error(MSE)and peak signal to noise ratio(PSNR)are pixel based methods which are not consistent well with the subjective image quality evaluation methods.Developing more accurate objective image quality evaluation method is very important for future visual information processing and communication applications.Image quality assessment methods based on image structure and HVS characteristics have become very popular in recent years.A series of new algorithms have been proposed.In this paper,two image quality evaluation methods based on image structure and human visual characteristics fusion are proposed.One is based on image edge and Log-Gabor frequency domain features.Compared with other methods which only use gradient to reflect the edge feature,edge strength is also considered in this method.0 °,45 °,90 °,135 °four directions are considered in edge strength calculation,so edge strength can better detect the edge region.In addition,salient region detection mechanism can be well approximated by integrating log-Gabor filter responses from opponent color channels.This method combines the structural features of the image and the visual characteristics of the human eye;the other is based on the structure contrast and the frequency domain features.The structural contrast can effectively reflect the complexity of the image texture and the masking effect of the human visual system(HVS).Frequency domain feature can represent the contrast sensitivity function of human eyes.Structural features of the image and the visual characteristics of the human eye are also combined in this method.Finally we get the objective score through features fusion.Further experiments are conducted on TID2013,LIVE and every type of distortions in TID2013.We compared these two methods with other 7 methods.Compared to methods based solely on image structure or human visual characteristic,these two methods with high computional accuracy consist well with subjective evaluation methods.