Research On The No-reference Quality Assessment Method Of Blur Distortion Image

In the process of image acquisition,processing,transmission and storage,the images will appear different levels of distortion,which will affect human perception and increase the difficulty to get the image information.As a consequence,establishing a good image quality evaluation mechanism exerts significant function in the image processing field.Since it is difficult to obtain the reference image in the practical application,the research on the no-reference image quality evaluation method is of more practical value.Blurred image is the most common type of image distortion.This paper is based on the study of quality evaluation methods of the blurred image,combining the structural similarity theory,image region division and image feature extraction,accordingly proposed a no-reference image quality assessment method and improved method for the blurred image.Since no-reference image quality evaluation method is unable to valuate image quality effectively,on the basis of study of related references and image quality evaluation methods,this paper introduces a kind of no-reference image quality evaluation method,which is based on region division.To begin with,the blurred image is subjected to low-pass filtering,and the second blurred image is obtained as a reference image.Secondly,the blurred image and reference image are regionally divided.This paper introduces the three region division method and the four area division method,the four area division method is the improvement on the three regions division method.It divides the image into reserved edge regions,changing edge regions,texture regions and flat regions through gray scale values of the image.And then we calculate the structural similarity(SSIM)of each region that will be used as the image quality evaluation index.Finally,the support vector regression(SVR)model is established to predict the input data to get the final objective evaluation value.This method can obtain better evaluation result.As for the blurred image,the change of edge feature,texture feature and structure feature has great influence on image quality.This paper proposes a blur no-reference image quality evaluation method based on region division that firstly uses the low pass filter to construct the reference image by subjecting the blurred image to the secondary blurred processing.The image is then divided into a reserved edge region,a modified edge region,and a flat area.And then we use the local binary patterns(LBP)algorithm to extract the texture area of the image for replacing the texture area in the four-region division method.By calculating the edge gradient similarity of the reserved edge region,each components of the Tamura texture area as the texture feature of the image,and the structural similarity of the four regions,we fuse the edge gradient similarity,texture similarity,and structural similarity as the image quality evaluation index.Then we get final objective quality evaluation value by the prediction of training SVR model.The Pearson correlation coefficient and Spearman level correlation coefficient of this method can reach more than 0.96,which has a better algorithm performance.