Image Quality Assessment Algorithm Research Based On Visual Feature

With the development of multimedia and computer networking, image qualityassessment as a new technology, this becomes one of the key technologies forbenchmarking the image compression or image processing algorithms and becomes animportant role in the multimedia application system. On the basis of preious study, wefocus on the image quality assessment algorithm based on visual perception.Innovations of this thesis are given as follows.An image quality assessment algorithm based on the degree of visual perception isproposed. First, an image is partitioned into three parts: edges、textures and flat regionsby the content of image. Edges are the most important feature of an image. Texturesplay in human perception is the second important following edges. The flat region is thelast important region. Thus, second, we calculate the structural similarity (SSIM) ofthree region. Finally, the quality assessment value is obtained based on fuzzy integralswith the information fusion for the three value of SSIM.Experimental results show thatthe effectiveness of the proposed method.(2)An image quality assessment method is presented based on the localcharacteristics in the wavelet domain. First, according to the degree of important of theimage edges、textures and flat regions,the original image and distortion image arepartitioned into three parts. Then the discrete wavelet transform is applied to decomposethe three regions respectively. The subband wavelet coefficients of different scale arefiltered by the contrast sensitivity function and masking. Second, the just noticeabledistortion(JND) is computed by the correlation between the subband images, theperception coefficients are obtained by using the JND filtering, the normalizedhistogram is then constructed. Finally, the image quality assessment value is obtained bythe nonlinear transform. Experimental results show that the proposed algorithm andsubjective assessment method have good consistency.