| The structural similarity image quality assessment(SSIM) models the main function ofthe HVS is to extract structural information from viewing fields. As the SSIM is simple andefficient, many scholars have in-depth study on it, and propose variety of improved methodsbased on SSIM. After study the SSIM and its advancements on gray image, color image,video image quality assessment, the work of this paper is show as follows:This paper introduces the KL transform to collect the main characters of the structureinformation, and avoid the interfere of some characters that do not contribute to HVS. Afterthe KL transform, the wavelet transform is used to simulate the multi-channel of the HVS. Aseach channel has its own signal character, this paper calculates the edge comparison and thetexture comparison in the high frequency component with the Harris Responds and theGray-level Co-occurrence Matrix. The contrast comparison is calculated in the low frequencycomponents, and combined with a special rule.This paper applies the new image quality assessment to the gray image, the color image,and the video image. For the gray image, the new assessment is used to measure the quality ofthe gray images in the LIVE database. For the color image, this paper introduces the colorcomparison into the new assessment. According to the character information of each colorspace, the new assessment is fused to the color comparison in different way. For the videoimage, the assessment model has three levels: the local level, the frame level, and the videolevel. We use the new assessment to calculate the quality score of the image locals, andcombine the local scores with their information to get the frame scores, finally, we togetherthe frame scores with their move vectors and edge information to get the video score. Theexperiment result shows that the new assessment is closer to the human perception than otherimage quality assessments for the gray image, the color image, and the video image. |
PDF Full Text Request