Image Quality Assessment Based On Local Invariant Feature

Recently, with the rapid development of computer technology, the popularizationof multimedia technology and network technology, more and more interminal wereappeared which use images as the medium. So image as an indispensable of informationare in our daily life. However, images were unavoidable distoried in the process ofcollecting、saving、compress、 transmitting,Therefore the image quality was damagedin different degree.So to evaluate image quality is a hot topic in the field of the presentstudy. At present, image quality assessment is divided into two methods of subjectiveevaluation and objective evaluation, subjective evaluation method is accurate but thisprocess takes too long and influenced by subjective factors easily, objective evaluationmethod is rapid, but it can’t satisfy the human visual characteristics very well.In view of the above problems，in this paper, doing the following research work:First, According to the theory of local invariant features, combined with thecharacteristics of the human eye is sensitive to image structure characteristics,Programming to extract the local invariant feature points of the image, let its applicationin the field of image quality assessment.Second, based on the structure feature of gray image,According to the principle oflocal invariant features, analysis of the feature points between distoried image andoriginal image feature, a new kind of image quality assessment algorithm based on localinvariant features was proposed.Third, according to the characteristics of the color image and the HVS，using colorinformation of color image and color space conversion, analysis of the feature pointsbetween distoried image and original image feature, a new color image quality assessment algorithm which based on color invariant features descriptor was proposed.Forth，the above two kinds of image quality assessment algorithm has doneexperimental simulation,the experimental results expresses that the evaluate result fromLIFIQAand WSCIQAis consistent with the subjective evaluation result,and canevaluate different types of distorted images,and the evaluation results are better than therecently popular image quality assessment algorithm,these two methods are both localinvariant features a complement in the field of application area and provides a new wayfor image quality evaluation.