No-Reference Quality Assessment Algorithm For Compressed Images

Visual information is the dominant income formation for human.Images are usually used to store and deliver visual information.The wide application of images has induced the dramatically increasing of data,which results in the problems of storing,transmission and application.Compression is the most effective solution to reduce the image data size although it brings distortion which affects the human visual experience.In order to study the relationship between the change of human visual experience and the change of image quality,it is necessary to design efficient image quality assessment method to reasonably evaluate image quality.Image quality assessment methods can be divided into the subjective quality assessment method and the objective quality assessment method.The subjective quality assessment needs lots of participants and resources.However,its assessment results are often affected by the human subjective decision and not satisfied.The objective quality assessment method is implemented with smaller number of resources and not affected by the subjective factors.Meanwhile,its assessment results can be reproduced.Therefore,the objective quality assessment method is widely studied.Based on the study of a variety of features related to compressed image distortion,this thesis uses feature fusion to form features that can be effectively used for objective quality evaluation,and has achieved good results after application.Its contributions include:(1)According to the relationship between the block artifact of the compressed image and the change of image gradient,a new no reference image quality assessment method based on the change of directional gradient was proposed.To implement this method,the first-order directional gradient,the second-order directional gradient and the directional change of the image were studied.These features were fused together and then used to construct a quality assessment model which achieves a high assessment efficiency for the compressed images.(2)According to the relationship between the block artifact of image compression and the local energy distribution,an energy distribution-based no reference image quality assessment method was proposed.In this method,the relationship between image block energy change and block edge feature was studied,which is used to guide the construction of the quality assessment model based on the change of image block energy distribution.With this proposed method,an accurate quality assessment for the compressed images is also achieved.