Research On Full Reference Image Quality Assessment Based On Information Entropy

With the progress of Internet technology and communication technology,digital image plays an irreplaceable role in the field of information transmission.However,the image is damaged inevitably in the process of storage,transmission and processing,which leads to the degradation of image quality.Full reference image quality assessment is widely used in image restoration,image fusion,image recognition and other image processing research,and the effect of image quality assessment is generally good.This thesis studies the full reference image quality assessment from the perspective of information entropy,and tries to extract image features by using information entropy to evaluate image quality,so as to improve the consistency between objective image quality assessment score and human visual perception.Firstly,the information entropy theory is extended to the image field,and the performance of the image entropy theory in the image field is studied.In the experimental study,some conclusions are drawn about the relationship between information entropy and image quality,and two factors affecting the value of information entropy are found.Then,based on the discovery of information entropy,a color image entropy model which can be used to extract color information features of images is proposed,which can make up for the defect that entropy ignores color information of images.Based on this model,a contrast distortion image quality assessment method is proposed,which has a good assessment effect on image quality,and the experimental data show that the assessment score obtained by this method has a high consistency with human perception.In order to enhance the adaptability of this method to images,a framework combining this method with structural similarity was designed,and an image quality assessment method based on color image entropy and structural similarity was proposed.Finally,an image quality assessment method for noise and blur distortion is proposed based on two-dimensional entropy.This method is based on a two-dimensional entropy model to extract information features between images,which can detect the pixel gap between images keenly.The assessment effect of this method is consistent with that of human eyes and its form is innovative,which can provide a new idea for the subsequent study of image quality assessment methods.In this paper,the research results of the relationship between information entropy and image quality can provide a theoretical basis and new ideas for the subsequent application of information entropy in the field of image quality assessment.At the same time,based on the characteristics of information entropy,it is optimized and two kinds of image quality assessment methods are proposed to improve its performance.The image quality assessment score and the perception consistency of human eyes to the image are greatly improved,which can be well applied to remote sensing,medical,biological,industrial and other fields.