Research On Image Objective Quality

As electronic information technology develops rapidly as well as digital camera, personal computer, network communication, multimedia technology and so on in the twenty-first century, the instruments, which are used for collecting, showing, transmitting and accumulating images, become more and more sophisticated, and the quality of modern images is increasing. However, images are lack of fidelity during the manipulation of every step. As the important part of human visual information, the research of the quality evaluation of image information has become an important subject in the field of image information project. Because all the time the direct terminal of image processing is human visual system, basic reasonable evaluation methods of the quality of images are all based on subjective evaluation, but the methods of subjective evaluation are limited, which is vulnerable to the influence of many objective factors. In addition to that, the result seems like impression evaluation of an image. It is vague and hard to make a suggestion of improving the quality of the image. It is not conducive to being widely used in the image project. Therefore, scholars and research institutions mainly focus on objective evaluation methods of the quality of images.In this dissertation, firstly I introduce the relevant knowledge of the human visual system. Then the key points are to research the classic computational method of the quality of images——MSSIM computational method that is based on the similarity of image structure and verify simulate results that are produced by evaluating different types of distortion images and evaluation values that are better than MSE computational method. Later, in view of its inaccuracy of evaluating distortion images caused by vague and noise, I briefly introduce several kinds of improving computational methods that are base on MSSIM computational method, including the train of thought and the analysis of the upper hand. Last, I put forward an improving computational method that is based on wavelet transformation and HVS. Through simulate results on the platform of MATLAB, it is verified that its evaluation results are more accurate and more close to human subjective feelings in the process of evaluating distortion images caused by vague and noise.