Research Of Image Quality Assessment And Restoration System

Image quality assessment and restoration are two fundamental problems on the image processing field, doing research about them, not only has the theoretical value, but also has the its application. In one aspect, image quality assessment techniques can supply standards to evaluate the results of image restoration techniques; In another aspect, the image restoration techniques can improve the quality of the processing images. The image quality assessment and restoration subsystem can be viewed as the basis of the computer image analysis and understanding system, improving the performance of this subsystem means a lot to the whole system. In this paper, some research has been done on the field of image quality assessment and restoration.Firstly, some popular image quality assessment methods are introduced. They are divided into two groups: the subjective ones and the objective ones. From the usage of the reference image, the objective quality assessment methods can be divided into three parts: the full-reference ones(FR),the reduced reference ones(RR) and the non-reference ones(NR). As an example, the full-reference ones are introduced from three classes: the traditional signal processing based ones, the human visual error sensitivity based ones and the human visual region structure similarity ones, to show the present newest development of this field.Secondly, from the aspect of non-reference image quality assessment methods, the causes of image quality degradation are analyzed. Generally, the degradation of the image quality is caused by some kinds of external elements such as noise and blur. Analyzing those causes can give convenience to select the special restoration methods to eliminate those degradations. In this paper, some common causes are introduced and the measures of those causes are also supplied.Thirdly, the present popular image restoration methods are introduced. The restoration of the degraded images is always based on two theories: one is suitable for the lack of image prior knowledge. At this situation, the image restoration can be viewed as an estimation process. According to some certain image quality standard, the restored image can be estimated from the distorted images according to the prior knowledge of noise and degradation functions, this method can also be viewed as a filtering process; the other is suitable for having enough prior knowledge of reference images, then a mathematical model can be built. The restoration will be much more effective if the fitting is done with the model. After that, a new restoration method based on support vector regression (SVR) has been proposed.Finally, the super-resolution image reconstruction is introduced, which is a very effective way to improve the quality of the original images and in this paper, a new super-resolution method based on SVR has been proposed.For every image system, high quality input images are indispensible, in order to get excellent results, which shows the importance of image quality assessment and restoration. In this paper, some generalization and further research based on it have been done, the experiments also show the effectiveness of the methods proposed.