Research On Methods Of Video Image Assessment And Restoration

With the putting forward of "Smart City", as the "eyes"of a "Smart City’ cameras are ubiquitous. They are capuring the video image information every now and then. Whether the information is we wanted or not, we cannot judge by human only, for the information is countless. Research on methods of video image assessment and restoration based on computer vision has long been the research hotspot in the field of image. In this paper, after the research of image assessment and restoration methods of the research status, the assessment method based on improved model of human visual system has been proposed. Restoration of motion blur image, the dark stripe method is proposed to estimate motion lenth.First of all, degraded image is analyzed in this paper. Due to the change of environment, man-made destruction and the interference of signal, the image may be degraded while capturing. There are two reasons that cause the degrad of image. One is some additional information have increased to the image because the superposition of noise, so that the original information of image have been covered. Another is the information of image have lost because of the interference. Through the analysis of the characteristic of the common types of degraded image, the objective characteristic has been provided to the image assessment.Secondly, this paper analyzes the method of image quality assessment which is based on human visual system. The method extracts the characteristic of image from the point of human visual and obtains the informantion of image subjectively. Because the method does not consider the information characteristic of image sufficiently, the result of assessment is not accurate enough. On the basis of the analysis of the characteristic of degraded image, the human visual system is improved in this paper. The peak signal-to-noise ratio, edge gradient and color characteristic of image is added to the human visual system as the assessment parameters so that the assessment model of image combines the advantages of both human visual characteristics and image information. A large number of image experiment prove that the assessment method presented in this paper can achieve a result which is closer to the subjective assessment. It also prove that the method has the ability to identify the common types of degraded image.Finally, in the aspect of image restoration, the characteristic of FFT of motion blur image is analyzed in this paper. Due to the additivity of motion blur image, there are some light and dark stripes on the spectrum of motion blur image and the direction of stripes is perpendicular to the direction of motion, so that the problem of get direction of motion turn to the problem of straight angle. The direction of motion is obtained by the direction of stripes which is calculated by Radon transformation. On the method of calculate movement distance, the decomposition of the Fourier expression of point spread function shows that the function has the periodic zero. This is why dark stripes appear in motion blur image. Because of the dark stripes which is a column of pixeis is easier to obtain than the light stripes which have width and the position of dark stripes appeared is related to the distance of motion, this paper obtains the motion blur distance by calculate the positon of dark stripes.The test result shows that the metod is easy to realize and has a good restoration effect.