Research Of Old Film Video Image Quality Assessment Method

The old movie information is valuable human material heritage. However, a number of reasons make the most of the existing old video different degrees of damage, Very large number of movie films need to be repaired urgently, and the completion of the repair work has always been extremely time-consuming, laborious and huge amount of work.In order to improve the repair efficiency, the damaged old film video need to be classified according to the damages. The main work of this thesis is to study the rapid method of damage detection for screening the films which are really need to repair.The three kinds of common damages occurred on old films are blocks, scratches and flicker. The main studies of this thesis are as follows:Firstly, the features of damaged blocks and their mathematical model are examined. Several blocks detection algorithm in the video repair process are summarized. After that, the intuitive blocks detection algorithm is improved and the regional blocks detection algorithm is designed according to the discontinuity at time and the spatial gray consistency in nature. The test results show that the algorithm can quickly detect the blocks that meet a certain area and avoid use of complex motion estimation algorithm.Secondly, the characteristics and the reasons of formation of the scratch damage and its classic mathematical model are studied. Several classic scratch detection algorithms are outlined and the scratch detection algorithm based on grayscale projection in the vertical direction is proposed. Also the distinction between natural objects and scratches is made based on the scratches grayscale’s inconsistency and natural object grayscale’s consistency in the vertical direction. Experimental results show that the algorithm can quickly detect the long vertical scratches and can distinguish real scratches from natural objects.Finally, the flicker of old film videos, its forming reasons and mathematical description model are studied. According to the flicker existing time period, the interframe mean is selected as detection characteristics. Through experiment comparison and analysis, the law of the image mean value and its Fourier transform in the two cases is found. Through research and analysis of experimental data, the flicker detection method based on Fourier analysis is given. The method can quickly determine whether there is flicker or not and the algorithm is simple and effective.