The Research Of Video Restoration Technology Based On Film Archived

Video, as a kind of multimedia, is one of the major information obtained by people from outside. Meanwhile, vivid and concrete video archive provides the important source for our understanding of history as their unique form. However, affected by the limitation of filming condition or some factors, for instance, improper conservation and copy, many existing videos, especially the celluloid films, are in a fragile state because of various distorsion. It is extremely urgent to conserve and restore the video archive.With the development of computer soft hardware and the digital processing technology, digital processing approach is introduced to the problem of video restoration with its advantages such as convenient for conservation and no need to restore repeatedly. Simultaneously, due to the particularity of the video information's structure, the restoration of digital video becomes one the tough and hotspot research in digital processing field. There are a great variety of distorsions; among them the conventional ones are blotches, intensity flicker, scratch lines and unsteadiness. On account of the different causing reasons, the restoration method to each distorsion is varied. So the research on the resoration of video archive is considered as the main content of this dissertation, where the detection of blotches, the restoration of larger-scale impair, the correction of intensity flicker and the detection of scratch line are the key points of research:1. Based on the the analysis of the blotches'characteristic in the video archive and the existing blotch detection methods in video sequences, a blotch detection method based on texture matching and dual-time stepping adaptive threshold is developed. The false alarms are reduced through the introduction of texture matching and the correct detection results are improved by the employment of dual-time stepping adaprive threshold method, which achieves the enhancement of the ratio between the correct detection results and false alarms.2. Through the research of the image completion algorithms, a fast image completion method based on texture matching which can be applied to the restoration of lage-scale impair in videos is proposed. In this method, successive eliminate algorithms and texture matching based on the down-sampling multi-resolution images are adopted, which improves the efficiency in the region filling process. This method is a highly-efficient and reliable one to restore the lager blotches in the video achive.3. By analysising the rationality and feasibility of linear and nonlinear models in the restoration methods of intensity flicker, an intensity flicker correction method based on motion compensation and threshold method is developed. In this method, three steps search algorithm is employed to estimate motion vectors, which can detect motion regions more accurately. Besides, threshold method is used to detect flicker which can avoid the correction error in non-flicker regions, and the efficiency is improved. The proposed algorithm can preserve the wanted effects with similar features such as fade in and fade out.4. Considering the characteristic of scratch line, detection method based on time-space domain character and canny operator is proposed. This method take advantage of the discontinuous of the image's local feature in the edge of the images where the diversification of gray value is obviously, as usual we say the character of the sigal which is anomalously changed is used to detect and locate the scrach line. Meanwhile, the employment of canny operator enhance the correct detection rate by obtain the length, start point and end point of scratch line accurately.