Enhancement of digitised analogue video recordings using multiple image processing algorithms

This thesis deals with the enhancement of image quality in analogue videotape recordings from closed-circuit television cameras. The severity and frequency of twenty image-degrading factors was investigated. In order of ranking, the six most frequent problems were found to be blurring, jitter, local motion, illumination variation, low contrast and media faults. Both existing and improved algorithms for dealing with these problems are described. Most of the thesis contains an analysis of both existing and new algorithms for dealing with the worst problems. In general, it was found that only small improvements could be obtained with blurred images. However, much could be done to overcome the effect of line-to-line jitter, and a large portion of the thesis is devoted to existing and to new dejittering algorithms. A new method is described that is some two thousand times faster than the best previously reported dejittering algorithm. An extensive comparison of contrast enhancement methods was also carried out; media faults were treated with noise-reducing algorithms; and speckle noise, electrical noise and signal dropout were also examined. The best eight algorithms were then combined sequentially in all permutations of length three, and in some cases length four, in order to find those processing sequences that produced the most improvement in visual image quality. Guidelines are provided for the most appropriate order in which to apply the individual algorithms. It was found that significant improvements in image quality could be obtained in a limited number of cases - specifically for image jitter and illumination compensation - but that in general only modest gains were achievable