Verfahren zur Verbesserung der Qualität von Bewegungsvektoren für digitale Bildübertragung

摘要

2.1 The method is intended to enable identification of error-affected movement vectors without the need to perform extensive computing operations. Movement vectors which have an implausible value are to be replaced by a movement vector which has been determined through comparison and which very closely approximates the correct value. 2.2 The determined movement vector of an image block is compared with the movement vectors of the adjacent image blocks and checked for plausibility. The values of the movement vectors originating in adjacent blocks in each case with approximately the same amounts and the same direction are combined separately and a weighting is carried out which is dependent on the proximity relationship and on the similarity of the movement vectors. Mean values are formed from the amounts and directions of the movement vectors classified in this way and a weighting is allocated to them. The mean movement vector to which the greatest weight is allocated is used for comparison with the movement vector of the image block just processed. This mean movement vector is used as a replacement for a movement vector of the respective image block under consideration which deviates too far from plausible values. 2.3 An image subdivided into image blocks is processed in such a way that implausible movement vectors are eliminated. This leads to an increase in image quality in narrowband digital image transmission.