Encoding a contour of an object in a video signal

摘要

The converter 102 receives current contour information relating to the image and converts it to an extended contour. This information is provided to a match point detector 104 and a motion estimation and compensation unit 120. The motion estimation and compensation unit 120 generates a motion vector for the extended contour based on previous contour data stored in memory 118. The motion vector generated corresponds to a displacement which yields the maximum overlap between the extended contour and the previous contour. The previous contour is then shifted towards the extended contour by the motion vector to produce a predicted contour which is passed to the multiplexer 112 as well as to the match point detector 104. This detector 104 determines the intersection points between the predicted contour and the extended contour which are then passed to the multiplexer 112 and to a deviation detector to produce a set of errors between predicted and extended contours. These are then passed to a transform and quantisation unit 108 from where they are passed to an inverse transform and quantisation unit 114 and a variable length coder 110. The output of the inverse transform and quantisation unit 114 together with the predicted contour is used in contour reconstruction 116 to produce a contour for use with the next contour input.