Motion vector differential coding method with median prediction

摘要

The method involves determining components for vector motion and forming a vector prediction where each component is equal to the median value of components corresponding to at least three candidate movement vectors. A candidate motion vector is a vector which is associated with a previously coded macroblock. The components obtained are subtracted from coded vector components. The current image block (M0) has a macro block of vector motion components. Possible candidate components for vector motion are selected (MV1,MV2,MV3) from various positions around the image position. If the image has not moved, then no action is taken. Otherwise the candidate image motion is examined, and the median value taken. The median value is then differentiated with the prediction, coded and multiplexed.