Verfahren und Vorrichtung zum Kodieren eines Bewegungsvektors

摘要

A current motion vector CMV of a search block is encoded based on a plurality of reference motion vectors RMV's. A first and a second candidate predictors FCP and SCP having first components FCP-1 and SCP-1 and second components FCP-2 and SCP-2 are determined. The FCP-1 and the FCP-2 represent medians of the first and the second components RMV-1's and RMV-2's of the RMV's, respectively, while the SCP-1 and the SCP-2 represent a RMV-1 and a RMV-2 which yield a least first and a second absolute differences with respect to a first and a second components CMV-1 and CMV-2 of the CMV, respectively. A dispersion value of the RMV's is compared with a predetermined threshold to thereby generate a first selection signal if the dispersion value is less than the threshold and a second selection signal if otherwise. A third candidate predictor TCP having a first and a second components TCP-1 and TCP-2 is determined based on the number of encoded bits generated by encoding the CMV-1 with respect to each of the FCP-1 and the SCP-1 and the number of encoded bits generated by encoding the CMV-2 with respect to each of the FCP-2 and the SCP-2, wherein the TCP-1 is one of the FCP-1 and the SCP-1 and the TCP-2 is either the FCP-2 or the SCP-2. The CMV is encoded based on the FCP and the TCP in response to the first and the second selection signals, respectively, thereby generating encoded motion vector data for the CMV.