MOTION VECTOR CODING METHOD AND APPARATUS

摘要

PURPOSE: A method for coding motion vector and an apparatus thereof are provided to accurately code motion vector of a search block based on the deviation of motion vector of a reference block. CONSTITUTION: A memory(110) stores motion vectors to each search block by using position information. A reference motion vector selection part(120) determines a reference search block of the present search block based on the position information. A median filter(200) determined median vector based on reference motion vector. A dispersion calculation part(130) calculates horizontal and vertical direction components of the reference motion vector around median vector for supplying to a selection signal generation part(140). The selection signal generation part(140) generates the first or the second selection signal by comparing the sum of horizontal and vertical direction dispersions with a threshold value. If the sum is less than the threshold value, the first selection signal is supplied to switches(150,155) and if the sum is more than the threshold value, the second selection signal is supplied to the switches(150,155). A deviation calculation part(160) calculates the difference between motion vector and the reference motion vectors of the present search block. The minimum deviation selection part(170) determines the minimum difference in horizontal and vertical directions for supplying to a switch(150) and a comparator(180) as the second candidate predictor. The comparator(180) compares the horizontal and vertical direction components of the second candidate predictor with the horizontal and vertical direction components of the first candidate predictor, and provides the comparison result to a header encoder(190). The header encoder(190) generates a flag signal or an equal signal related to each comparison result. The switch(150) selects the optimum predictor in the first candidate predictor and the second candidate predictor for providing to a differential encoder(210). The differential encoder(210) calculates the difference in the horizontal and vertical direction components of the optimum predictor and the present motion vector.