Abstract:

Complex and time-consuming computation in motion estimation poses a great challenge to the implementation of real-time video coding. In order to improve the real-time performance in motion estimation , the distribution characteristics and spatial correlation of the motion vectors are analyzed , and a unit-rood fast motion estimation algorithm
based on the starting-point prediction is proposed. In the proposed algorithm , the early stop criterion is adopted to predict the starting-point of search by comparing the sum of absolute difference values of the candidate points , and a unit-rood pattem is used to perform the search. Experimental results show that the proposed algorithm, which is of a searching speed of 3 - 17 times that of the Three~Step Search algorithm , 2 - 9 times that of the Diamond Search algorithm and 1. 19 - 4. 42 times that of the Adaptive Rood Pattem Search algorithm , maintains the picture quality well. Moreover , as compared with other motion estimation algorithms , the proposed algorithm is easy to implement and is of less computational complexity , better real-time performance , and performs better in terms of small motion video sequence.

Key words: video coding, motion estimation, searching algorithm, block matching