Fast block matching algorithms for motion estimation

Abstract

Motion compensation is an effective method for reducing temporal redundancy found in video sequence compression. However, the complexity of the full-search block matching algorithm (BMA) is extremely high and a number of fast algorithms have been proposed to reduce the computational complexity of the BMA. In this paper, we propose three techniques for reducing the arithmetic complexity of the BMA. They are namely: 1) Domain Decimation, 2) Error Function Subsampling, and 3) Multiple Candidates Search. The domain decimation and error function decimation try to reduce the arithmetic complexity by reducing the number of operations in each comparison, the search range and number of locations searched. The multiple candidates search is a technique used to increase the robust of such algorithms. Combination of these techniques can generate algorithms with different tradeoff between arithmetic complexity and quality of prediction.