Motion Vector Calculation Using Block Matching Method with Three-Step Search Algorithm

This document presents a block matching method for motion vector calculation, implementing the classical three-step search algorithm for efficient motion estimation. The methodology includes comprehensive literature references to establish theoretical background and understanding. For threshold determination, we employ the MAD (Mean Absolute Difference) criterion, which calculates the average absolute differences between corresponding blocks in successive frames. The implementation involves dividing images into macroblocks and comparing each block with candidate blocks in the search window using the three-step search pattern, which reduces computational complexity by progressively refining the search area. Preliminary motion vectors between two consecutive images are computed through this process, providing fundamental parameters for subsequent global motion estimation algorithms. These calculated vectors enable accurate analysis of inter-frame motion patterns and establish a foundation for advanced motion processing research. The code structure typically includes functions for block partitioning, search window definition, MAD calculation, and vector refinement through the three-step hierarchy.