Low complexity recursive search based motion estimation algorithm for video coding applications

A new recursive-search based motion estimation algorithm has been proposed that uses only a maximum of seven candidate motion vectors. The coding quality in terms of signal to noise ratio and compression is quite close to full search motion estimator. The algorithm is extremely light with respect to the computational load. Proposed algorithm improves upon the original three-dimensional recursive search (3DRS) algorithm that is well known as a high performance motion estimator. Results obtained from various test sequences indicate around 50% improvement in computational load over the conventional 3DRS estimator. Simultaneously, the technique provides about 8% improvement in the coded bit-rates. The convergence behavior of our improved algorithm towards variations in the motion vector field is also superior.