New dynamic pattern search based fast motion estimation algorithm

In the development of fast block based motion estimation (BME) algorithms, the focus is always on reduction of computational burden with quality as good as that of Full Search algorithm. Fast fixed search BME algorithms like TSS, DS etc. have been proposed in the literature but these suffer from over or under search for slow or fast motion video sequences. This problem is eradicated by using the coherence of the neighboring blocks to predict the motion of current block. A new dynamic pattern search algorithm for fast BME is proposed in this manuscript, which uses the coherence of temporal right neighboring block along with the spatial and temporal left neighboring blocks and dynamically adapts its search pattern for motion vector estimation of the candidate block. Experimental results show that the proposed algorithm results in improvement in PSNR by 0.2116dB, 0.5043dB, and 2.0160dB with only 1.027, 1.058 and 1.43 times intensification in number of search points as compared to ARPS, DPS and MDPS algorithms respectively. Further proposed algorithm shows better bit compression ratio by 1.0047, 1.0014 and 1.0058, 1.0354 times and also better structural similarity index measurement by 1.0107, 1.0010, 1.0112 and 1.0376 times compared to DS, ARPS, DPS and MDPS algorithms respectively.