An Efficient Multilevel Transform-Domain Partial Distortion Search Algorithm
Kiran Kumar Vemula, S. Neeraja · Pattern Recognition and Image Analysis · 2022
Abstract This paper proposes a fast partial distortion-based block matching motion estimation algorithm in Walsh–Hadamard. The proposed algorithm divides the current block into subblocks and provides a sequence of fine-partial distortions to reject the impossible candidates using as less calculation as possible. In contrast to the previous algorithms in the Walsh–Hadamard domain, such as transform-domain successive elimination algorithm (TSEA) and multilevel transform-domain partial distortion search (MT-PDS) algorithm; this algorithm applies Walsh–Hadamard Transform (WHT) to only those subblocks that demand WHT instead of entire block or all subblocks. In addition, unlike in the TSEA and MT-PDS algorithms, the number of additional transform coefficients required to calculate partial distortion at any level is constant and small. The simulation results show that the proposed algorithm reduces the computational cost of TSEA and MT‑PDS algorithms while maintaining the motion prediction quality. Compared with the full search, 94.95% of the computational complexity is reduced by the proposed algorithm without any reduction in the motion prediction quality. Similarly, the proposed algorithm reduces the computational complexity of the TSEA and MT-PDS algorithms by 28.88 and 42.31% respectively without any loss in the motion prediction quality.