하다마드 변환의 적응적 적용을 이용한 고속 움직임 예측 알고리즘

초록

In this paper, we propose a new, effective, fast motion estimation algorithms using 4x4 pixels Hadamard transform. The Hadamard transform has the advantage of simplicity because it uses only addition and subtraction. Motion estimation is composed of three stages. First, it should be decided whether to terminate the search early and use a previous motion vector with DC (Direct Current) coefficients. Then the adaptive matching scan order for motion estimation should be determined according to the image complexity using AC (Alternating Current) coefficients. Experimentally, we adapted this algorithms to MVFAST and PMVFAST algorithms, and the proposed algorithms turns out to be very efficient in terms of computational speed while remaining almost the same in terms of PSNR (Peak Signal-to-Noise Ratio) compared to MVFAST and PMVFAST algorithms.