Weighted constrained one-bit transform based fast block motion estimation

Block based motion estimation (ME) techniques based on low bit-depth representation have been proposed in the literature to reduce the computational complexity of ME. One-bit transform based ME methods are pioneers in this field. These methods derive one-bit images using an adaptive thresholding scheme, and perform block based ME on one-bit images using simplified matching criteria. Recently, constrained one-bit transform (C-1BT) based ME has been proposed to increase ME performance of 1BT based ME, restraining pixels with values close to the adaptive threshold throughout the matching. In this work, a weighted constraint strategy is proposed to increase the matching performance of C-1BT based ME. Instead of the one-bit constraint mask used in the literature, a two-bit constraint is computed according to the difference between a pixel and its 1BT threshold value, novel to this paper. Experimental results show that the proposed method improves the ME performance of C-1BT, and outperforms other low bit-depth based ME methods at macro-block level.