E. Ojimi, K. Fukuda, and A. Kawanaka (Japan)
arbitrarily shaped video object, variable-size block motion compensation, rate-distortion optimization, video coding
We propose a coding method for arbitrarily shaped video objects with variable block-size motion compensation. Although the block matching algorithm (BMA) with fixed block size is widely used for motion compensation, this method has several drawbacks, such as unreliable motion estimation for cases in which regions with different movements exist in a fixed-sized block. Therefore, based on rate-distortion properties, our estimation algorithm yields an adaptive block model. Motion vectors are modified so that a Lagrangian-type cost function with respect to the motion compensation errors and the bit rate for motion vectors will be minimized. Moreover, the method determines whether the motion compensated prediction error for each block is coded or not according to bit rate. The experiments show that the proposed method gives results that are subjectively and objectively better than the coding method that uses a fixed block-size BMA. The proposed method provides a gain of 2 dB compared to the coding method with the conventional BMA at low bit rates.
Important Links:
Go Back
