디인터레이싱 기법에 따른 다해상도 고속 계층적 움직임 추정기 = Multi-resolution fast Hierarchical Motion Estimation by de-interlacing Algorithm

다국어 초록 (Multilingual Abstract)

As the search area increased, a full-search algorithm(FSA) requires large amount of computation which causes time delay or very complex architecture for real time VLSI system implementation. In motion picture coding technique, many fast algorithms have been developed. However, these fast algorithms are often designed to merely reduce arithmetic operation without considering their overall performance in VLSI implementation as like silicon area, input/output requirement, and image quality. And it would be maintain the performance with a little PSNR degradation in the reconstructed and predicted images as a compare with the FSA. In this paper, we improved the multi-resolution fast hierarchical motion estimation by using de-interlacing algorithm, that is effective in term of both performance and VLSI implementation, is proposed so as to cover large search area field-based as well as frame based motion estimation(ME). The main concept is to combine both spatial/temporal correlation of motion vector and merits of hierarchical search in a computationally simple way, maintaining a similar rate-distortion performance to that of full search block matching algorithm. There is strong spatial /temporal correlation between the motion vector of adjacent blocks. The applied main idea is to compose a set of candidates from MVs of spatially /temporal adjacent blocks and then choose the best one of them as the initial estimate for further refinement according to a certain rule.
To show the validity of our basic idea, the performance of some experimental sequences were compared with that of full search block matching algorithm. The simulation results shows that the proposed algorithm method, which appropriate VLSI architecture as Multi-resolution fast Hierarchical Motion Estimation(MHME) by using de-interlaced reconstruction frame, provide better performance than the other methods as like same parity reconstruction frame or sub-sampled reconstruction frame.