Low-rate Video Coding Method Based On Block Motion Estimation

In the past few years, there has been significant interest in digital video applications like video conferencing and video e-mailing over Internet, video telephony over public switched telephone networks (PSTN) and wireless networks, etc.By removing spatio-temporal redundancies existing in adjacent frames, motion estimation can reduce the coding bit-rate significantly. The block matching algorithm (BMA) is used as a motion estimation method for most of the video coding systems. Its goal is to find a block that is most similar to a current block within a pre-defined search area in a reference frame.As a straightforward method, the full search algorithm (FS) is widely used because of its high performance .Usually, FS is computationally expensive in a video encoder. As a result, in order to reduce the heavy computational load of FS, active research has focused on fast BMAs for a long time and several sub-optimal search algorithms for block-based motion estimation have been developed. These include: Three Step Search (TSS), Four Step Search (FSS), Two-D Logarithmic Search, Orthogonal Search, Cross Search ,etc. They are the conventional BMAs, a significant local minimum phenomenon occurs.In this paper, a fast multi-resolution adaptive motion estimation algorithm is proposed here which using multiple motion vector (MV) candidates according to the spatial-temporal correlation in MV fields.The choice of an algorithm for motion estimation is governed by the 'speed-quality-bitrate' tradeoff. A good algorithm is one that has a low computational complexity, provides a high quality of motion compensation and also ensures that the bitstream is as small as possible. Previous results in literatures do not always consider all the three factors.The multi-resolution schemes are based on the idea of predicting an initial estimate at the coarse level and refining the estimate at the fine level. This helps in reducing the search space.The proposed algorithm consists of three resolution levels. Level numbers are ordered from 0 to 2, and level 0 and level 2 represent the finest and coarsest level, respectively.The algorithm limits MV candidates to only three at the middle level, and describes the algorithm with three resolution levels. At the coarsest level, two MV candidates are obtained on the basis of minimum matching error for the next search level. At the middle level, the two candidates selected at the coarsest level and the other one based on spatial MV correlation at the finest level are used as center points for local searches, and a MV candidate is chosen for the next search level. Then, at the finest level, the final MV is obtained from local search around the single candidate obtained at the middle level. Thereby, a single candidate at the final-level search is enough to provide the desired performance. As a result, the algorithm performs only one localsearch at the finest level, and its overall computational cost and data bandwidth burden decrease.According to different motion area, the search methods dynamically are also introduced .The algorithm introduced in this paper provides prospective PSNR performance that is close to the FSBMA.The algorithm tries to achieve better performance in terms of all three parameters in the speed-quality-bitrate tradeoff, so satisfies high estimation performance and efficient implementation and can be used for real time low bit-rate video coding.