| Summary: | 碩士 === 義守大學 === 資訊工程學系碩士班 === 97 === Nowadays, the technology of video compression is a very important technique for digital video data transmission and storage. In general, we use block motion estimation in MPEG4 standard. In order to obtain higher performance of encoding, H.264/AVC adopts seven block modes for motion estimation. However, the mode decision of H.264/AVC is too complex that influence the whole algorithm to perform video encoding inefficiently.
In this thesis, we proposed a novel and fast scheme for block mode decision that adopts macro and micro block types as baseline. According to the RD-Cost, the macro or micro block types can be classified. Furthermore, we use early termination to reduce the complexity of encoding. In the experiments, we combined the proposed mode decision scheme and the fast motion estimation algorithm, and then applied Kalman Filter to refine motion vector. From the observation of experimental results, our proposed scheme indeed improves the performance of encoding rate and preserves video quality simultaneously.
|
|---|
