MPEG-2 中時間調變模式的運動估測

Abstract

中，為求增加傳輸效率以縮短傳輸時間，和在視訊儲存時， 間，視訊壓 縮就成了當然的選擇。運動估測則是視訊壓縮 重複性以達成高壓縮率的 核心。在MPEG-2的時間調變模式 建影像會被使用作為增益層影像運動估 測時的參考影像，因 像品質將會影響增益層影像運動估測的精確度，進 而影響了 品質和整體編碼位元數。要提高運動估測的準確度，使得預 接 的方法就是增加運動估測時的搜尋範圍，但是計算複雜度 本論文中所提 出的方法，是根據時間調變模式的特性，利用 改進搜尋起始點的方式， 在不增加計算複雜度的情形下，降 差；並在不增加預測誤差的情況下， 減少增益層的計算複雜度。 Video compression is essential in increasing the video transmission efficiency and reducing video storage. In video compression, motion estimation is the kernel to remove temporal redundancy to get high compression gain. The accuracy of motion estimation will affect the quality of reconstructed images and total coding bit-rate. For the temporal scalability mode in MPEG-2, the reconstructed images in base layer can be used as the reference images for the motion estimation in enhancement layer. Therefore, the accuracy of the motion estiamtion in enhancement layer will be affected by the quality of reconstructed images in base layer. To enhance the accuracy of motion estimation and to reduce the prediction error, the straight-forward method is to enlarge the search area of motin estimation; however, this method leads to increased computational complexity. The new method proposed in this thesis is an efficient motion estimation algorithm for the temporal scalability in MPEG-2. The idea behind the new method is to improve the initial search position based on the temporal correlation between base layer pictures and enhancement layer pictures, and hence enlarge the effective search area of motion estimation . Conducted experiments show that the prediction error in base layer will be reduced without increasing computational complexity while the computational complexity in enhancement layer will be reduced without increasing the prediction error.