Efficient Motion Estimation and Mode Decision Algorithms for Advanced Video Coding.

摘要

H.264/AVC video compression standard achieved significant improvements in coding efficiency, but the computational complexity of the H.264/AVC encoder is drastically high. The main complexity of encoder comes from variable block size motion estimation (ME) and rate-distortion optimized (RDO) mode decision methods.;This dissertation proposes three different methods to reduce computation of motion estimation. Firstly, the computation of each distortion measure is reduced by proposing a novel two step edge based partial distortion search (TS-EPDS) algorithm. In this algorithm, the entire macroblock is divided into different sub-blocks and the calculation order of partial distortion is determined based on the edge strength of the sub-blocks. Secondly, we have developed an early termination algorithm that features an adaptive threshold based on the statistical characteristics of rate-distortion (RD) cost regarding current block and previously processed blocks and modes. Thirdly, this dissertation presents a novel adaptive search area selection method by utilizing the information of the previously computed motion vector differences (MVDs).;In H.264/AVC intra coding, DC mode is used to predict regions with no unified direction and the predicted pixel values are same and thus smooth varying regions are not well de-correlated. This dissertation proposes an improved DC prediction (IDCP) mode based on the distance between the predicted and reference pixels. On the other hand, using the nine prediction modes in intra 4x4 and 8x8 block units needs a lot of overhead bits. In order to reduce the number of overhead bits, an intra mode bit rate reduction method is suggested. This dissertation also proposes an enhanced algorithm to estimate the most probable mode (MPM) of each block. The MPM is derived from the prediction mode direction of neighboring blocks which have different weights according to their positions. This dissertation also suggests a fast enhanced cost function for mode decision of intra encoder. The enhanced cost function uses sum of absolute Hadamard-transformed differences (SATD) and mean absolute deviation of the residual block to estimate distortion part of the cost function. A threshold based large coefficients count is also used for estimating the bit-rate part.