The H.264 Advanced Video Compression Standard, Second Edition

Chapter 6. H.264 Prediction

Introduction

Perhaps the most important reason for the widespread adoption of H.264/AVC is its compression performance. An H.264 codec, if properly designed and used, can out-perform many other video codecs in terms of compression ratio for a given image quality. Much of the performance gain compared with previous standards is due to H.264/AVC's efficient prediction methods. For every macroblock, a prediction is created, an attempt to duplicate the information contained in the macroblock using previously coded data, and subtracted from the macroblock to form a residual. The efficiency or accuracy of this prediction process has a significant impact on compression performance. An accurate prediction means that the residual contains very little data and this in turn leads to good compression performance.

H.264/AVC supports a wide range of prediction options – intra prediction using data within the current frame, inter prediction using motion compensated prediction from previously coded frames, multiple prediction block sizes, multiple reference frames and special modes such as Direct and Weighted prediction. Together with sub-pixel interpolation and a built-in filter to reduce compression artefacts, these features give an H.264 encoder a great deal of flexibility in the prediction process. By selecting the best prediction options for an individual macroblock, an encoder can minimize the residual size to produce a highly compressed bitstream.

We start with an ...