H.264 is a powerful video compression standard using advanced spatial and temporal encoding techniques. Compared to the MPEG4 Part 2 standard, H.264 achieves the same visual quality at half of the bitrate.
The coding efficiency of H.264 makes it well suitable for quality- and time-critical video applications such as low-latency video transmission.
This work concentrates on a part of the H.264 coding process that is used in many of these applications - the spatial (Intra) prediction.
This coding tool propagates pixels in a frame for predicting unknown neighbouring pixel regions. H.264 supports multiple propagation patterns which are referred to as coding modes. The challenge of most real-time video encoders is to choose an efficient coding mode in a reasonable amount of time. This process is called coding mode selection (CMS) and involves the computation of all possible coding modes and choosing the best one.
In this thesis, we focus on reducing the computational complexity of the CMS. First, we investigate the behaviour of the CMS on a set of test sequences. The impact on the coding efficiency when using only a reduced number of coding modes is analyzed. Second, we develop two methods for speeding up the CMS. Both methods skip the computation of coding modes that are unlikely to improve the coding efficiency.
This results in a significant reduction in the computational complexity of the encoder. For each of the two proposed CMS algorithms, we evaluate the impact on the coding efficiency, the runtime reduction and the quality. We demonstrate that they reduce the computational complexity of the CMS with no signicant degradation of the bitrate and quality.