Computational Complexity Management of H.264/AVC Video Coding Standard

The new H.264/AVC video coding standard achieves significantly improved compression efficiency compared to previous standards by adopting highly advanced and flexible encoding techniques at the expense of increased complexity. However, the high computational complexity of H.264/AVC is a big concern primarily for low-power devices with limited processing capabilities. This thesis presents new techniques to reduce and/or control the computational complexity of an H.264/AVC encoder.;A complexity scalable encoding framework is proposed for controlling the encoder complexity at a macroblock level using a single parameter. The framework uses a special macroblock grouping technique called the "wave-front macroblock scheduling". The computational resources are allocated to the macroblocks within a wave-front. The resource allocation is further developed by adopting the Lagrangian rate-distortion cost prediction into the framework. Results demonstrate significant improvements in the rate-distortion performance of the encoder operating at limited complexity. Finally, the complexity reduction algorithms are installed into the complexity scalable encoding framework. Simulations show that these algorithms equip the complexity scalable encoder with additional complexity control.;These novel algorithms are designed with the target of enabling the H.264/AVC implementations in computationally constrained environments such as the hand-held devices with limited processing capabilities and limited battery life.;A new prediction method is developed to estimate the Lagrangian rate-distortion cost of a macroblock. The prediction method is used in the design of two complexity reduction algorithms for H.264/AVC. The first algorithm uses the predicted rate-distortion costs to identify the SKIP coded macroblocks prior to any INTRA or INTER mode trial. Simulation results show that the algorithm achieves significant complexity savings with negligible loss in rate-distortion performance. Similarly, the second algorithm seeks to further reduce the encoder complexity by using the predicted costs to identify not only SKIP coded but also the INTRA and INTER coded macroblocks at earlier stages. Results indicate greater reductions in the encoder complexity at the expense of slightly larger loss in rate-distortion performance.