Associate Professor Maja Bystrom
Information Systems and Sciences Laboratory

The new H.264/AVC video coding standard can deliver significantly improved compression efficiency compared with previous standards, supporting higher quality video over lower bit rate channels. Due to this improved compression efficiency and increased flexibility of coding and transmission, H.264 has the potential to enable new video services such as mobile video phones and multimedia streaming over mobile networks. However, the performance gains of H.264 come at a price of increased computational complexity. The processing overhead required to implement H.264 is likely to be a major problem for power constrained devices such as camera-phones and mobile monitoring devices. There is a need to develop low complexity implementations of H.264/AVC that offer the performance and flexibility advantages of the standard without an excessive computational cost.

We continue to develop algorithms that can significantly reduce computation in an H.264 encoder without any significant loss of rate-distortion performance as compared with a “baseline” non-optimized encoder. The new algorithm is based on well-known rate distortion optimization methods and enables an H.264 encoder to decide whether to encode or “skip” each coding unit (termed a macroblock), prior to processing the macroblock data. The algorithm estimates rate-distortion costs and employs a model of the expected rate-distortion cost of skipping versus coding the macroblock to determine whether the macroblock should be coded. Computational complexity reduction is achieved by avoiding motion estimation and further processing stages for the “skipped” macroblocks. Results shown in the following table illustrate that significant savings can be achieved, at little quality penalty, particularly for low-motion video sequences.

Coding time of the proposed method relative to the baseline encoder and a rate distortion optimized encoder for four sequences. Savings in computation of up to 67% is achieved.

Sequence name	Reduced complexity encoder	Baseline encoder	RD Optimized encoder
Carphone	0.64 - 0.73	1.0	1.54 - 1.77
Foreman	0.71 - 0.81	1.0	1.56 - 1.87
Claire	0.33 - 0.38	1.0	1.5 - 1.6
Mother & Daughter	0.46 - 0.47	1.0	1.51 - 1.69

Above: PSNR of representative  video sequences using the baseline codecs and the proposed method. It can be observed that the proposed complexity-reducing method can even outperform the baseline H.264 codec.