Fast video coding at low bit-rates for mobile devices

Abstract

Digital video coding standards such as MPEG-4 simple profile and H.263 are becoming popular in mobile multimedia applications. Mobile multimedia applications demand low power and low processing power. MPEG-4 simple profile and H.263 encoding requires huge amount of processing power, motion estimation being the most complex of the encoding process. Extensive research has led to the development of fast motion estimation algorithms. Fast motion estimation algorithm consumes about 40% of the encoding time as compared to 90% taken by exhaustive search algorithm. In this paper, low bit-rate video data has been analyzed and based upon the statistical characteristics, Discrete Cosine Transform, Quantization, Variable Length Encoding, Inverse Discrete Cosine Transform and Inverse Quantization are optimized. The experiments are conducted on ARM9TDMI core, which is widely used in 2.5G & 3G handheld and mobile devices. The results show that significant gain in encoding speed can be obtained without any subjective quality degradation and negligible objective quality degradation.

Efficient predictive model of zero quantized DCT coefficients for fast video encoding

“…As a consequence, the computational complexity of video encoding is significantly reduced with very little degradation of video quality. In [13], the Laplacian distribution based model in [12] is extendedly applied to the chrominance component in a similar way as the luminance component. In fact, about the DCT coefficients, various studies [14][15][16][17][18][19][20][21][22] have been carried out on their distributions for still images.…”