Simplified video coding for digital mobile devices

Abstract

Most mobile devices are currently suffering from the lack of computational power and energy-consumption constraints, there is significant interest and research in reducing the computational complexity for fast video coding. This paper proposes a statistical model to predict the zero-quantized DCT (ZQDCT) coefficients for 3-D DCT video coding and thus to speech up the encoding process. First, each 8times8times8 pixel cube at the input of DCT is decomposed into a mean value and a residual 8times8times8 pixel cube. Subsequently, a statistical Laplacian model is mathematically developed to predict the ZQDCT coefficients. Finally, the redundant calculations for these ZQDCT coefficients are skipped. Compared to the baseline encoding method, the proposed model can significantly simplify the complexity and achieve better real-time performance at the expense of negligible visual degradation. Computational reduction also implies longer battery lifetime and energy economy for portable devices.