一种改进的TMN8率失真模型

The Improved TMN8 R-D Model

在视频压缩中,码率控制是必不可少的.通常码率控制算法使用一定的率失真模型来决策量化参数从而控制码率.H.263视频编码标准使用传统的基于DCT系数一致分布假设的TMN8率失真模型,但实际上DCT系数更接近拉普拉斯分布.为此,在基于DCT系数符合拉普拉斯分布的假设下,改进了传统的TMN8模型,得到了一种新的率失真模型.实验证明,这种改进的率失真模型在码率控制的精度和编码效率上都有较好的效果.

In the video compression, rate control plays an important role. Generally, the rate distortion model is employed to compute quantization parameter for the controlling bit rate. In the H. 263 standard, the rate control algorithm employs the traditional TMN8 rate distortion model with the assumption of uniform distribution of DCT coefficients. In this paper, the traditional TMN8 R-D model is improved based on the assumption of the Laplacian distribution of the residual coefficients. The experiments show that the proposed R-D model is better than the traditional one both in the bit rate controlling and coding efficiency.