论文部分内容阅读
在H.264的码率控制中,主要通过调整编码参数的量化步长来控制码率输出,改善图像质量。求解量化步长先要通过线性模型,预测当前帧的平均绝对差值。在JVT-G012中,线性模型的系数值通过对数据点的过去统计选择更新,无法体现实时场景的变化。采用线性判决函数对参考数据点分类,综合考虑场景变化和计算复杂度,能合理利用码率资源,平滑图像质量。
In the H.264 rate control, the code rate output is controlled mainly by adjusting the quantization step size of the coding parameters to improve the image quality. The first step in solving the quantization step is to predict the average absolute difference of the current frame by using a linear model. In JVT-G012, the coefficient values of the linear model are updated by the past statistical selection of data points and do not reflect the real-time scene changes. The linear decision function is used to classify the reference data points. Considering the scene change and computational complexity, the rate resources can be rationally used to smooth the image quality.