基于重建图像压缩率的量化方法

为了适应传输通道不同的带宽限制,目前各种图像编码算法通过调整量化参数QP的值来控制量化粒度,但是压缩算法的输出码率与量化参数QP值并没有显式的关系,造成QP值的选取很难与带宽约束相匹配。本文提出一种依据输出码率来控制量化粒度的方法QCR,可使压缩算法的输出码率尽可能贴近带宽约束。因为量化参数QP要映射成量化阶距Qstep,本文首先确定压缩算法的量化阶距Qstep与输出码率的关系,然后通过逆向插值确定64个给定的输出码率所对应的量化阶距Qstep,并根据这64个给定的输出码率修改相应的量化表和反量化表,从而建立了利用输出码率控制量化粒度的方法。实验部分采用AVS压缩算法,同样也适用于AVS2。由于AVS的量化表中含有整数DCT变换的小数部分,因此QCR方法用于调整H.264/H.265的量化表和反量化表时,必须与AVS的整数DCT变换有所不同。实验表明本文通过重建图像压缩率控制量化粒度的QCR解决方案是有效可行的。

A quantization method based on reconstructed image compression Ratio

In order to adapt to the different bandwidth limitation of the transmission channel, various image coding algorithms are used to control the quantization granularity by adjusting the value of the quantization parameter QP. However, the output bit rate of the compression algorithm does not have the explicit relationship with the QP value , the selection of quantization parameter is difficult to match with bandwidth constraints. In this paper, we propose a method QCR to control the quantization granularity based on the output bit rate, which can make the output bit rate of the compression algorithm as close to the bandwidth constraint as possible. Since the quantization parameter QP is mapped to the quantization step Qstep, this paper determines the relationship between the quantization step Qstep and the output bit rate of the compression algorithm first, and then determines the 64 quantization step Qstep corresponding to given output bit rates by inverse interpolation, And the corresponding quantization table and inverse quantization table are modified according to the 64 given output bit rates, thus establishing the method of controling quantization granularity by using the output bit rate The experimental part uses AVS compression algorithm, also applies to AVS2. Since the quantization table of the AVS contains the fractional part of the integer DCT transform, the QCR method must be different from the integer DCT transform of the AVS when adjusting the quantization table and the inverse quantization table of H.264 / H.265. Experiments show that the QCR solution of this paper that control the quantization granularity by reconstruct the image compression rate is feasible