一种通过重建图像压缩率控制量化粒度的方法

发布时间：2018-03-20 10:09

本文选题：量化参数　切入点：量化阶距　出处：《科学技术与工程》2017年32期 　论文类型：期刊论文

【摘要】：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。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,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。
[Abstract]:In order to adapt to different bandwidth constraints of transmission channel, various image coding algorithms control quantization granularity by adjusting the quantization parameter QP. However, there is no explicit relation between the output bit rate of the compression algorithm and the QP value of the quantization parameter. The QP value is difficult to match with the bandwidth constraint. A quantization granularity control method QCRs based on the output bit rate is proposed. The output bit rate of the contractive algorithm is as close to the bandwidth constraint as possible. Because the quantization parameter QP has to be mapped to the quantization step Qstep, the relationship between the quantization step Qstep and the output bit rate of the compression algorithm is first determined. Then the quantization order Qsteps corresponding to 64 given output bit rates are determined by inverse interpolation, and the corresponding quantization tables and inverse quantization tables are modified according to the 64 given output bit rates. Thus, a method of controlling quantization granularity by output bit rate is established. In the experiment part, AVS compression algorithm is adopted, which is also suitable for AVS2.Because the quantization table of AVS contains the fractional part of integer DCT transform, Therefore, when the QCR method is used to adjust the quantization table and inverse quantization table of H.264 / H.265, it must be different from the integer DCT transformation of AVS. The experiment shows that the QCR solution of controlling quantization granularity by reconstructing image compression ratio is effective and feasible.
【作者单位】：太原理工大学信息工程学院;太原科技大学;
【分类号】：TN919.81

【相似文献】