A fast RDOQ quantization zero coefficient proportional estimation method

Abstract

The invention discloses a fast RDOQ quantization zero coefficient proportion estimation method. In order to embody the difference of the different position coefficients, the invention defines a different weighting coefficient for the different position coefficients, constructs the weighting coefficient as a function of the position adaptive dead-time offset delta, and weights SATD, namely WSATD, based on this metric. Considering the complexity of RDOQ algorithm, only considering q and WSATD is not enough to construct accurate Rho model. Assuming Chi = WSATD / q, it is found that different Chi samples have different values of rho, and the invention takes the mean value theta of Chi as another characteristic parameter for constructing rho model. In the HEVC encoder, there are different types of transform blocks (TU), and in order to distinguish this difference, the present invention models the different types of TU blocks separately. Finally, the three-dimensional Rho model is constructedby using the statistical analysis curve fitting method. Through this model, the zero coefficient ratio Rho after RDOQ can be estimated quickly.

Description

technical field [0001] The invention belongs to the field of video encoders, in particular to a fast RDOQ quantization zero coefficient ratio estimation method. Background technique [0002] In the process of video encoder design, the rate-distortion (R-D) function model is often used in algorithmic rate-distortion optimization (RDO), such as rate control. Computing R and D accurately will increase the computational complexity of video encoders, especially in the latest HEVC standard encoders that support computationally intensive RDOQ and CABAC entropy coding techniques. The development of fast estimation methods for R and D has aroused extensive research interest in the past two decades, and several typical R-D models have been proposed in academia and industry. [0003] Since quantization directly determines the encoding distortion and bit rate consumption, some scholars describe the rate-distortion model as a function of the quantization parameter QP. This type of metho...

AI Technical Summary

Problems solved by technology

[0015] The all-zero block pre-decision method proposed in this paper is based on the hard-decision quantization (HDQ) algorithm on the one hand, so it does not support the soft-decision quantization (SDQ) algorithm with good coding performance, such as RDOQ; on the other hand, the method It is an all-zero block pre-judgment method at the macroblock level, which is not applicable to the fine-grained TU block level

Images