A Linear Code Rate Estimation Method Based on Element Category

Abstract

The present invention provides a linear code rate estimation method based on element category, which relates to the technical field of video coding. Firstly, the coefficients are divided into two groups of luma and chrominance, and then respectively in the luma coding group and the chroma coding group according to the element type and The TU size is further divided, and then the characteristic parameters are selected in each group, and the linear regression equation between the estimated bit rate and the parameters is established, and finally the video sequence is selected for training to obtain the linear regression coefficients of each group. The invention eliminates the operational redundancy caused by the coding structure in the code rate calculation, reduces the complexity of the code rate calculation, carefully groups the syntax elements, and establishes a linear regression equation in each group to estimate the code rate, with higher accuracy. High, which effectively reduces the complexity of RDO bit rate calculation and improves coding efficiency; the bit rate estimation accuracy is high, and the coding quality loss is small; the robustness is strong, and it has better optimization effects in different video sequences.

Description

technical field [0001] The invention relates to the technical field of video coding, in particular to a code rate estimation method. Background technique [0002] HEVC uses Rate Distortion Optimization (RDO) to select the mode with the best combination of compression rate and coding quality. This process mainly includes the calculation of distortion and bit rate. The former needs to be transformed, quantized, inversely transformed, and inversely quantized. The difference between the reconstructed video and the original video, i.e. distortion; the latter requires entropy coding of the syntax elements involved to derive the bitrate consumed. Therefore, the RDO process is very cumbersome, and simplifying the operation of the process is of great significance for reducing the complexity of HEVC coding. Research in this area in domestic and foreign academic and industrial circles mainly focuses on the simplification of the bit rate calculation process, especially for context-based...

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the deficiencies of the existing technology and solve the shortcomings of previous code rate estimation algorithms such as low prediction accuracy, poor robustness, and low estimation efficiency, the present invention proposes a linear code rate estimation model based on the type of syntax element, using the transformation unit (Transform Unit, TU) is the basic processing unit, which performs fast code rate estimation for the quantized coefficients. According to the characteristics of using different context probability models under different encoding conditions, the coefficients are first divided into luma and chrominance groups, and then In the luminance coding group and the chroma coding group, it is further divided according to the element type and TU size, and then the characteristic parameters are selected in each group, and the linear regression equation between the estimated code rate and these parameters is established, and finally the complexity is high. The PeopleOnStreet video sequence is trained to obtain the linear regression coefficients of each group

Images