Parallel processing method of entropy coding link in hevc based on cpu+gpu heterogeneous platform

Abstract

Provided is an efficient parallel processing method for implementing entropy coding link in HEVC based on a CPU+GPU heterogeneous platform. The method comprises: in a process of encoding a video imagesequence according to a HEVC protocol standard, performing parallel processing on a final entropy coding link of the current frame image and all other links except the final entropy coding link in processing of the next frame image, that is, using a CPU + GPU computing platform, processing prediction, transformation, quantization, rate-distortion optimization, inverse quantization, inverse transformation, filtering and image reconstruction and all other links of the next frame image by the GPU, and the CPU and GPU simultaneously performing parallel computing. Through using the parallel processing scheme, processing time consumed by the link between the above two (that is, a final entropy coding link) which is short in time consumption can be eliminated, so as to significantly increase overall computational speed of a HEVC encoder.

Description

technical field [0001] The invention belongs to the technical field of digital video compression coding, and relates to how to realize efficient parallel processing in the entropy coding link in the HEVC (High Efficiency Video Coding, also known as H.265 or HEVC / H.265) protocol standard method to achieve the purpose of significantly improving the computational efficiency of the HEVC encoder. Background technique [0002] With the rapid development of the Internet and information technology, multimedia technology has played an increasingly important role in social life. As an important information carrier, video has the advantages of intuition, certainty, efficiency and universality, and is widely used in various fields of society. With the continuous improvement of people's demand for video resolution and clarity, the development of digital video has developed from the initial 352×240 resolution to high-definition (1920×1080) and then to ultra-high-definition (4k×2k and abo...

AI Technical Summary

Problems solved by technology

This method can improve the computational efficiency of the HEVC encoder to a certain extent, but because fine-grained parallelism only involves CU-level parallelism, its parallelism is small, and it does not perform parallel processing for the entropy encoding process, so its acceleration effect cannot be achieved. Meet the need for real-time

A paper published in the Journal of Zhejiang University in 2014 (Zhou Chengtao, Tian Xiang, Chen Yaowu. HEVC coding unit size fast selection algorithm [J]. Zhejiang University Journal: Engineering Edition, 2014, 48(8):1451-1460.) proposed a deep The concept of singleness, using the selection of the encoding depth of adjacent units, skips some depths with a relatively small probability of occurrence, thereby speeding up the speed of CU block division. Although this method increases the encoding speed by about 25%, but due to skipping some CU Divide the depth, so that the final encoding quality of the HEVC encoder is affected to a certain extent

However, this method is only slice-level or layer-level parallel, and the degree of parallelism is limited, so it cannot improve its computing efficiency on a larger scale.

The above method does not involve the parallel processing of the final entropy encoding process of the entire frame image after the optimal encoding parameters have been determined.

As mentioned earlier, this link itself takes an average of 14--17 milliseconds when processing a frame of high-definition images (1080p)

Images