GPU decoding method for multipath h.264 video conference

Abstract

The invention discloses a GPU decoding method for a multipath h.264 video conference. The method comprises the steps of (S1) inputting multipath video conference stream through MCUs of various branch meeting places; (S2) distinguishing each meeting place according to IP information in the video stream; (S3) decoding each meeting place by adopting a CPU and GPU master-slave thread cooperative way separately; (S4) storing the decoded data on a server; (S5) recoding the decoded data for a request of a client and transmitting the data to the client; and (S6) displaying statistical information and the video stream transmitted by the server by the client. The problems of real-time transmission, decoding and playing of multipath video bitstream during meetings in a plurality of meeting places at the same time are solved, mutual waiting state of the CPU and the GPU is prohibited, simultaneous parallel working of the CPU and the GPU is achieved, and the decoded image is recoded and stored to be played at the client, so that the usability and the functionality are expanded and the GPU decoding method has relatively high popularization and application value.

Description

technical field [0001] The invention relates to the technical field of video conferences, in particular to a GPU decoding method for multi-channel h.264 video conferences. Background technique [0002] The H.264 video codec standard jointly proposed by ITU-T and IOS / IEC in 2003 is a practical video codec standard with the best compression performance in the field of video compression today. Compared with previous international standards such as H.263 and MPEG-4, the biggest advantages are reflected in the following four aspects: 1. Separate each video frame into blocks composed of pixels, so the encoding process of video frames can be Reach the block level; 2. Use the method of spatial redundancy to perform spatial prediction, conversion, optimization and entropy coding (variable length coding) on ​​some original blocks of the video frame; 3. Use temporary storage for different blocks of consecutive frames In this way, only the changed parts in consecutive frames need to be...

AI Technical Summary

Problems solved by technology

For example, the MMX / SSE instruction set provided by Intel for Pentium series processors uses the CPU with SIMD processing mechanism for video encoding and decoding, which has achieved a certain acceleration effect, but the problem of CPU ultra-high load operation has not been solved, so the system finishing still underutilized

In addition, some researchers use DSP chips and other hardware circuits to realize the acceleration of video encoding and decoding, and some experiments have achieved good results, but due to reasons such as expensive hardware equipment, they cannot be popularized and used.

[0006] In view of this, it is necessary to propose a new method to solve the problem of improving the decoding speed of multiple real-time video streams

Images