A method for large-format multi-layer real-time editing based on intra-frame segmentation and clustering

Abstract

The invention discloses a method for realizing large-scale multi-layer real-time editing based on clustering of intra-frame segmentation, and relates to the technical field of video editing. The invention includes the following steps: S1: using multiple machines to form an InfiniBand network cluster, each machine Configure the InfiniBand network card; S2: Set each machine to have the same timeline information, so that the timeline data structures of all machines are synchronized; S3: Decode and edit the input video file to obtain the decoded video data, and then press the decoded video data The space is divided into multiple areas, and the decoded video data in each area is rendered in parallel to finally obtain the rendering results of each area; S4: Stitching the rendering results of each area into a complete picture and completing the editing, the present invention adopts Multiple machines form a cluster, and the intra-frame segmentation method is used to split the heavy computing tasks of multi-layer large-format editing to different machines for processing, realizing free and smooth multi-layer large-format real-time editing.

Description

technical field [0001] The invention relates to the technical field of video editing, and more specifically relates to a method for realizing large-format multi-layer real-time editing based on intra-frame segmentation and clustering. Background technique [0002] Large-format video has massive amounts of data. Taking 8K as an example, a non-linear editing system needs to edit a layer of 8K4:2:210bit50P video, and needs to process more than 50Gbit of data per second, and these processes include very complex dizziness, such as decoding and rendering . However, the hardware capability of the current stand-alone nonlinear editing system is completely insufficient for editing 8K video, which is specifically reflected in the following aspects: [0003] 1. The CPU decoding ability is insufficient. At present, Intel's high-end dual-channel Xeon Gold processor can only decode 8K video in real-time for one layer, not two-layer real-time, let alone multi-layer real-time; [0004] 2....

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to solve the problem that the non-linear editing products currently on the market are all stand-alone, and the hardware capabilities of the stand-alone non-linear editing system cannot meet the user's editing requirements for 8K video. Segmented clustering realizes the method of large-format multi-layer real-time editing. Multiple machines are used to form a cluster, and the heavy computing tasks of multi-layer 8K large-format editing are split into different machines for processing by means of intra-frame segmentation. Use very high-speed network technology to realize data transmission, so that users can realize free and smooth multi-layer large-format real-time editing, and the operation experience is not much different from 4K editing, realizing unlimited format, unlimited layers, and unlimited frames rate free editing

Images