Parallel Video Encoding via Dual Graphics Adapters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video streaming technologies face high latency issues due to increased encoding delays when improving video stream quality, making them impractical for live video streaming applications, especially those using I-frames followed by P-frames without B-frames.
Innovation Solution
The use of two or more graphics adapters for parallel encoding of live video streams, dividing incoming frames between them to reduce frame dependencies and improve encoding throughput, allowing for real-time or near-real-time generation of low-delay video bitstreams.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If video stream quality is improved by increasing resolution or bitrate, then video quality is improved, but encoding latency increases
Solution Approach 1:
The patent divides the video stream into I-frames and P-frames, with I-frames encoded independently and P-frames encoded with reference to previous frames. This segmentation allows certain frames to be encoded faster while maintaining overall quality, reducing the average encoding latency while preserving video quality.
2Manufacturing precision
If encoding techniques are used to improve video stream quality, then video quality is improved, but encoding delay increases making it impractical for live streaming
Solution Approach 1:
The patent performs motion estimation and compensation in advance during the encoding of I-frames and previous P-frames, storing the results for rapid reference during subsequent P-frame encoding. This preliminary action reduces the computational burden during real-time encoding, maintaining video quality while improving encoding throughput for live streaming.
3Productivity
If I-frames followed by P-frames encoding pattern is used, then video compression efficiency is improved, but latency increases for live streaming applications
Solution Approach 1:
The patent dynamically adjusts the frequency of I-frames based on scene complexity and motion levels. In low-motion scenes, I-frames are spaced further apart, reducing the number of full-frame encodings and lowering latency. In high-motion scenes, I-frames are inserted more frequently to maintain quality. This parameter change optimizes the balance between compression efficiency and latency for live streaming.
Data Source
AI summary
Technology for low-delay encoding of live video streams in multi-adapter systems can include a host processor and memory comprising instructions which, when executed, cause a computing system to encode, via a first graphics adapter, an I-frame of a live video signal and a first subset of P-frames of the live video signal, encode, via a second graphics adapter, a second subset of P-frames of the live video signal, and combine the encoded video frames from the first adapter and the encoded video frames from the second adapter into an output video bitstream. Frames of the live video signal are divided into the first and second subsets of P-frames on an alternating frame basis, and the I-frame copied from the first graphics adapter to the second graphics adapter prior to encoding the second subset of P-frames, such that encoding the first and second subsets of P-frames occurs essentially in parallel.