AI Video Encoding with Residual Hyperprior Bitrate Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding technologies face challenges in reducing bitrate overheads, particularly in high-definition and ultra-high-definition video transmission, due to the inefficiencies in handling time and spatial redundancy in video frames.
Innovation Solution
The proposed method involves determining optical flow hyperprior information and residual information based on feature and prior information of video frames, using AI video encoding and decoding frameworks to optimize probability distributions and reduce bitrate by encoding only a portion of the frame data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If conventional video compression technologies (H265, H266) are used, then video quality can be maintained, but bandwidth and storage requirements increase significantly
Solution Approach 1:
The patent segments video frames into different components (optical flow, residual information, hyperprior information) and processes them separately through different encoding paths. This allows selective compression of different frame components, reducing overall bitrate while maintaining visual quality by focusing compression efforts on less perceptually critical elements.
Solution Approach 2:
The patent performs entropy encoding only on a portion of the frame data (specifically on optical flow and residual information) rather than all frame data. By applying partial encoding actions to the most critical components and using hyperprior information for the rest, the system achieves significant bitrate reduction while preserving essential video quality.
2Quantity of substance
If inter encoding is performed on all video frames to reduce bitrate, then bandwidth requirements decrease, but encoding complexity and processing time increase
Solution Approach 1:
The patent applies inter encoding and entropy encoding selectively only to optical flow and residual information components rather than all frame data. This partial application of complex encoding operations significantly reduces computational complexity compared to full-frame entropy encoding while still achieving bitrate reduction through compression of the most variable frame components.
Solution Approach 2:
The patent performs motion estimation and generates hyperprior information before the main entropy encoding process. This preliminary action prepares probability distributions and reference data in advance, simplifying the subsequent entropy encoding operations and reducing overall processing complexity by pre-organizing the data structure for efficient compression.
3Quantity of substance
If entropy encoding is performed on all frame data, then compression ratio increases, but encoding time and computational resources increase
Solution Approach 1:
The patent performs entropy encoding only on optical flow and residual information (portions of frame data) rather than all frame data. This selective partial encoding achieves meaningful compression ratios by focusing on the most compressible and visually critical components, while avoiding the prohibitive computational cost of entropy encoding entire frames.
Solution Approach 2:
The patent extracts and separately encodes the optical flow and residual information components from the full frame data. By taking out these specific components for targeted entropy encoding and using hyperprior information for the remaining data, the system achieves efficient compression with reduced encoding time compared to processing all frame data through entropy encoding.
Data Source
AI summary
This application relates to an encoding method, a decoding method, and an electronic device. An example method includes: performing inter prediction on a current frame, to obtain prediction information of the current frame; determining residual information of the current frame based on the prediction information of the current frame and an original picture of the current frame; determining first residual hyperprior information of the current frame based on the residual information of the current frame and prior information of the residual information of the current frame; encoding the first residual hyperprior information of the current frame, to obtain a first bitstream; and performing probability estimation based on second residual hyperprior information of the current frame and prior information of the residual information of the current frame, to obtain a residual probability distribution of the current frame.