AI Image Encoding and Decoding for Optical Flow Artifact Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding and decoding technologies using artificial intelligence (AI) face issues with increased bitrate due to artifacts in optical flow, leading to deteriorated image quality and inefficient encoding/decoding processes.
Innovation Solution
Implementing a neural network-based system that adaptively determines encoding and decoding modes by using optical flow, predicted images, and correction data to prevent artifacts and optimize bitrate, thereby improving image quality and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If optical flow is used in AI-based image encoding/decoding, then image quality can be improved through better motion compensation, but bitrate is unnecessarily increased due to artifacts in the optical flow
Solution Approach 1:
The patent extracts and removes artifacts from the optical flow data before it is used in the neural network-based decoding process. By separating the useful motion information from the harmful artifacts, the system maintains improved image quality while preventing bitrate increase from artifact-related corrections.
Solution Approach 2:
The patent detects artifacts in the optical flow and converts this harmful element into a benefit by using the artifact detection information to adjust the decoding process. The artifacts are not merely removed but their presence is leveraged to improve the overall encoding efficiency by preventing unnecessary bitrate consumption.
2Manufacturing precision
If artifact correction is applied to optical flow, then image quality is improved, but encoding complexity and processing time increase
Solution Approach 1:
The patent performs artifact detection and correction in advance, during the encoding phase, so that the corrected optical flow is ready for use in decoding without requiring additional complex processing during the decoding phase. This preliminary action simplifies the overall system complexity while maintaining image quality improvements.
Solution Approach 2:
The encoding system performs self-correction of optical flow artifacts by automatically detecting and correcting them during the encoding process itself, without requiring external intervention or complex post-processing. This self-service approach reduces overall system complexity while improving image quality.
3Loss of information
If traditional encoding methods are used, then bitrate is kept low, but image quality deteriorates due to inability to handle optical flow artifacts
Solution Approach 1:
The patent introduces an intermediary artifact detection and correction module between the optical flow generation and the neural network decoding process. This intermediary component enables the system to maintain low bitrate by preventing artifact propagation, while still achieving improved image quality through the neural network's ability to process corrected optical flow data.
Data Source
AI summary
An image decoding method using artificial intelligence (AI), including obtaining, from a bitstream, a current optical flow and correction data which are generated based on a current predicted image and a current image; obtaining the current predicted image based on a previous reconstructed image and the current optical flow; obtaining feature data of the current predicted image by applying the current predicted image to a neural network-based predicted image encoder; and obtaining a current reconstructed image corresponding to the current image by applying the correction data and the feature data of the current predicted image to a neural network-based image decoder.