Neural-Network Post-Filter Signaling for Reliable Image Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution and high-quality images leads to a significant increase in transmission and storage costs due to the rise in the amount of transmitted information, necessitating high-efficient image compression technology.
Innovation Solution
An image encoding/decoding method that includes obtaining and signaling post-filter-based output picture information through neural-network post-filter characteristics (NNPFC) SEI messages, with specific conditions for output picture generation, to improve encoding/decoding efficiency and clarify the meaning of information related to the output picture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If image resolution and quality are improved to meet high-definition and ultra high-definition demands, then image quality is improved, but transmission cost and storage cost increase due to increased amount of transmitted information
Solution Approach 1:
The patent extracts and removes redundant information from the bitstream through intelligent post-processing filters. The neural network-based post-filter selectively eliminates unnecessary data components while preserving essential image quality characteristics, thereby reducing the amount of transmitted information without compromising image quality.
Solution Approach 2:
The patent changes the parameter representation of image data by transforming spatial domain information into frequency domain or latent space representations through neural network processing. This parameter transformation enables more efficient compression by representing image content with fewer bits while maintaining perceptual quality.
2Productivity
If neural-network post-filter characteristics SEI messages are used to specify output picture information, then coding quality and efficiency are improved, but decoder errors may occur due to ambiguous meaning of information
Solution Approach 1:
The patent implements feedback mechanisms where the encoder provides explicit signaling information about the output picture characteristics generated by the neural network post-filter. The decoder uses this feedback information to accurately reconstruct and interpret the filtered output, ensuring synchronization between encoder and decoder operations and preventing interpretation errors.
Solution Approach 2:
The patent performs preliminary action by pre-defining and signaling the characteristics of the neural network post-filter output picture in advance through SEI messages. This preliminary specification of output picture properties (such as resolution, format, and processing parameters) enables the decoder to prepare appropriate processing routines and avoid errors during actual decoding operations.
3Quantity of substance
If high-efficient image compression technology is implemented to reduce transmission and storage costs, then transmission cost and storage cost are reduced, but image quality may deteriorate due to compression artifacts
Solution Approach 1:
The patent replaces traditional mechanical filtering methods with neural network-based intelligent filtering. The neural network learns optimal filtering strategies from training data and adapts to different image content types, achieving superior compression efficiency while minimizing quality loss compared to conventional fixed algorithms.
Solution Approach 2:
The patent combines multiple processing stages including traditional compression algorithms with neural network post-filtering in a composite system. This hybrid approach leverages the strengths of both conventional and AI-based methods to achieve optimal balance between compression ratio and image quality preservation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided are an image encoding/decoding method, a method for transmitting a bitstream, and a computer-readable recording medium for storing a bitstream. The image decoding method according to the present disclosure may comprise the steps of: obtaining, from a neural-network post-filter characteristics (NNPFC) supplemental enhancement information (SEI) message, post-filter-based corresponding output picture information for an input picture; and obtaining a corresponding output picture for the input picture on the basis of the corresponding output picture information, wherein the corresponding output picture information may include output picture generation information about whether to generate a corresponding output picture of a post-filter for the input picture.