Two-Layer Image Encoding for Low-Bitrate Reconstruction Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional image coding standards and neural network-based encoding technologies suffer from significant information loss at low bitrates, leading to poor image reconstruction quality and visual experience due to the use of a single code stream.
Innovation Solution
An image processing method that extracts a one-dimensional feature vector from an original image block, transforms it into a multi-dimensional feature map, quantizes the one-dimensional feature vector into a first code stream, and discretely encodes the multi-dimensional feature map into a second code stream, enabling efficient compression of different levels of image information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If a single code stream is used for image encoding, then the encoding process is simple, but information loss is significant at low bitrates resulting in poor image reconstruction quality
Solution Approach 1:
The patent divides the image encoding process into two separate code streams: a first code stream for low-frequency information and a second code stream for high-frequency information. This segmentation allows each stream to be optimized independently, reducing information loss at low bitrates while maintaining manageable encoding complexity through modular processing.
Solution Approach 2:
The patent transforms the traditional single-dimension encoding approach into a two-dimensional encoding structure by separating frequency domains (low-frequency and high-frequency components). This dimensional change enables more efficient information preservation by allocating bits according to frequency importance, significantly reducing information loss at low bitrates.
2Adaptability or versatility
If traditional image coding standards are used, then the encoding method is well-established, but they do not adequately support machine vision requirements
Solution Approach 1:
The patent applies different encoding strategies to different frequency components of the image. The first code stream uses encoding optimized for low-frequency information (suitable for human vision), while the second code stream uses encoding optimized for high-frequency information (suitable for machine vision). This local quality approach allows the system to meet diverse requirements without overhauling the entire encoding framework.
Data Source
AI summary
This application discloses an image processing method, device, system, electronic equipment, and storage medium, applied at the encoding end. The method includes extracting a one-dimensional feature vector from an original image block; transforming the original image block into a multidimensional feature map based on the one-dimensional feature vector; quantizing and encoding the one-dimensional feature vector to generate a first code stream; discretely encoding the multidimensional feature map to generate a second code stream, thereby efficiently compressing the spatial-independent vector and the multidimensional feature map; and sending the first and second code streams to the decoding end. Since the encoding stream comprises two layers each representing different types of image information, image reconstruction from the two-layer code streams maintains information integrity even at low bit rates, thus improving visual effects and experience.