Neural Image Compression With Adaptive Entropy Coding Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in efficiently managing high-resolution video and image data compression while minimizing bandwidth and storage requirements, particularly in devices with limited hardware resources.
Innovation Solution
An image processing method using a neural network that adjusts latent vector ranges based on a preset target compression ratio, employs adaptive entropy coding, and utilizes a bypass flag to determine whether to perform entropy coding, thereby optimizing data transmission and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If entropy coding is applied to compress latent vectors, then compression ratio is improved, but processing complexity and time increase
Solution Approach 1:
The patent implements dynamic selection between entropy coding and bypass modes based on compression ratio requirements. The system adaptively determines whether to apply entropy coding to each latent vector or its components, allowing the processing complexity to vary dynamically according to the needed compression level rather than applying a fixed complex encoding process to all data
Solution Approach 2:
The patent changes the parameter of compression ratio as a control variable to determine the encoding strategy. By adjusting the target compression ratio parameter, the system can switch between different processing modes (full entropy coding, partial entropy coding, or bypass), thereby balancing compression effectiveness against processing complexity
2Quantity of substance
If high compression ratio is achieved through entropy coding, then bandwidth usage is reduced, but image quality may deteriorate
Solution Approach 1:
The patent applies different processing qualities to different parts of the latent representation. Instead of uniformly applying entropy coding to the entire latent vector, the system can selectively apply entropy coding to specific components or subsets of the latent vector, preserving more critical information components in higher quality while compressing less critical components more aggressively
Solution Approach 2:
The patent implements partial entropy coding where only a portion of the latent vector components undergo entropy coding while others are transmitted without compression or with different compression levels. This partial application of the compression technique allows the system to achieve some bandwidth reduction while maintaining image quality above a certain threshold
3Productivity
If latent vector range is adjusted to meet compression ratio, then compression efficiency is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary adjustment of the latent vector range before entropy coding is applied. By pre-processing the latent vectors to optimize their range and distribution, the system reduces the complexity and time required for the subsequent entropy coding step, thereby improving overall compression efficiency while minimizing total processing time
Data Source
AI summary
An image processing method using a neural network includes receiving input image data, generating a first latent vector corresponding to the input image data by inputting the input image data to the neural network and encoding the input image data, and generating a second latent vector based on the first latent vector, where a range of the first latent vector is adjusted based on a preset target compression ratio.