Multi-Scale Image Detail Enhancement With Noise-Suppressed Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image enhancement and detail enhancement algorithms based on neural networks suffer from low processing fineness, poor flexibility, and are easily affected by image noise.
Innovation Solution
A multi-scale detail enhancement model that performs multi-scale decomposition on an input image, extracts feature maps, residual features, and fuses them to enhance details while suppressing noise, using a trained model with sequential operation blocks and convolution layers for improved image processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If existing neural network models are used for image enhancement, then processing speed is maintained, but processing fineness is low and detail enhancement capability is poor
Solution Approach 1:
The patent divides the image into multiple scales (coarse scale and fine scale) and processes each scale separately through dedicated neural network branches. This segmentation allows each branch to focus on specific detail levels, improving overall processing fineness without requiring a single overly complex model to handle all scales simultaneously.
Solution Approach 2:
The patent introduces a multi-scale dimensional framework by decomposing the image into different scale layers. This adds a scale dimension to the processing, allowing the system to capture details at multiple resolutions simultaneously, thereby improving processing fineness beyond what a single-scale model can achieve.
2Manufacturing precision
If detail enhancement is performed on noisy images, then detail information is enhanced, but noise is also amplified
Solution Approach 1:
The patent separates detail enhancement and noise suppression into different processing branches operating at different scales. The coarse-scale branch handles noise suppression while the fine-scale branch focuses on detail enhancement, allowing both functions to be performed simultaneously without mutual interference, thus enhancing details without amplifying noise.
Solution Approach 2:
The patent applies different processing characteristics to different scale layers: coarse-scale layers receive stronger noise suppression processing while fine-scale layers receive enhanced detail processing. This local quality differentiation ensures that noise suppression and detail enhancement are optimized for their respective scales, preventing noise amplification while maintaining detail quality.
3Manufacturing precision
If single-scale processing is used, then model complexity is low, but processing fineness and flexibility are insufficient
Solution Approach 1:
The patent segments the processing into multiple scale branches (coarse and fine scales), with each branch having its own neural network structure. This segmentation improves processing fineness by capturing details at different resolutions while keeping each individual branch relatively simple, avoiding the need for a single overly complex model.
Solution Approach 2:
The patent creates a multi-scale framework where each scale branch can be independently configured and trained for specific tasks. This universal structure allows the same basic architecture to handle different scale requirements, improving flexibility and processing fineness without proportionally increasing overall model complexity.
4Adaptability or versatility
If existing algorithms are used, then implementation is simple, but they are easily affected by image noise and lack flexibility
Solution Approach 1:
The patent segments the image processing into multiple independent scale branches that can be selectively activated or configured based on input image characteristics. This segmentation provides flexibility to adapt processing to different scenarios while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The patent implements a dynamic multi-scale framework where the processing can adaptively adjust which scales to emphasize based on image content and noise levels. This dynamic capability improves flexibility and adaptability while the modular structure keeps implementation complexity manageable through systematic organization.
Data Source
AI summary
An image processing method, comprising: by using a trained multi-scale detail enhancement model, performing detail enhancement on an input image to be processed; wherein multi-scale decomposition is performed on the input image to obtain a base layer image and at least one detail layer image; a first residual feature corresponding to a first feature map is acquired, and a second residual feature corresponding to a second feature map of each detail layer image is acquired; a base layer output image is obtained according to the first residual feature, each second residual feature and the first feature map, and a detail layer output image corresponding to the detail layer image is obtained according to the first residual feature, each second residual feature and the second feature map; and image fusion is performed on the base layer output image and each detail layer output image to obtain an output image.