Jointly Trained Image Processing Network for Multi-Stage Enhancement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing models trained separately for different optimization tasks, such as super-resolution, image quality enhancement, and face enhancement, often exhibit conflicting effects when applied sequentially, leading to weakened overall optimization performance.
Innovation Solution
A multi-task joint training framework is employed, where super-resolution, image quality enhancement, and face enhancement networks are cascaded and trained together using a unified loss function, allowing for mutual constraint and promotion of network parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If multiple image models are trained separately for different optimization tasks, then each model can be optimized for its specific task, but the models may have reverse optimization effects on each other when used together, weakening the overall optimization effect
Solution Approach 1:
The patent merges multiple image processing models (super-resolution, quality enhancement, face enhancement, and sharpening models) into a single unified model. This model contains multiple processing branches that can simultaneously perform different optimization tasks on the same input image, ensuring that all optimizations work together harmoniously rather than conflicting with each other, thus eliminating reverse optimization effects while maintaining task-specific optimization capabilities
Solution Approach 2:
The unified image processing model is designed to perform multiple functions simultaneously - super-resolution, quality enhancement, face enhancement, and sharpening - all within a single model structure. This multi-functional design allows the model to process different aspects of image optimization in an integrated manner, preventing conflicts between separate models while achieving comprehensive image optimization
2Adaptability or versatility
If multiple image models are used for different optimization tasks, then various image properties can be enhanced, but the device complexity increases due to multiple separate models
Solution Approach 1:
The patent combines multiple separate image processing models into one unified model with multiple processing branches. This merging reduces device complexity by eliminating the need to manage, deploy, and coordinate multiple separate models, while still providing comprehensive optimization capabilities across different image properties through the integrated branches within the single model
Data Source
AI summary
An image processing method is performed by a computer device, which includes: converting initial image data into super-resolution image data using a trained image processing network, a resolution of the super-resolution image data being greater than or equal to a target resolution; performing image quality enhancement processing on the super-resolution image data using the trained image processing network, to obtain first enhanced image data; when there is a face image in the first enhanced image data, performing face enhancement on the face image in the first enhanced image data using the trained image processing network to obtain second enhanced image data; and performing image sharpening processing on the second enhanced image data using the trained image processing network to obtain sharpened image data.