Attention-Map Image Enhancement for Noise Reduction Without Detail Loss
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies struggle to effectively remove noise and artifacts while maintaining image quality and resolution, particularly in low-quality or low-resolution images.
Innovation Solution
An image processing apparatus utilizing one or more neural networks to perform denoising and enhance image quality by generating attention maps and feature information through a series of processing stages, including convolutional neural networks, multi-layer perceptrons, and adaptive instance normalization, to produce higher quality images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If existing image processing technologies are used to remove noise and artifacts, then image quality may be improved, but image resolution and detail are lost
Solution Approach 1:
The patent applies local quality by using attention maps that identify and prioritize specific regions of the image requiring different processing treatments. The system divides the image into local regions and applies targeted denoising operations only where needed, preserving important details while removing noise in less critical areas. This regional differentiation allows simultaneous improvement of image quality and preservation of resolution.
Solution Approach 2:
The patent introduces an attention map as an intermediary element that guides the denoising process. This attention map serves as a mediator between the original image and the processed output, directing where noise removal should be aggressive and where detail preservation is prioritized. The attention map enables the system to balance quality improvement and resolution maintenance by providing spatially-varying processing guidance.
2Manufacturing precision
If deep neural networks are used for image processing, then image quality can be enhanced, but computational complexity and processing time increase
Solution Approach 1:
The patent segments the image processing task into distinct functional components: feature extraction, attention map generation, and denoising application. By dividing the deep neural network processing into these separate stages, the system can optimize each component independently and apply processing only where necessary, reducing overall computational complexity while maintaining image quality enhancement capabilities.
Solution Approach 2:
The patent performs preliminary action by generating attention maps before applying the full denoising process. These attention maps are computed in advance to identify regions of interest, allowing the subsequent denoising operations to be targeted and efficient. This preliminary guidance reduces the computational burden of the main processing stage by avoiding unnecessary calculations in already-identified critical regions.
3Object-affected harmful factors
If aggressive denoising is applied to low-quality images, then noise is reduced, but important image details and artifacts are removed along with noise
Solution Approach 1:
The patent applies local quality by using attention maps that identify and prioritize specific regions of the image requiring different processing treatments. The system divides the image into local regions and applies targeted denoising operations only where needed, preserving important details while removing noise in less critical areas. This regional differentiation allows simultaneous improvement of image quality and preservation of resolution.
Solution Approach 2:
The patent introduces an attention map as an intermediary element that guides the denoising process. This attention map serves as a mediator between the original image and the processed output, directing where noise removal should be aggressive and where detail preservation is prioritized. The attention map enables the system to balance quality improvement and resolution maintenance by providing spatially-varying processing guidance.
Data Source
AI summary
An image processing apparatus for performing an image by using one or more neural networks may include a memory storing one or more instructions and at least one processor configured to execute the one or more instructions stored in the memory to obtain first feature information of a first image, generate an intermediate output image for the first image by performing first image processing on the first feature information, generate an attention map, based on the first image and the intermediate output image, obtain second feature information by performing second image processing on the first feature information, obtain fourth feature information by performing third image processing on third feature information extracted during the first image processing, and generate a second image having a higher quality than the first image, based on the attention map, the second feature information, and the fourth feature information.