Blurred Image Patch Dictionary Creation for Accurate Restoration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies, such as learning-based super-resolution methods, face challenges in accurately generating high-resolution images from blurred images, particularly when the blur type is not easily estimable, leading to erroneous selection of high-resolution patches.
Innovation Solution
A dictionary creation device and image processing system that generates and registers patch pairs, including restoration and blurred patches, along with corresponding blur parameters, allowing for flexible selection of restoration patches based on feature similarity and blur parameter similarity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a learning-based super resolution process is used to generate high resolution images from blurred images, then image restoration capability is improved, but accuracy deteriorates when blur type is not easily estimable
Solution Approach 1:
The image is divided into multiple patches, and each patch is processed independently to find corresponding patches in the dictionary. This segmentation allows the system to handle different blur types in different regions of the image, improving overall accuracy while maintaining restoration capability.
Solution Approach 2:
The invention introduces blur parameter as an additional dimension for dictionary matching. By changing the matching criteria from only patch similarity to both patch similarity and blur parameter similarity, the system can accurately handle images with different blur types, resolving the accuracy deterioration problem.
2Ease of operation
If patch similarity alone is used for selecting restoration patches, then selection simplicity is improved, but accuracy deteriorates due to erroneous selection when blur types differ
Solution Approach 1:
The invention introduces blur parameter as an intermediary criterion between patch similarity and final patch selection. This intermediary layer ensures that patches with similar blur characteristics are selected, preventing erroneous selections while maintaining a relatively simple dictionary-based matching process.
3Quantity of substance
If a dictionary with one-to-many relation is used to aggregate similar low frequency feature quantities, then memory amount is reduced, but accuracy deteriorates due to loss of blur-specific information
Solution Approach 1:
The invention creates multiple copies of dictionary patches organized by blur parameter categories. Instead of aggregating all similar patches into one representative, the system maintains separate copies for different blur types, preserving blur-specific information while keeping memory usage manageable through structured organization.
Data Source
AI summary
A dictionary creation device including a blurred image generation unit which outputs a blurred image generated by performing a blurring process to a learning image together with a blur parameter indicating a blurring state of the blurred image, a patch pair generation unit which generates a restoration patch and a blurred patch as a patch pair that is composed of the patches located at the corresponding positions of the learning image and the blurred image, and a registration unit which associates the patch pair with a blur parameter corresponding to the blurred patch in the patch pair and registers them in a dictionary.