Image Restoration via Candidate Clustering and Reliability Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods, such as learning-type super resolution arts, often fail to accurately restore the original content of low-quality images, leading to incorrect outputs, especially when image degradation is significant, as they prioritize increasing inconsistency over selecting multiple restored images that may indicate the original content.
Innovation Solution
An image processing device and method that generates multiple restored image candidates using a dictionary of patch pairs associated with degraded and restored image patches, calculates their reliability degrees, and clusters these candidates to select the most accurate restored images based on reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single restored image is generated using learning-type super resolution, then the processing is simple and fast, but the accuracy of restoring the original content deteriorates when image degradation is significant
Solution Approach 1:
The system dynamically adjusts the number of restored image candidates generated based on the degree of image degradation. When degradation is significant, it generates more candidates (e.g., 5 or more) to ensure the original content is captured. When degradation is minor, it generates fewer candidates to reduce processing complexity. This dynamic adjustment resolves the contradiction between reliability and device complexity.
Solution Approach 2:
The system changes the parameter of candidate image quantity based on degradation assessment. By varying this parameter according to input image quality, the system achieves high restoration accuracy for severely degraded images while maintaining reasonable processing efficiency for mildly degraded images, thus resolving the contradiction between reliability and complexity.
2Reliability
If multiple restored image candidates are generated to ensure accuracy, then the reliability of restoring original content improves, but the processing time and computational load increase
Solution Approach 1:
The system dynamically controls the number of candidates generated based on degradation level. For severely degraded images, it generates more candidates (5 or more) to ensure accuracy. For mildly degraded images, it generates fewer candidates (1-3) to reduce processing time. This dynamic approach resolves the contradiction between reliability and time loss.
Solution Approach 2:
The system changes the candidate quantity parameter according to degradation assessment results. This parameter adaptation allows the system to maintain high reliability when needed while minimizing processing time when degradation is minor, effectively resolving the time-reliability contradiction.
3Productivity
If high frequency component is selected only to increase inconsistency, then the super resolution effect is enhanced, but the accuracy of corresponding to the original input image deteriorates
Solution Approach 1:
The system uses degradation assessment as feedback to control the generation of restored image candidates. By assessing the degradation level first, the system can adjust the number of candidates generated, ensuring that high productivity (super resolution effect) is achieved while maintaining reliability (accuracy) through appropriate candidate quantity selection.
Solution Approach 2:
The system performs preliminary degradation assessment before generating restored image candidates. This preliminary action allows the system to determine the appropriate number of candidates to generate, ensuring both super resolution effect and accuracy are achieved by preparing the correct number of candidates in advance based on degradation level.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention provides an image processing device whereby the probability of outputting a restored image which accurately corresponds to an original image which is included in a low-quality input image is improved. This image processing device comprises: an image group generating means for generating, from the input image, using a dictionary which stores a plurality of patch pairs wherein a degradation patch which is a patch of a degraded image wherein a prescribed image is degraded is associated with a restoration patch which is a patch of this prescribed image, a plurality of restored image candidates including a plurality of different instances of content which have a possibility of being the original content of the input image; and an image selection presentation means for clustering the generated plurality of restored image candidates, and selecting and outputting an image candidate on the basis of the result of this clustering.