Target Area Blur Estimation for Reference Image Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional multiframe super resolution techniques often fail to produce intended high resolution images due to inappropriate selection of reference images, leading to variations in image quality and composition, even when selecting based on whole-image blur amounts.
Innovation Solution
An image processing apparatus that includes a target area setting unit, an image index estimating unit, and a selecting unit to identify and select a reference image based on estimated image indices specific to target areas within low resolution images, ensuring optimal high resolution image generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a reference image is selected based on the blur amount of the whole image, then the selection process is simplified, but the image quality in specific target areas may deteriorate
Solution Approach 1:
The patent divides the image evaluation process into two levels: whole-image blur amount calculation for initial filtering, and target-area blur amount calculation for final selection. This segmentation allows the system to first narrow down candidates using a simple whole-image metric, then apply a more sophisticated target-area metric to ensure high quality in specific regions of interest.
Solution Approach 2:
The patent implements local quality assessment by calculating blur amounts specifically in target areas rather than uniformly across the entire image. This allows different parts of the image to have different quality requirements, with the reference image selection optimized for the importance degree of specific target areas while maintaining overall image quality.
2Device complexity
If the first low resolution image is selected as the reference image, then the processing is simplified and consistent, but the image quality and composition of the high resolution image deteriorate
Solution Approach 1:
The patent introduces feedback mechanisms where the system evaluates the blur amount in target areas of each low resolution image and uses this information to select the optimal reference image. This feedback loop ensures that the selected reference image has the smallest blur amount in the target area, thereby improving the quality of the generated high resolution image while maintaining automated decision-making.
3Stability of the object's composition
If a reference image with small whole-image blur amount is selected, then the overall image stability is improved, but the clarity of specific important areas may deteriorate
Solution Approach 1:
The patent implements local quality assessment by calculating blur amounts specifically in target areas rather than uniformly across the entire image. This allows different parts of the image to have different quality requirements, with the reference image selection optimized for the importance degree of specific target areas while maintaining overall image quality.
Solution Approach 2:
The patent changes the evaluation parameter from whole-image blur amount to target-area blur amount, and further refines it by incorporating importance degree weighting. This parameter change allows the system to prioritize clarity in specific important areas while still considering overall image stability, achieving a more nuanced selection criterion.
Data Source
AI summary
To generate a high resolution image of higher quality depending on a target area. An image processing apparatus includes a blur amount estimating unit configured to estimate a blur amount in a set target area in each of images indicated by a plurality of low resolution image data, and a reference image selecting unit configured to select a low resolution image to be a reference for generating the high resolution image depending on the estimated blur amount.