Asymmetric Unsharp Mask for Complex PSF Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional unsharp mask processing struggles with sharpening images deteriorated by complex Point Spread Functions (PSF) with asymmetric aberrations and sagittal halos, as it relies on rotationally symmetric filters, which fail to accurately correct asymmetries and require excessive information for processing.
Innovation Solution
An image processing method that reconstructs a discretized PSF using coefficient data to perform unsharp mask processing, allowing for a rotationally asymmetric unsharp mask that matches the PSF of the optical system, thereby reducing the required information amount and improving sharpening accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a rotationally symmetric filter is used for unsharp mask processing, then the processing is simple, but it cannot adequately correct asymmetric aberrations and sagittal halos in the PSF
Solution Approach 1:
The patent applies asymmetry by designing the unsharp mask filter to match the asymmetric characteristics of the optical system's PSF. The filter coefficients are determined based on the specific asymmetric aberrations (coma, astigmatism, sagittal halos) present in the optical system, allowing the filter to correctly compensate for these asymmetric degradations rather than applying a generic symmetric blur.
Solution Approach 2:
The patent implements local quality by allowing different regions of the image to have different filter characteristics. The unsharp mask filter is designed with spatially varying coefficients that adapt to the local PSF characteristics at different field positions, enabling accurate correction of position-dependent asymmetric aberrations while maintaining processing efficiency.
2Productivity
If conventional unsharp mask processing is used, then the processing speed is fast, but the sharpening accuracy is insufficient for images with complex PSF
Solution Approach 1:
The patent applies preliminary action by pre-calculating the optimal unsharp mask filter coefficients based on the known PSF characteristics of the optical system. These pre-determined coefficients are stored and directly applied during image processing, eliminating the need for complex real-time calculations while maintaining high sharpening accuracy tailored to the specific optical system.
Solution Approach 2:
The patent utilizes parameter changes by adapting the unsharp mask filter parameters (coefficients, kernel size, sigma values) to match the specific PSF characteristics of the optical system. By adjusting these parameters based on the measured or calculated PSF, the filter achieves optimal sharpening performance for images degraded by complex asymmetric aberrations.
3Manufacturing precision
If asymmetric aberration correction is increased, then the correction in the azimuth direction with large aberration improves, but undershoot occurs in the azimuth direction with small aberration
Solution Approach 1:
The patent implements feedback by using the known PSF characteristics as a reference to guide the unsharp mask filter design. The filter coefficients are determined based on the actual measured or calculated PSF, which provides feedback information about the true aberration levels in different directions. This feedback mechanism enables the filter to automatically balance the correction strength across different azimuth directions, preventing both overcorrection and undershoot.
Data Source
AI summary
An image processing apparatus includes an acquirer configured to acquire a captured image generated through imaging by an optical system, a reconstruction processor configured to reconstruct a discretized point spread function of the optical system using coefficient data used to approximate the point spread function, and a sharpening processor configured to perform unsharp mask processing for the captured image based on information on the reconstructed point spread function. A discretization interval of the reconstructed point spread function is different from a pixel pitch in an image sensor used for the imaging.