Image Restoration Filter Using Segmented PSF Windowing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Constrained Least Square (CLS) filters for digital image restoration require a Fourier transform of the entire input image, resulting in a complex structure and high processing demands.
Innovation Solution
The proposed solution involves generating an image restoration filter coefficient by multiplying the CLS filter coefficient calculated from the Point Spread Function (PSF) by a window function, using a filter coefficient generation unit and a Finite Impulse Response (FIR) filter for convolution, to reduce processing requirements while maintaining effective image restoration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a conventional CLS filter is used for image restoration, then image restoration quality is maintained, but the device complexity and processing power requirements increase
Solution Approach 1:
The patent divides the large CLS filter into multiple smaller sub-filters arranged in a matrix structure. Each sub-filter processes a local region of the image independently, then results are combined through weighted summation. This segmentation reduces the computational complexity from O(N²) for a full N×N filter to O(M×k²) where M is the number of sub-filters and k is the sub-filter size, making the system more manageable while preserving restoration quality.
Solution Approach 2:
The patent applies window functions to each sub-filter to concentrate processing effort on the most relevant regions. By using Gaussian or other window functions, the filter coefficients are weighted to emphasize central regions and de-emphasize peripheral regions, achieving effective restoration with reduced filter support. This partial action approach maintains quality while reducing the effective processing area.
2Reliability
If a conventional CLS filter is used for image restoration, then image restoration quality is maintained, but the power consumption increases
Solution Approach 1:
By segmenting the filter into smaller sub-filters, the patent reduces the total number of multiplications and additions required. Each sub-filter operates on a smaller data set, and the segmented structure allows for parallel processing that reduces overall computational load and power consumption compared to a single large filter processing the entire image.
Solution Approach 2:
The patent applies different window function weights to different regions of the filter, concentrating computational resources on the most important local regions. This local quality approach ensures that power is spent primarily on processing the most critical image regions, reducing overall power consumption while maintaining restoration quality in the most important areas.
3Productivity
If a smaller filter size is used for image restoration, then processing speed increases, but the restoration quality may deteriorate
Solution Approach 1:
The patent uses multiple small sub-filters instead of one large filter, achieving the benefits of small filter size (fast processing) while maintaining the effective coverage of a large filter through the matrix arrangement. The combined effect of multiple sub-filters with appropriate weighting reproduces the restoration quality of a full CLS filter at a fraction of the computational cost.
Solution Approach 2:
The patent transforms the problem from a single large 2D filter into a 3D structure with multiple sub-filters arranged in a matrix. This dimensional transformation allows the system to achieve the computational efficiency of small filters while maintaining the effective aperture and restoration quality of a large filter through the additional dimension of multiple filter instances.
Data Source
AI summary
An image restoration device, an image restoration method, and an image restoration system are provided. The image restoration device includes a point spread function estimation unit, and an image restoration filter unit. The point spread function estimation unit receives an image signal and estimates a Point Spread Function (PSF) from the received image signal. The image restoration filter unit receives the image signal and the PSF, generates an image restoration filter coefficient from the PSF, and filters the image signal according to the image restoration filter coefficient to output the filtered image signal.