Iterative Image Restoration via Dynamic Filter Coefficients
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image restoration methods, such as DTVF, are limited to denoising and cannot effectively handle image restoration tasks like super resolution and deblurring due to their design focusing on minimizing the sum of squares of differences between degraded and restored images, which is not suitable for large differences in deblurred images.
Innovation Solution
An information processing apparatus and method that iteratively processes images by calculating local variations, filter coefficients, and reconfiguration errors to create a restored image, capable of handling denoising, super resolution, and deblurring by degrading the input image and updating pixel values based on filter coefficients and errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If DTVF method is used to minimize the sum of squares of differences between degraded and restored images, then denoising performance is improved, but applicability to super resolution and deblurring is lost
Solution Approach 1:
The patent introduces dynamic adjustment of filter coefficients based on local image characteristics. The filter coefficient calculation unit dynamically computes coefficients according to the local variation calculated from the input image, allowing the system to adapt to different image degradation types (noise, blur, low resolution) rather than using a fixed filter design optimized only for denoising.
Solution Approach 2:
The patent changes the parameter of filter coefficient calculation from a fixed approach to a dynamic approach based on local variation. By calculating filter coefficients according to the local variation in each region, the system can adjust its behavior to suit different restoration tasks, thereby extending applicability beyond denoising to super resolution and deblurring.
2Productivity
If simple nonlinear filter processing is used, then calculation cost is reduced, but restoration accuracy for deblurred images is worsened
Solution Approach 1:
The patent segments the image processing into distinct functional units: variation calculation unit, filter coefficient calculation unit, and filter processing unit. This segmentation allows each unit to perform a specific operation efficiently, maintaining computational simplicity while improving overall restoration accuracy through coordinated processing of local variations and filter coefficients.
Solution Approach 2:
The patent implements a feedback mechanism where the local variation is calculated from the input image to determine the filter coefficients, which are then used in the filter processing. This feedback loop ensures that the filter adapts to the specific characteristics of the input image, improving restoration accuracy without significantly increasing calculation cost.
Data Source
AI summary
An information processing apparatus may include a memory storing instructions and at least one processor configured to process the instructions to receive an input image. The input image includes either a first image or a provisional image created by iteratively image processing the first image. The instructions further provide for the processor to calculate a local variation of a focused pixel in the input image based on a difference in pixel value between the focused pixel and a surrounding pixel of the focused pixel, to calculate a filter coefficient for suppressing a variation between neighboring pixels in the input image based on the local variation, to create a degraded image by degrading the input image, to calculate a reconfiguration error between the input image and the degraded image, and to create the provisional image based on the filter coefficient and the reconfiguration error.