Digital Image Noise Restoration via Pixel-Level Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current blur techniques in digital image processing result in spatially varying noise reduction, making it difficult to accurately restore noise lost during the blurring process, which affects the aesthetics of the image by introducing inconsistent noise patterns.
Innovation Solution
A method is introduced to calculate a noise reduction factor for each pixel using a root of the sum of squares equation, allowing for accurate noise compensation to restore the original noise level across the image, ensuring consistent noise distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If blur is applied to reduce noise in digital image, then noise reduction is achieved, but noise becomes spatially varying and inconsistent which detracts from image aesthetics
Solution Approach 1:
The patent applies local quality by calculating a noise reduction factor for each pixel based on its specific blur amount. The noise compensation value is determined individually for each pixel using its noise reduction factor and the target noise level, ensuring that each pixel receives appropriate noise restoration tailored to its local blur characteristics. This resolves the contradiction by making noise compensation adaptive to local conditions rather than uniform across the entire image.
2Adaptability or versatility
If current blur techniques are used, then spatially varying blur is achieved, but accurate noise restoration becomes difficult
Solution Approach 1:
The patent changes the parameter of noise compensation by introducing a noise reduction factor that varies with the blur amount. The noise compensation value is calculated as a function of the noise reduction factor and target noise level, allowing the compensation to adapt dynamically to different blur conditions. This resolves the contradiction by making the noise restoration process sensitive to the specific blur parameters applied to each region.
Solution Approach 2:
The patent implements feedback by using the calculated noise reduction factor to determine the appropriate noise compensation value. The process continuously adjusts noise compensation based on the actual blur amount experienced by each pixel, creating a closed-loop system that optimizes noise restoration accuracy. This feedback mechanism enables precise noise restoration even when spatially varying blur is applied.
3Quantity of substance
If noise is added back to blurred image, then noise level is restored, but inconsistent noise patterns are created that are readily discernable to human eye
Solution Approach 1:
The patent changes the parameter of noise distribution by using the noise reduction factor to modulate the noise compensation value for each pixel. Instead of adding uniform noise across the image, the compensation value is adjusted based on local blur characteristics, ensuring that noise is distributed consistently according to the target noise level. This resolves the contradiction by making noise addition parameter-dependent rather than uniform.
Data Source
AI summary
Embodiments of the present invention provide systems, methods, and computer storage media directed to restoring digital image noise lost in a blur process. In one embodiment, a noise reduction factor is calculated for each pixel of the digital image that has had a blur applied thereto. The noise reduction factor for a respective pixel is indicative of an estimated reduction in noise resulting from an amount of blur applied to the respective pixel. A noise compensation value can then be determined for each of the pixels, based on the noise reduction factor calculated for each of the pixels and a target noise level for the digital image as a whole. Once the noise compensation value is determined, noise can be applied to each of the plurality of pixels in accordance with the determined noise compensation value. Other embodiments may be described and/or claimed.