Adaptive Noise Reduction for Image Edge Preservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current noise reduction methods for image data do not effectively consider brightness or direction, leading to edge damage and image blurring, especially in low-light conditions.
Innovation Solution
An apparatus and method that adaptively reduce noise by using a flexible sigma filter with varying sizes, determining whether each pixel is an edge in multiple directions, performing low-pass filtering only for non-edge pixels, and selecting filters based on average brightness to prevent blurring and maintain image sharpness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If noise reduction filtering is applied to all pixels uniformly, then noise is reduced, but edges are damaged and image becomes blurred
Solution Approach 1:
The patent applies different filtering operations to different regions of the image based on local characteristics. Edge pixels are identified and excluded from certain filtering operations, while non-edge pixels receive noise reduction filtering. This local differentiation allows noise reduction in flat regions while preserving edge sharpness, directly resolving the contradiction between noise reduction and edge preservation.
Solution Approach 2:
The image is segmented into edge regions and non-edge regions through edge detection algorithms. This segmentation allows the system to apply appropriate filtering strategies to each region independently - preserving edges in edge regions while applying noise reduction in non-edge regions, thereby solving the contradiction between uniform noise reduction and edge preservation.
2Adaptability or versatility
If fixed-size filter is used for noise reduction, then processing is simple, but cannot adapt to different brightness regions
Solution Approach 1:
The patent employs dynamic filter size selection based on local image characteristics such as brightness and noise levels. Instead of using a fixed filter size, the system adjusts the filter kernel size adaptively for different regions, allowing larger filters in uniform low-noise areas and smaller filters near edges or high-contrast regions. This dynamic adaptation resolves the contradiction between adaptability and complexity by implementing complexity only where necessary.
Solution Approach 2:
The system changes filter parameters (size, type) based on local image properties including brightness and gradient magnitude. In dark regions with high noise, larger filters are applied; in bright regions or near edges, smaller filters are used. This parameter adaptation allows the system to handle varying noise characteristics across different brightness regions while maintaining edge sharpness, resolving the contradiction between adaptability and simplicity.
Data Source
AI summary
An apparatus of reducing noise includes: an edge determining module determining whether each pixel of an image corresponds to an edge; a low pass filter module performing low pass filtering of the image in two directions and performing low pass filtering only for the pixel determined not to be the edge; a filter selecting module selecting a filter having a flexible size to be applied to a mask region for each of the images low pass filtered based on the average brightness of the entire region and the average brightness of the mask region having a predetermined size; a sigma filter module sigma filtering the mask region using the selected filter for each of the images low pass filtered; and an averaging module averaging the image sigma filtered in the two directions.