Chroma Noise Reduction via Adaptive Low-Pass Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital image processing techniques struggle to effectively reduce chroma noise without causing color bleeding, especially in darker regions, as they often rely on uncorrelated noise assumptions and large blur kernels which can desaturate genuine chrominance details.
Innovation Solution
A method that applies a low-pass filter to chrominance signals, modulating its strength based on the dynamic range of luminance and chrominance signals to restrict filtering to flat-image areas, using down-sampling and up-sampling to employ small filter kernels and adaptively weight chrominance values to prevent color bleeding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If standard gray-scale image-noise-reduction techniques are applied to each color channel separately, then noise reduction is achieved, but excessive desaturation of genuine chrominance details occurs
Solution Approach 1:
The patent applies different filtering strengths to different regions of the image based on edge detection. In regions with edges, filtering is reduced or eliminated to preserve chrominance details, while in flat regions filtering is applied to remove noise. This spatially adaptive approach resolves the contradiction by making the filtering behavior locally optimized rather than uniformly applied.
Solution Approach 2:
The patent dynamically adjusts the filtering parameters based on the detected edge map. The filter strength is modulated according to the presence or absence of edges in different regions, allowing the system to adapt its denoising behavior in real-time. This dynamic adjustment enables the system to maintain chrominance saturation in edge regions while removing noise in flat regions.
2Reliability
If large blur kernels are used to remove low-frequency chroma blobs, then chroma noise reduction is improved, but color bleeding at sharp colored edges increases
Solution Approach 1:
The patent applies large blur kernels only in regions where edges are not detected, while using small or no kernels in edge regions. This spatially selective application of filtering allows effective chroma noise reduction in flat areas without causing color bleeding at sharp colored edges, directly resolving the contradiction between noise reduction effectiveness and color bleeding prevention.
Solution Approach 2:
The patent segments the image into edge regions and flat regions based on edge detection. Different filtering strategies are applied to each segment: aggressive filtering in flat regions and conservative filtering in edge regions. This segmentation approach allows the system to optimize noise reduction where needed while preserving edges, eliminating the trade-off between chroma noise reduction and color bleeding.
3Reliability
If aggressive filtering is applied to remove chroma noise, then low-frequency chroma blobs are reduced, but sharp colored edges begin to bleed color
Solution Approach 1:
The patent applies aggressive filtering only in flat regions where edges are not present, while preserving edge sharpness in regions where edges are detected. This local differentiation of filtering strength allows effective chroma noise removal in flat areas without compromising edge sharpness, resolving the contradiction between noise removal effectiveness and edge preservation.
Data Source
AI summary
An embodiment of a method for reducing chroma noise in digital image data and of a corresponding image processor. Chrominance components are subjected to low-pass filtering. The strength of the low-pass filtering is modulated in accordance with the dynamic range of the luminance signal and the dynamic range of each of the two chrominance signals in order to avoid color bleeding at image-object edges. Moreover, the low-pass filtering is selectively applied to pixels with similar luminance and chrominance values only. A combination of down-sampling and up-sampling units is employed so that comparatively small filter kernels may be used for removing chroma noise with low spatial frequency.