Digital Image Noise Reduction via Flat Noise Space Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional noise filtering methods for digital images face challenges in signal-dependent noise reduction, often blurring edges or failing to smooth non-detail areas, as they need to determine pixel brightness to apply noise reduction effectively.
Innovation Solution
The method transforms image values from linear space to a flat noise space where noise is independent of signal values, allowing for edge-preserving noise filtering techniques to be applied, and then transforms back to linear space, with separate filtering methods used for luminance and color channels in luminance-chrominance representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional noise filtering methods are applied to reduce noise in digital images, then noise reduction is achieved, but edges are blurred or fine details are lost
Solution Approach 1:
The patent applies different filtering strengths to different regions of the image by transforming to uniform noise space. In this transformed space, the filtering operation can be applied uniformly while the transformation itself ensures that edges and fine details are preserved because the transformation adapts to local signal characteristics. This resolves the contradiction by making the filtering quality local (preserving edges where needed) while using a global uniform filtering operation.
Solution Approach 2:
The patent changes the parameter space by transforming image values from linear space to uniform noise space using a nonlinear transformation. This parameter change allows the noise filtering to be applied in a space where noise is uniform rather than signal-dependent, thereby enabling effective noise reduction without the need for complex adaptive filtering that would blur edges. The transformation itself adapts to local signal levels, preserving edge sharpness.
2Object-affected harmful factors
If conventional noise filtering methods are applied to reduce noise in digital images, then noise reduction is achieved, but the filtering process becomes complex due to signal-dependent noise characteristics
Solution Approach 1:
The patent transforms the image from linear space to uniform noise space using a parameter change (nonlinear transformation). This transforms the problem from one with signal-dependent noise (requiring complex adaptive filtering) to one with uniform noise (allowing simple uniform filtering). The complexity is reduced because the transformation handles the signal-dependent nature of noise, allowing straightforward filtering in the transformed space.
Solution Approach 2:
The uniform noise space acts as an intermediary between the original linear space and the final processed image. By introducing this intermediate representation, the patent simplifies the filtering process: the transformation to uniform noise space handles the complexity of signal-dependent noise, allowing simple filtering operations, and the inverse transformation returns to the original space. This intermediary eliminates the need for complex signal-dependent filtering logic.
3Shape
If noise filtering is applied to preserve edges, then edge sharpness is maintained, but insufficient smoothing is achieved in non-detail areas
Solution Approach 1:
The patent achieves different filtering behaviors in different regions through the uniform noise space transformation. In edge regions, the transformation preserves sharp transitions, while in non-detail (smooth) regions, the transformation enables effective noise smoothing. This local differentiation is achieved through the transformation's adaptation to local signal characteristics rather than through region-specific filtering operations.
Solution Approach 2:
By changing to uniform noise space, the patent creates a parameter space where the same filtering operation produces locally adaptive results. In smooth regions, the transformation allows strong smoothing, while near edges, the transformation preserves sharpness. This eliminates the need for complex edge-detection and adaptive-filtering logic, as the parameter transformation itself provides the local adaptability.
Data Source
AI summary
Methods and apparatus for reducing or removing noise in digital images. Image noise reduction methods are described that may be applied to grayscale and color images, for example RGB images. An image noise reduction method may, before applying a noise filtering technique, transform the image values from linear space to flat noise space in which the noise is independent of the signal. An edge-preserving noise filtering technique may then be applied to the image in flat noise space. After noise filtering is applied, the image is transformed from flat noise space back to linear space. For color images, the flat noise space may be converted from linear color space to luminance-chrominance space before applying the noise filtering technique so that different filters can be applied to luminance and color channels. After applying the noise filtering technique, the image is converted back to linear color space.