Adaptive Image Noise Reduction Using Edge-Weighted Spatial and Frequency Compositing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods fail to effectively combine noise reduction using spatial and frequency information based on edge strength, leading to suboptimal noise reduction and contrast maintenance in images.
Innovation Solution
An image processing device and method that detect edge-strength information, apply noise-reduction processing using both spatial and frequency information, and composite processed images using adaptive weights, where the compositing ratio of the first processed image (using spatial information) is higher in regions with strong edges and the compositing ratio of the second processed image (using frequency information) is higher in regions with weak edges.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If noise-reduction processing using spatial information is applied uniformly to the entire image, then noise is reduced in all regions, but edge regions suffer from contrast degradation and loss of detail
Solution Approach 1:
The patent applies different noise reduction processing to different regions of the image based on edge strength detection. Specifically, regions with strong edges use one processing method while regions with weak edges use another method, allowing each region to have optimized quality characteristics appropriate to its local features.
Solution Approach 2:
The patent segments the image into different regions based on edge strength thresholds. By dividing the image processing into distinct zones (edge regions vs. non-edge regions), the system can apply specialized processing to each segment, preventing uniform processing from degrading edge quality.
2Object-affected harmful factors
If noise-reduction processing using frequency information is applied uniformly to the entire image, then noise is reduced in all regions, but flat regions suffer from over-smoothing and loss of texture
Solution Approach 1:
The patent applies frequency-based noise reduction selectively to regions with weak edges where it is most effective, while avoiding its application to edge regions where it would cause blurring. This local adaptation preserves texture in flat regions while maintaining edge sharpness.
Solution Approach 2:
The patent segments the image based on edge strength to identify flat regions suitable for frequency-based processing. By separating flat regions from edge regions, the system can apply frequency-domain noise reduction to flat areas without the harmful over-smoothing effect that would occur in edge areas.
3Device complexity
If a single noise-reduction method is used for the entire image, then processing complexity is low, but processing precision and adaptability to different regions are insufficient
Solution Approach 1:
The patent dynamically selects the appropriate noise reduction method for each region based on real-time edge strength detection. The processing approach changes adaptively according to local image characteristics, allowing the system to optimize quality without excessive complexity through automated region-based decision making.
Data Source
AI summary
An image processing device includes a computer that is configured to: detect edge-strength information expressing an edge strength in an acquired image; apply noise-reduction processing using spatial information to the image; apply noise-reduction processing using frequency information to the image; and composite a first processed image subjected to the noise-reduction processing using spatial information and a second processed image subjected to the noise-reduction processing using frequency information, by using weights in which a compositing ratio of the first processed image becomes higher than a compositing ratio of the second processed image, in a region where the edge strength, which is expressed by the detected edge-strength information, is greater than a predetermined threshold, and in which the compositing ratio of the second processed image becomes higher than the compositing ratio of the first processed image, in a region where the edge strength is less than the threshold.