Multi-layer Image Processing for Low-brightness Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies fail to effectively remove low-frequency noise with periodicity equal to or more than the size of a smoothing filter and exhibit insufficient noise reduction performance in low-brightness regions where noise components are easily noticeable.
Innovation Solution
An image processing method that calculates pixel statistical values and edge information across multi-layers with progressively decreasing ranges, corrects difference information using edge information, and iteratively refines pixel values until the area range reduces from maximum to minimum, employing correction functions to enhance noise suppression in low-brightness areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a smoothing filter is used to reduce random noise, then noise reduction performance is improved, but low-frequency noise with periodicity equal to or more than the filter size cannot be removed
Solution Approach 1:
The image processing is divided into multiple layers with different smoothing filter sizes. The noise reduction is performed in segments across different layers, allowing both high-frequency and low-frequency noise to be addressed at appropriate scales. Each layer processes noise at its specific frequency range, and the results are combined to achieve comprehensive noise reduction.
Solution Approach 2:
The patent introduces a multi-layer dimensional structure where noise reduction is performed at different scales. By adding the layer dimension, the system can simultaneously handle noise of different frequencies and periodicities that would be impossible to address with a single filter size.
2Reliability
If conventional noise reduction processing is applied, then general noise is reduced, but noise components remain noticeable in low-brightness regions
Solution Approach 1:
The patent applies different noise reduction processing to different regions of the image based on their brightness characteristics. Low-brightness regions receive specialized processing that preserves more detail and reduces noise more effectively, while other regions use standard processing. This local adaptation ensures optimal noise reduction performance across all brightness levels.
Solution Approach 2:
The noise reduction processing is made dynamic by adapting the processing strength and method based on local image characteristics, particularly brightness levels. The system automatically adjusts its behavior to match the local conditions, applying stronger noise reduction where needed in low-brightness regions while maintaining detail preservation.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
The present invention is an image processing method comprising: calculating a pixel statistical value and edge information of pixels for each of areas of a multi-layer, the areas each containing a target pixel and having a successively decreased range; correcting difference information between a pixel statistical value of an area of a specific layer and a pixel statistical value of an area of a layer that is wider than the area of the specific layer using the edge information; correcting the pixel statistical value of the area of the specific layer using post-correction difference information, the pixel statistical value of the area that is wider than the area of the specific layer, and a pixel statistical value of an area that is wider than the areas of other layers; and correcting the target pixel by repeating correction and recorrection of the pixel statistical value of the area of the specific layer successively in each layer until the area reduces its range from the maximum range to the minimum range.