Multi-resolution Image Noise Removal via Segmented Frequency Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise removal methods from high-frequency subbands in digital images often result in residual noise as projecting points or streaks, while methods targeting low-frequency subbands can make images appear flat and lose texture, and gamma ray images present unique challenges due to differing characteristics. Additionally, current methods struggle with effective noise removal in color images using multiple-channel frequency bands.
Innovation Solution
An image processing method that involves multiple resolution image generation, where noise is individually removed from both low-frequency and high-frequency images, with specific noise removal processing in real space and varying resolution levels to optimize noise extraction and preservation of image structure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If noise removal processing is applied to high-frequency subbands, then noise is reduced, but residual noise appears as projecting points or streaks
Solution Approach 1:
The patent segments noise removal into two distinct stages: first removing noise from high-frequency subbands (LH, HL, HH), then removing residual noise from the low-frequency subband (LL). This segmentation allows each stage to target specific noise characteristics, with the second stage specifically addressing residual noise that persists after the first stage, thereby reducing projecting points and streaks while maintaining image structure.
Solution Approach 2:
The patent applies preliminary noise removal to high-frequency subbands before processing the low-frequency subband. By removing the majority of noise in the high-frequency components first, the subsequent low-frequency processing only needs to address residual noise, making the overall noise removal more effective and reducing harmful artifacts.
2Reliability
If noise removal processing is applied to low-frequency subbands, then noise is reduced, but images appear flat and lose texture
Solution Approach 1:
The patent segments the noise removal process by first handling high-frequency subbands where texture information resides, and only then processing the low-frequency subband. This ensures that texture characteristics are preserved in the high-frequency components while noise removal is applied to the low-frequency components, preventing the image from appearing flat.
Solution Approach 2:
The patent applies different noise removal strategies to different frequency subbands: aggressive noise removal for high-frequency subbands (where noise is prominent but texture is less critical) and more conservative noise removal for the low-frequency subband (where texture and overall image structure are critical). This local differentiation preserves image quality while effectively removing noise.
3Measurement precision
If multiple resolution transformation is used for noise removal, then noise extraction is improved, but processing complexity increases
Solution Approach 1:
The patent segments the multiple resolution transformation into a standard wavelet decomposition followed by a separate noise removal processing stage. By using conventional wavelet transformation and then applying noise removal to the decomposed subbands, the patent achieves precise noise extraction without requiring complex custom transformation algorithms, thus managing processing complexity.
Data Source
AI summary
An image processing method adopted to remove noise present in an image, includes: an image input step in which an original image constituted of a plurality of pixels is input; a multiple resolution image generation step in which a plurality of low-frequency images with resolutions decreasing in sequence and a plurality of high-frequency images with the resolutions decreasing in sequence are generated by decomposing the input original image; a noise removal processing step in which noise removal processing is individually executed on the low-frequency images and the high-frequency images; and an image acquisition step in which a noise-free image of the original image is obtained based upon both the low-frequency images and the high-frequency images having undergone the noise removal processing.