Learning-Based DCT Noise Filter for Low-Light Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise filters for images captured in low light conditions struggle due to changing noise profiles with light conditions and sensor/optical configurations, and AI-based filters face high computational complexity and lack of tunability.
Innovation Solution
A learning-based discrete cosine noise filter that applies a discrete cosine transform (DCT) process to input images, including gamma correction, separation of luminance and chroma channels, division into overlapping patches, suppression of patches with DCT coefficients within a threshold range, and inverse DCT transformation to generate a noise-filtered output image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI-based noise filters are used, then noise suppression performance is improved, but computational complexity increases significantly
Solution Approach 1:
The patent divides the image into multiple overlapping patches and processes each patch independently through DCT transformation. This segmentation allows the noise filtering to be applied locally rather than globally, reducing the computational burden while maintaining effective noise suppression. The overlapping patches ensure that edges and details are preserved across patch boundaries.
Solution Approach 2:
The patent applies gamma correction to transform the input image from original space to perceptual space, changing the parameter space of the image data. This transformation modifies the noise characteristics and enhances the effectiveness of subsequent DCT-based filtering. The parameter change enables better noise suppression while reducing computational complexity compared to direct AI-based methods.
2Adaptability or versatility
If a single noise filter is designed for different light conditions, then adaptability is improved, but filter performance deteriorates due to noise profile changes
Solution Approach 1:
The patent implements a dynamic noise filtering approach where the DCT threshold is adaptively adjusted based on the input image characteristics. The filtering parameters are not fixed but can be tuned according to lighting conditions and sensor configurations. This dynamic adjustment allows the same filter to maintain optimal performance across different light conditions without requiring multiple specialized filters.
Solution Approach 2:
The DCT-based noise filter is designed as a universal filter that can handle multiple light conditions and sensor configurations through a single unified approach. By using perceptual space transformation and adaptive thresholding, the filter achieves multi-functionality without sacrificing performance, eliminating the need for separate filters for different lighting scenarios.
3Reliability
If gamma correction and DCT processing are applied, then noise suppression is improved, but processing time increases
Solution Approach 1:
By dividing the image into overlapping patches and processing them in parallel, the patent reduces the effective processing time. The segmentation allows for efficient memory access patterns and enables concurrent processing of multiple regions, significantly reducing the time penalty associated with gamma correction and DCT transformations while maintaining comprehensive noise suppression.
Solution Approach 2:
The patent applies different processing intensities to different regions of the image based on local characteristics. Areas with higher noise content receive more aggressive filtering, while regions with fewer details are processed more lightly. This local quality approach optimizes the balance between noise suppression and processing time by avoiding unnecessary computations in low-noise regions.
Data Source
AI summary
A method includes filtering an input image through a discrete cosine transform based noise filter (DCT-NF). The DCT-NF converts the input image from an original space into a perceptual space, applying gamma correction. The DCT-NF separates luminance channels of the input image from chroma channels of the input image. The DCT-NF divides the input image into overlapping patches and computes DCT transform of the patches. Each patch is a different partial portion of the input image. The DCT-NF suppresses patches that include an input DCT coefficient within a threshold range. The DCT-NF applies an inverse discrete cosine transform (IDCT) to the suppressed patches and remaining overlapping patches that include an input DCT coefficient outside the threshold range. The DCT-NF re-combines luminance and chroma channels of the IDCT-transformed patches. The DCT-NF generates a DCT noise-filtered output image by re-converting the IDCT-transformed patches to the original space by applying inverse gamma correction.