Adaptive Bilateral Filter for Low-Light Image Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital images captured in low-light conditions often suffer from noise, leading to degraded quality and authenticity, especially in inexpensive camera sensors like those found in smartphones, where conventional denoising filters fail to effectively reduce noise without blurring edges.
Innovation Solution
An adaptive bilateral filter is applied to digital images, combining a domain filter and a range filter that are intensity-dependent, allowing for adaptive kernel sizes based on pixel intensity, effectively reducing noise while preserving edges.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional denoising filters are applied to reduce noise in low-light images, then noise reduction is achieved, but edge blurring occurs and image quality deteriorates
Solution Approach 1:
The patent applies different filtering strengths to different regions of the image based on local noise characteristics. The bilateral filter adapts its kernel size and filtering intensity locally, applying stronger denoising in high-noise regions while preserving edges in low-noise regions, thereby achieving noise reduction without uniform edge blurring
Solution Approach 2:
The patent employs an adaptive bilateral filter where the filter parameters (kernel size, sigma values) are dynamically adjusted based on local image characteristics such as gradient magnitude and noise variance. This dynamic adaptation allows the filter to respond to varying noise levels across different image regions, maintaining edge sharpness while reducing noise
2Object-affected harmful factors
If stronger denoising is applied to reduce noise effectively, then noise reduction improves, but edge preservation deteriorates
Solution Approach 1:
The filter applies different processing intensity to different image regions: stronger filtering in homogeneous regions with high noise and weaker filtering near edges. The domain filter and range filter work together to selectively attenuate noise while preserving edge structures through local adaptation
Solution Approach 2:
The patent changes filter parameters (sigma domain, sigma range, kernel size) based on local image characteristics. By adapting these parameters to local noise variance and gradient information, the filter achieves stronger noise reduction in appropriate regions while maintaining edge sharpness through parameter modulation
3Object-affected harmful factors
If adaptive kernel sizes based on pixel intensity are used, then denoising effectiveness improves, but computational complexity increases
Solution Approach 1:
The patent segments the image processing into distinct computational stages: gradient calculation, noise variance estimation, adaptive parameter determination, and filtering application. This segmentation allows for optimized computation at each stage and facilitates parallel processing implementations
Data Source
AI summary
An apparatus comprising one or more processors configured to receive a digital image comprising a current pixel and a plurality of nearby pixels, determine a pixel intensity of the current pixel, wherein the pixel intensity comprises a noise component, and reduce the noise component by applying a bilateral filter as a combination of a domain filter and a range filter on the current pixel, wherein the domain filter is dependent on the pixel intensity and geometric closeness between the current pixel and the nearby pixels, and wherein the range filter is dependent on the pixel intensity and photometric similarity between the current pixel and the nearby pixels.