Hierarchical Noise Filtering for Digital Camera Image Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital cameras and camera phones capture images with significant noise levels under low-light conditions, especially with CMOS sensors, which degrade image quality.
Innovation Solution
A hierarchical noise filtering method using two-tap filters and non-linear mapping to create a hierarchical representation of images, modifying coefficients at different scales to remove noise while preserving edge structures, implemented in digital cameras and camera phones using DSPs or other processors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If high ISO gain is used to capture images under low-light conditions, then image brightness is improved, but noise level increases significantly
Solution Approach 1:
The image is decomposed into multiple scales using a hierarchical representation (wavelet transform), separating the image into approximation coefficients (LL) and detail coefficients (LH, HL, HH) at different levels. This segmentation allows noise filtering to be applied selectively at different frequency bands and scales, preserving important image features while removing noise.
Solution Approach 2:
Different filtering strengths are applied to different regions and scales of the image. The modification function adjusts coefficient values based on local characteristics, applying stronger noise reduction in homogeneous regions while preserving edges and important features in other regions. This local adaptation optimizes the balance between noise reduction and feature preservation.
2Object-affected harmful factors
If noise filtering is applied to reduce noise levels, then image quality is improved, but edge structures may be destroyed or blurred
Solution Approach 1:
The filtering process is made adaptive and dynamic through the modification function that adjusts coefficient attenuation based on local image characteristics. The function dynamically determines the appropriate filtering strength for each coefficient based on its magnitude and local context, allowing the filter to adapt to different regions (smooth areas vs. edges) rather than applying a fixed filtering strength throughout.
Solution Approach 2:
The modification function changes the parameters (coefficient values) in a non-linear manner based on their magnitude and local statistics. By using a non-linear modification function rather than simple linear thresholding, the system can preserve important features while removing noise, as the function adapts its behavior based on the local image characteristics.
Data Source
AI summary
Image noise reduction filtering by low-pass/high-pass filtering to get a hierarchical representation, modifying coefficients in each hierarchy level for noise suppression, and the modified level combination to yield a noise-filtered image. The noise suppression within levels preserves edges which the representation preserves.