Image Sensor Noise Reduction Preserving Edges
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Solid state image sensors face challenges in noise reduction, especially with small pixel sizes, as existing methods either blur edges or require significant computational resources, and there is a need for an efficient denoising method that does not affect image quality.
Innovation Solution
A noise reduction method that compares the difference between a target pixel and its neighboring pixels in the same color, replacing the target pixel value with an average of neighboring pixels only if the difference is below a set threshold, thereby averaging and reducing noise without blurring edges, and can be applied in various color spaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If local smoothing filters (averaging, medium, Gaussian filters) are applied to reduce noise, then noise is reduced, but edges of the image are blurred
Solution Approach 1:
The patent applies different processing strategies to different regions of the image based on local characteristics. By detecting edges through gradient calculations and comparing them against thresholds, the system applies smoothing only to regions identified as noise (where gradients are small) while preserving edges (where gradients are large). This local differentiation allows simultaneous noise reduction and edge preservation.
Solution Approach 2:
The patent dynamically adjusts processing parameters based on local image characteristics. The threshold for gradient comparison is adapted based on the specific region being processed, and the degree of smoothing is modified according to the detected edge strength. This parameter adaptation enables the system to distinguish between noise and edges and apply appropriate processing accordingly.
2Object-affected harmful factors
If frequency domain filters (DFT filters, wavelength transform filters) are used for image denoising, then noise is reduced, but computational resources are significantly increased
Solution Approach 1:
The patent replaces complex frequency domain processing with simpler spatial domain operations. Instead of using computationally intensive Fourier transforms and frequency filtering, the system performs direct spatial comparisons and averaging operations on pixel values and their neighbors. This substitution maintains effective noise reduction while dramatically reducing computational complexity for silicon-on-a-chip implementation.
Solution Approach 2:
The patent divides the image processing into discrete, manageable operations: edge detection through gradient calculation, threshold comparison, and selective averaging. This segmentation of the processing task into simple, sequential steps makes the algorithm suitable for implementation in resource-constrained imaging devices while maintaining effectiveness.
3Device complexity
If simple noise reduction algorithms are used, then computational resources are reduced, but edges of the images are blurred
Solution Approach 1:
The algorithm applies selective smoothing based on local edge detection. By calculating gradients and comparing them to thresholds, the system identifies regions containing edges versus regions containing noise. Smoothing is applied only to noise regions, while edge regions are preserved. This local quality differentiation achieves effective noise reduction without the computational burden of complex global algorithms.
Solution Approach 2:
The processing threshold and smoothing application are dynamically adjusted based on the detected edge strength in each region. The system adapts its behavior in real-time based on the image content, applying more aggressive smoothing where noise is detected and preserving edges where gradient information indicates boundary structures. This dynamic adaptation enables simple algorithms to achieve complex results.
Data Source
AI summary
A method of reducing noise in an image including steps for obtaining a first value for a target pixel, obtaining a respective second value for each neighboring pixel surrounding the target pixel and having the same color as the target pixel, for each neighboring pixel, comparing a difference between said first value and said second value to a threshold value, and replacing the first value with an average value obtained from the first value and at all second values from the neighboring pixels which have an associated difference which is less than or equal to the threshold value based on a result of the comparing step.