Homomorphic Image Processing for Night Vision Contrast Enhancement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-dynamic, high-resolution digital video imaging systems, especially in low-light conditions, face challenges such as noise amplification, degradation of wide luminous dynamics, appearance of artifacts, and incompatibility with real-time processing at high frame rates, leading to decreased image quality and loss of scene aspect.
Innovation Solution
A homomorphic filtering method that processes images by calculating the logarithm of the initial image, applying a low-frequency Gaussian filter, determining intensity values, and performing a linear combination to enhance contrast without increasing noise, suitable for real-time operation and dynamic adaptation between sensor and display.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional contrast enhancement methods are applied to low-light digital video images, then image contrast is improved, but digital noise is strongly increased
Solution Approach 1:
The image processing method segments the image into different frequency components using a low-pass filter. The low-frequency components (representing overall illumination) are separated from high-frequency components (representing fine details and noise). This segmentation allows selective enhancement of contrast in low-frequency regions without amplifying noise in high-frequency regions.
Solution Approach 2:
The patent applies different processing characteristics to different frequency regions. The low-pass filtering operation applies a smoothing kernel that preserves low-frequency information while attenuating high-frequency noise. The contrast enhancement is then applied selectively to the filtered low-frequency components, creating local quality differentiation between frequency domains.
2Measurement precision
If conventional contrast enhancement methods are applied, then low gray level contrasts are enhanced, but saturation of overexposed areas occurs
Solution Approach 1:
By segmenting the image into low-frequency and high-frequency components, the method processes different regions with different characteristics. The low-pass filter separates the overall illumination pattern from local variations, allowing contrast enhancement to be applied primarily to the low-frequency component without affecting the dynamic range of overexposed high-frequency regions.
3Measurement precision
If conventional image processing methods are applied, then contrast is enhanced, but processing compatibility with real-time operation at high frame rates is lost
Solution Approach 1:
The method segments the computationally intensive contrast enhancement task into a two-stage process: first applying a low-pass filter to obtain low-frequency components, then enhancing contrast on this simplified representation. This segmentation reduces the overall computational burden compared to applying complex contrast enhancement to the full-resolution image.
Solution Approach 2:
The patent applies partial action by enhancing contrast only on the low-frequency component obtained through low-pass filtering, rather than processing the entire image at full resolution. This partial processing approach significantly reduces computational requirements while maintaining perceptual quality, enabling real-time operation at high frame rates.
4Measurement precision
If conventional contrast enhancement methods are applied, then local contrasts are improved, but resolution of the initial image decreases
Solution Approach 1:
The patent segments the image processing into frequency domains using low-pass filtering. The contrast enhancement is applied to the low-frequency component, while the high-frequency component (containing fine detail and resolution information) is preserved. This segmentation allows local contrast improvement without degrading the resolution carried by high-frequency components.
Solution Approach 2:
After separate processing of the low-frequency component (contrast enhancement via low-pass filtering), the method merges the processed low-frequency component back with the original high-frequency component. This merging operation combines the enhanced local contrast from the low-frequency part with the fine detail and resolution information from the high-frequency part, achieving both goals simultaneously.
Data Source
Figure 1
Figure 2
AI summary
The general field of the invention is that of methods for processing an initial digital image IIN(p) composed of pixels (p). The method according to the invention comprises the following successive steps: Step 1: Calculation of the logarithm of the initial image to obtain a first intermediate image denoted Log(IIN(p)); Step 2: Filtering of said first intermediate image by a low-frequency filter to obtain a second intermediate image denoted BF(p); Step 3: Determination of the minimum intensity value and the maximum intensity value in the low-frequency image; Step 4: Calculation of a third intermediate image denoted Log(IOUT(p)) by the following linear combination: LogIOUTp=Log(IINp-BFp.K1.BFp+K2+K3.BFp+K4 K1, K2, K3 and K4 being constants; Step 5: Calculation of the final image IOUT(p) by applying the exponential function to the third intermediate image.