Multispectral Image Contrast via Segmented Fisher Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current multispectral image processing methods, such as Fisher projection, face challenges in effectively detecting targets in complex terrestrial environments, especially when targets are camouflaged or have varying spectral characteristics, leading to reduced contrast and potential missed detections due to variegated targets, small targets, parasitic scene elements, and environmental factors like movement and shake.
Innovation Solution
The method enhances target detection by optimizing contrast within separate target areas within an analysis window, using distinct background zones and weighting intensity values based on distance to minimize spectral interference, allowing for improved contrast enhancement and robustness against environmental factors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single analysis window with uniform background zone is used for Fisher projection, then the processing is simple and fast, but the contrast enhancement is reduced when targets are variegated, small, or surrounded by parasitic elements
Solution Approach 1:
The patent divides the analysis window into multiple target areas, each with its own dedicated background zone. This segmentation allows each target to be processed independently with optimized background subtraction, improving contrast enhancement for diverse target types while maintaining computational efficiency through localized processing.
Solution Approach 2:
The patent applies different processing parameters and background zones to different target areas within the same image. Each target area receives customized contrast enhancement based on its local characteristics, allowing optimal detection of variegated, small, or camouflaged targets without compromising overall processing performance.
2Reliability
If the analysis window is positioned to capture potential targets, then target detection is improved, but the detection becomes sensitive to positioning errors caused by environmental factors like movement and shake
Solution Approach 1:
By dividing the analysis window into multiple target areas with individual background zones, the patent reduces the sensitivity to positioning errors. Each segmented area can be independently optimized, so that minor positioning shifts do not cause complete loss of target detection across the entire window.
Solution Approach 2:
The patent dynamically adjusts processing parameters for each target area based on local characteristics and positioning conditions. This adaptability allows the system to maintain reliable target detection even when environmental factors cause shifts in analysis window positioning.
3Ease of manufacture
If background zones are selected to represent typical background, then processing is straightforward, but parasitic scene elements with different spectral characteristics reduce contrast and cause missed detections
Solution Approach 1:
The patent selects background zones specifically for each target area based on local spectral characteristics rather than using a uniform background model. This local adaptation allows the system to account for parasitic elements and spectral variations in different regions, improving contrast accuracy without significantly complicating the processing workflow.
Solution Approach 2:
The patent performs preliminary analysis of background zones to identify and exclude parasitic elements before conducting Fisher projection. This preliminary action ensures that background representation is optimized for each target area, improving detection accuracy while maintaining processing efficiency.
Data Source
Figure 1~2
Figure 3
Figure 4a~4c
AI summary
A method for visualizing a multispectral image employs several target areas (T1, T2, T3, T4) selected within an analysis window (AF). A Fisher projection is determined separately for each target area to optimize the contrast of the multispectral image within the analysis window. Contrast-enhanced elementary images are then obtained for each target area within the same analysis window. A composite image is then constructed from these elementary images and displayed on an image display unit. The invention improves the visualization of the multispectral image, even for specific configurations of the scene content captured within the multispectral image.