Image Fusion Using Gray Level Distribution Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for combining digital images often result in output images with low contrast due to 'clumped' data sets, where the distribution of gray levels tends to concentrate in the central range rather than the outer edges, leading to visually unappealing results.
Innovation Solution
A look-up-table (LUT) and algebraic approach are used to control the distribution of gray levels in the fused image, ensuring increased contrast by calculating gray level values based on a probability distribution function and cumulative distribution function, without requiring large memory storage or processing power.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional weighted combination methods are used to fuse two digital images, then the computational complexity is minimized, but the output image contrast deteriorates due to clumped data distribution in the central gray level range
Solution Approach 1:
The patent transforms the output gray level values using a monotonic transformation function that redistributes the clumped central values to expand the dynamic range. This parameter transformation changes the distribution of output values without requiring complex computational operations, thereby improving contrast while maintaining computational efficiency.
Solution Approach 2:
Instead of directly computing the weighted combination and then enhancing contrast through complex post-processing, the patent inverts the approach by applying a transformation function that pre-distributes the values to achieve uniform spacing. This inversion of the conventional workflow simplifies the overall process while achieving better contrast.
2Ease of operation
If conventional fusion methods are used, then the processing workflow is simple, but extensive downstream processing like dynamic range adjustment and gamma correction is required to achieve visually appealing images
Solution Approach 1:
The patent performs the contrast optimization action in advance by applying the monotonic transformation function during the fusion process itself. This preliminary action ensures that the output image is already optimally distributed, eliminating the need for subsequent dynamic range adjustment and gamma correction steps.
Solution Approach 2:
The patent merges the contrast optimization step with the image fusion operation by integrating the monotonic transformation into the fusion algorithm. This combination eliminates separate processing steps, simplifying the overall workflow and reducing downstream processing requirements.
3Illumination intensity
If a look-up table approach is used to achieve fixed distribution, then the contrast is maintained, but large memory storage and processing power are required
Solution Approach 1:
The patent replaces the mechanical look-up table system with a mathematical monotonic transformation function. Instead of storing pre-computed values in a large memory table, the system uses a compact mathematical formula that can be evaluated on-the-fly, significantly reducing memory storage requirements while maintaining the same contrast enhancement效果.
Solution Approach 2:
The patent creates a simplified mathematical model (the monotonic transformation function) that replicates the effect of the complex look-up table approach. This mathematical copy achieves the same contrast optimization purpose with minimal computational resources and no large memory storage requirements.
Data Source
AI summary
A method of forming an output image based on first and second input images includes the steps of: (a) selecting a desired mix-ratio between the first and second input images; (b) determining an angle φ based on the selected mix-ratio; and (c) extending a diagonal line, at the angle of φ between a vertical axis and a horizontal axis, in which the vertical axis includes a first set of gray levels from the first input image and the horizontal axis includes a second set of gray levels from the second input image. The method further includes the steps of: (d) computing an area A included within the horizontal and vertical axes and the diagonal line, in order to determine a gray level value of a pixel of the output image based on a probability value of k=A/n2, where, n is a maximum gray level value of each of the input images, and k varies from 0 to 1. In one embodiment, the fixed probability distribution is uniform, and the input and output image bit depths are equal to each other. Then the gray level value of z of the output image becomes A/n.