Thermal Image Shading Correction Using Multi-Scale Downsampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional thermal imaging systems face challenges in efficiently removing shading caused by extreme temperatures, rain, snow, moisture, and heat sources, which obscure objects of interest and require computationally intensive methods not suitable for portable systems.
Innovation Solution
A method and system that accurately calculate and correct shading in thermal images by adjusting pixel values and performing tone optimization, using equations to compress and expand pixel ranges, downsampling and upsampling images, and combining weighted sums to generate a shading image for subtraction from the original image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional shading removal approaches are used, then shading can be removed from thermal images, but the processing becomes computationally intensive and not suitable for portable systems
Solution Approach 1:
The patent transforms the shading removal problem from a complex spatial filtering task into a simpler parameter-based correction by modeling shading as a multiplicative factor that varies across image coordinates. The solution changes the approach from analyzing spatial patterns to applying coordinate-based parameter adjustments, making the computation suitable for portable systems while maintaining effectiveness.
Solution Approach 2:
The patent creates a simplified computational model that copies the essential shading pattern characteristics without requiring full complex processing. By estimating shading parameters from image statistics and applying a simplified correction model, the system achieves shading removal equivalent to complex methods but with much lower computational requirements.
2Reliability
If conventional shading removal approaches are used, then shading can be removed from thermal images, but the methods are not readily implementable in portable systems with limited processing resources
Solution Approach 1:
The patent reduces energy consumption by changing from computationally intensive spatial filtering operations to simple parameter calculations and multiplicative corrections. The solution computes shading parameters using basic image statistics and applies efficient pixel-wise corrections, dramatically reducing processing energy requirements for portable systems while maintaining shading removal effectiveness.
3Loss of information
If pixel values are adjusted to reduce shading, then contrast and clarity of objects of interest improve, but the dynamic range of pixel values must be carefully managed
Solution Approach 1:
The patent carefully manages pixel value precision by modeling shading as a multiplicative parameter and applying corrections that preserve the relative relationships between pixel values. The method calculates shading parameters from image statistics and applies coordinated compression and expansion to different pixel value ranges, maintaining precision while recovering detail visibility in shaded regions.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
Various techniques are provided to reduce shading in thermal images. In one example, a method includes receiving a captured thermal image having an original resolution, wherein the captured thermal image comprises scene information and shading information. The method includes processing the captured thermal image to generate a plurality of downscaled images each having an associated reduced resolution lower than the original resolution and different from each other, wherein the downscaled images exhibit reduced scene information in relation to the captured thermal image. The method includes generating a shading image using the downscaled images. The method includes adjusting the captured thermal image using the shading image to generate a processed thermal image with reduced shading information in relation to the captured thermal image. Additional methods and systems are also provided.