Thermal Image Noise Reduction via Digital Column Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Infrared imaging devices suffer from noise issues, particularly high spatial frequency fixed pattern noise (FPN) that can be correlated to rows and columns, which hinders the ability to distinguish scene features and is challenging to address due to mechanical failures in conventional shutter-based solutions, especially in devices with small form factors.
Innovation Solution
A method and system that process image frames to determine column correction terms based on relative relationships between pixels in a neighborhood, reducing noise without the need for a mechanical shutter, by using a processor to analyze and correct thermal image data in infrared imaging devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If a mechanical shutter is used to remove fixed pattern noise, then noise reduction is achieved, but device complexity and reliability deteriorate due to mechanical failure risks
Solution Approach 1:
The patent replaces the mechanical shutter system with a digital signal processing approach. The processor analyzes pixel data from the infrared sensor array and applies computational algorithms to remove fixed pattern noise, eliminating moving mechanical parts and their associated reliability issues while achieving the same noise reduction goal
Solution Approach 2:
The patent creates a digital model of the fixed pattern noise by analyzing pixel responses and generates correction terms that replicate the noise pattern. These correction terms are then subtracted from the raw pixel data, effectively copying and removing the noise without requiring physical intervention
2Object-affected harmful factors
If a mechanical shutter is used to remove fixed pattern noise, then noise reduction is achieved, but device size and cost increase
Solution Approach 1:
The patent replaces the mechanical shutter system with a digital signal processing approach. The processor analyzes pixel data from the infrared sensor array and applies computational algorithms to remove fixed pattern noise, eliminating moving mechanical parts and their associated reliability issues while achieving the same noise reduction goal
Solution Approach 2:
The patent enables the infrared imaging device to self-correct its own noise issues through onboard processing. The processor within the device analyzes the pixel data and applies noise removal algorithms, allowing the system to service itself without requiring external mechanical components or additional hardware
3Object-affected harmful factors
If conventional noise filtering is applied, then noise reduction is achieved, but scene features may be distorted
Solution Approach 1:
The patent segments the noise removal process into distinct components: determining correction terms based on pixel relationships, identifying high spatial frequency fixed pattern noise specifically, and applying targeted corrections. This segmentation allows selective noise removal while preserving genuine scene features that do not match the FPN pattern
Solution Approach 2:
The patent applies different processing approaches to different spatial frequencies and noise characteristics. By focusing specifically on high spatial frequency fixed pattern noise and using local pixel relationships to determine correction terms, the method preserves low spatial frequency scene features while removing only the problematic noise components
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods and systems are provided to reduce noise in thermal images. In one example, a method includes receiving an image frame comprising a plurality of pixels arranged in a plurality of rows and columns. The pixels comprise thermal image data associated with a scene and noise introduced by an infrared imaging device. The image frame may be processed to determine a plurality of column correction terms, each associated with a corresponding one of the columns and determined based on relative relationships between the pixels of the corresponding column and the pixels of a neighborhood of columns. In another example, the image frame may be processed to determine a plurality of non-uniformity correction terms, each associated with a corresponding one of the pixels and determined based on relative relationships between the corresponding one of the pixels and associated neighborhood pixels within a selected distance.