Thermal Image Noise Reduction via Frame Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Thermal imaging cameras face challenges in distinguishing features in low contrast thermal images due to noise, which can be exacerbated by existing image averaging techniques that introduce issues like blurred edges and ghosting effects.
Innovation Solution
A hand-held thermal imaging camera and method that utilize a processor to capture and align multiple IR frames, reducing noise by averaging them after alignment to improve image quality, with the option to also use visible light frames for alignment, thereby enhancing the accuracy of the thermal image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image averaging techniques are used to reduce thermal noise, then noise reduction is improved, but image quality deteriorates due to blurred edges and ghosting effects
Solution Approach 1:
The patent applies preliminary alignment action by performing rigid transformation (translation and rotation) on multiple thermal images before averaging them. This preprocessing step ensures that corresponding features across images are properly positioned, preventing the blurred edges and ghosting effects that would otherwise occur during averaging. The alignment is calculated using feature detection and matching algorithms that identify and correct spatial misalignments before the noise reduction averaging process.
2Measurement precision
If multiple frames are averaged to reduce noise, then signal-to-noise ratio is improved, but processing complexity increases
Solution Approach 1:
The patent segments the image processing task into distinct stages: feature detection, feature matching, alignment calculation, transformation application, and averaging. This segmentation allows each step to be optimized independently and enables parallel processing of multiple images during the alignment and averaging stages, reducing overall processing complexity while maintaining high signal-to-noise ratio improvement.
Solution Approach 2:
The patent implements dynamic processing by adaptively selecting the number of frames to average based on noise levels and by using real-time feature matching to determine alignment parameters. The system dynamically adjusts processing intensity - using fewer frames when noise is low and more frames when noise is high - thereby optimizing the balance between processing complexity and signal-to-noise ratio improvement.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Devices and methods for generating infrared (IR) images with improved quality are disclosed. In embodiments of the invention, temporal averaging techniques are used to reduce temporal noise that may be present in thermal images. This is especially useful in low-contrast thermal scenes, where a relatively small amount of thermal noise may become exceedingly prevalent. In order to average properly, some embodiments of the invention provide methods or means for aligning multiple images that are to be averaged together to eliminate inaccuracies and misrepresentations that may result from averaging misaligned images.