Real-time infrared non-uniformity correction method and device based on fuzzy registration and medium

Abstract

The invention discloses a real-time infrared non-uniformity correction method and device based on fuzzy registration and a medium. The method comprises the steps that two adjacent frames of infrared images are acquired from an infrared image sequence; performing image registration on the two frames of infrared images through an image matching method to obtain an overlapping region of the two frames of infrared images; searching the nearest pixel value as an expected value in a set size window taking a corresponding point in the overlapping region as a center, and establishing a mean square error matrix; and updating the correction parameter matrix through the established mean square error matrix to obtain a corrected output image. In this way, the actual infrared thermal imaging video sequence is processed through a fuzzy registration method; according to the invention, non-uniform real-time correction can be realized, and since the pixel corresponding relationship in the overlapping region is not simple one-to-one mapping, image degradation caused by sub-pixel offset, rotation and heat residue in actual use can be effectively overcome, and generation of ghosting is effectively inhibited.

Description

technical field [0001] The invention relates to the technical field of infrared thermal imaging, in particular to a real-time infrared non-uniformity correction method, equipment and medium based on fuzzy registration. Background technique [0002] Infrared thermal imaging technology can obtain information related to thermal motion, and is widely used in many fields such as military, aerospace, industry, medical treatment, and security. However, affected by factors such as the manufacturing process, materials, electronic architecture, and optical system architecture of infrared focal plane detectors, the non-uniformity of infrared images has become a research hotspot in infrared thermal imaging technology. [0003] At present, non-uniformity correction methods are divided into two types: calibration correction based on reference source and adaptive correction based on scene. Among them, the calibration correction method based on the reference source usually needs to obtain ...

AI Technical Summary

Problems solved by technology

Among them, the neural network non-uniformity correction method is difficult to achieve a balance between the convergence speed and the correction effect. When the non-uniformity is strong, the image is prone to serious degradation and "ghosting"; the constant constant statistical method and the time-domain high-pass filter The method requires the randomness of the image information, and its practical application is limited; the correction effect of the inter-frame registration method is affected by the registration accuracy, and the slight rotation of the image in actual use is unavoidable. The cost of improving the registration accuracy is to increase the amount of calculation, which is difficult to achieve Real-time correction of non-uniformity, while reducing the amount of calculation, inaccurate matching will introduce "ghosting"

Images