Infrared fringe non-uniformity correction method based on steering kernel estimation

Abstract

The invention discloses an infrared fringe non-uniformity correction method based on steering kernel estimation. The fringe degree of an area with pixels is represented by a steering kernel neighborhood according to non-uniformity of infrared column fringes, a steering kernel estimation result is used as an expected image, least square fitting is performed for the column vector of the expected image and the column vector of a current image, non-uniformity correction parameters of columns are iteratively solved, accordingly, non-uniformity of column fringes of infrared images is corrected, and image edges can be reserved while fringe non-uniformity is removed.

Description

technical field [0001] The invention relates to an image processing technology of an infrared imaging system, in particular to an infrared fringe non-uniformity correction method based on steering kernel estimation. Background technique [0002] Due to the limitations of technology, materials, and device working environment, the response of infrared detector pixels under the same irradiance is not completely consistent, resulting in the problem of non-uniformity in infrared images. Among them, the readout circuit of the infrared scanning line column and the non-refrigerated staring type focal plane usually adopts the column output mode in consideration of the process and cost factors, and the image noise exhibited is mainly the non-uniform interference mainly in the column direction stripe shape. [0003] Usually, within a certain illumination range, the pixel response of the infrared detector under the interference of columnar stripe non-uniformity is approximately a linear...

AI Technical Summary

Problems solved by technology

At present, the scene-based adaptive correction methods mainly include artificial neural network and constant statistical evaluation method, etc. The disadvantage is that the convergence of the algorithm is affected by the iterative step size and initial value, and the calculation amount of the algorithm is relatively large.

For infrared fringe non-uniformity correction methods, the correlation between image sequences is mostly used. For example, Ren Jianle et al. proposed a registration-based correction method in the article "Registration-based Infrared Focal Plane Array Stripe Non-uniformity Correction". Under the assumption that the brightness of the same scene in two adjacent frames of the same infrared detector is constant, the correction parameters are obtained. The algorithm can effectively remove the non-uniformity of stripes between the accurately registered image sequences, but the algorithm has a high accuracy for registration. Require

Images