A time-domain high-pass non-uniformity correction method based on gray-scale correlation

Abstract

The invention discloses a time-domain high-pass non-uniformity correction method based on grayscale correlation, relates to a time-domain high-pass non-uniformity correction method based on grayscale correlation used in the field of infrared imaging, and belongs to the technical field of infrared imaging. The invention adopts a one-point non-uniformity correction model related to the incident radiation value, and uses the spatial low-pass filtering result with edge protection as a correction reference source to pre-correct the input image. Combined with time-domain high-pass filtering, the correction bias value of each frame is calculated, and the mapping relationship between the bias value and the gray level is changed according to the change of the incident radiation at the same position of each frame to complete the correction bias value of the next frame, and eliminate the """ in the correction process. Ghosting" to improve infrared imaging quality. The present invention can reduce the probability of "ghosting" and "over-correction" appearing in the non-uniformity correction algorithm of the real-time infrared imaging system, and improve the quality of infrared imaging.

Description

technical field [0001] The invention relates to a time-domain high-pass non-uniformity correction method based on gray-scale correlation, in particular to a gray-scale correlation-based time-domain high-pass non-uniformity correction method used in the field of infrared imaging, and belongs to the technical field of infrared imaging. Background technique [0002] The fixed pattern noise caused by the non-uniformity of infrared imaging devices is a key factor affecting its imaging quality, and it is necessary to introduce the method of non-uniformity correction (Nonuniformity Correction) in the subsequent image processing to eliminate the noise. The methods of non-uniformity correction mainly include Calibration Based Non-uniformity Correction (CBNUC) and Scene Based Non-uniformity Correction (SBNUC). [0003] CBNUC needs to add a radiation baffle in front of the detector as a uniform radiation reference source required by the calibration algorithm. However, because the algor...

AI Technical Summary

Problems solved by technology

However, these adaptive algorithms are prone to "ghosting", and unreasonable convergence conditions during the calibration process will cause "over-correction"

For example, the neural network algorithm based on the mean value filter is easy to lose details in the filtering window to generate an improper reference source, causing the appearance of "ghosting"; the subsequent neural network algorithm based on bilateral filtering can maintain details to a certain extent to generate a stable reference source, However, in the case of large scene changes in the front and back frames, the correction parameters of the neural network are difficult to update correctly, causing "ghosting"

Images