Infrared image non-uniformity correction method

Abstract

The invention discloses an infrared image non-uniformity correction method, which is used for processing an infrared image of a measured object obtained by an acquisition system, and comprises the following steps: step 1, correcting the obtained infrared image by adopting a normalized BP neural network; and step 2, correcting the infrared image corrected in the step 1 by adopting definition improvement based on region attribute classification. On the basis of a traditional artificial neural network, the application of a classic BP neural network algorithm in heterogeneity is optimized from twoaspects, namely a normalization neural network algorithm and a heterogeneity correction algorithm based on regional attribute classification; wherein the normalized BP artificial neural network algorithm aims to eliminate the difference between the order of magnitudes of gain and offset adjustment amount and improve the convergence rate and correction effect of the algorithm, and the latter focuses on improving the definition of the image, so that the efficiency and quality of the algorithm are jointly optimized, and the effect of advantage complementation is achieved.

Description

technical field [0001] The invention relates to infrared image processing, in particular to an infrared image non-uniformity correction method. Background technique [0002] The non-uniformity of the infrared image refers to the inconsistency of the response output of the detector array. The non-uniformity of the infrared image is caused by the non-uniformity of the device itself, the non-uniformity introduced by the working state of the device, and the non-uniformity related to the external input, which is serious. It affects the imaging quality of the infrared sensor, so infrared non-uniformity correction must be performed. [0003] There are many ways to compensate the non-uniformity of the focal plane detector, but because the source of this kind of noise is different from the transient noise, the usual processing method cannot be adopted. So suppressing inherent spatial noise is more difficult than suppressing transient noise. So far, no algorithm with strong adaptabi...

AI Technical Summary

Problems solved by technology

However, this algorithm has two main defects: one is that in the process of adjusting the gain, the selection of the step size and the speed of convergence are mutually restricted; the other is that in the process of estimating the real image, the pixel extraction does not take into account Area attributes, resulting in low correction accuracy, which affects the clarity of the image

Images