Infrared image noise time domain filtering method based on registration

Abstract

The invention discloses an infrared image noise time domain filtering method based on registration. The method comprises the following steps of: acquiring an original infrared image, and calculating an initial noise threshold value Fth(0) of different coordinate positions of the original infrared image; acquiring the original infrared image of the kth frame Forg (k), and calculating the noise threshold value Fth(k) of the original infrared image of the kth frame; carrying out the low-pass filtering on the original infrared image Forg (k) of the kth frame and the time dome filter image Ffilt (k-1) of the (k-1)th frame to obtain a relative displacement (delta x, delta y) between the low-pass image calculated low-pass images for showing the relative displacement of Forg(k) and Ffilt (k-1); and outputting a time domain filter image Ffilt (k) of the kth frame.

Description

technical field [0001] The invention relates to an infrared image processing technology, in particular to a registration-based infrared image noise time-domain filtering method suitable for hardware real-time implementation. Background technique [0002] Infrared thermal imaging system has excellent performance such as all-weather work, long working distance and good concealment, and has been widely used in military and civilian fields. However, in the actual use of the thermal imaging system, if the observed target is too far away, or the weather conditions are relatively poor, the energy of the target will decrease, the noise will increase relatively, and the signal-to-noise ratio of the infrared image obtained will decrease. [0003] The research on image denoising has a long history. The main idea is to model the image pixels in the space or transformation domain, analyze the feature information of the image, and then process the image. In the spatial domain, denoising ...

AI Technical Summary

Problems solved by technology

The transform domain denoising algorithm can be divided into signal-based adaptive algorithm and non-adaptive algorithm according to the selection of basis functions: independent component analysis is a classic signal adaptive algorithm, which linearly decomposes the observed signal and converts It is decomposed into statistically independent components, which have been successfully implemented in data denoising, but the disadvantage is high computational complexity; the frequency domain denoising algorithm based on Fourier transform is a non-adaptive transform domain denoising algorithm, through fast Fourier transform Transform transforms the signal into the frequency domain, and denoises the image by applying a low-pass filter

The existing image noise filtering algorithms have the following disadvantages: (1) most of the existing spatial domain filtering algorithms do not consider the correlation between image frames, and will lose the high-frequency detail information of the image; (2) most of the existing temporal domain filtering algorithms are in Statistical characteristics of noise are not considered in image motion estimation, which will cause motion blur and "ghosting" phenomenon in the process of image noise suppression; (3) Most existing temporal filtering algorithms require large storage space and computational complexity Large, not easy to realize in real time by hardware

Images