Infrared small and dim target image background suppression method based on variational Bayes model

Abstract

The invention discloses an infrared small and dim target image background suppression method based on a variational Bayes model. The method comprises the following steps that: according to a sliding window, decomposing an infrared image, and obtaining a plurality of image blocks which have the same size and are mutually overlapped; then, according to a kNN (k-Nearest Neighbor) method, carrying outclustering processing on the obtained image blocks which have the same size and are mutually overlapped, obtaining a plurality of different infrared image block clusters, and according to a variational Bayes theoretical model, inhibiting background ingredients in the obtained infrared image block clusters; and finally, reconstructing the infrared image block clusters subjected to background suppression to obtain an image subjected to the background suppression. By use of the method, an infrared background which has a high gray level as well as rises and falls violently in the infrared image can be effectively suppressed, and target information can be highlighted so as to bring convenience for subsequent target segmentation and detection.

Description

technical field [0001] The invention belongs to the field of digital image processing, and in particular relates to a method for suppressing the background of an infrared weak and small target image based on a variational Bayesian model. Background technique [0002] Because the infrared imaging detection system has the ability of passive detection, high concealment and all-day detection, it is widely used in automatic target recognition equipment. When the target of interest is far away from the infrared imaging detection system, it usually appears as a few pixels in the complex background in the infrared image. At the same time, due to atmospheric attenuation and interference, the contrast and signal-to-noise ratio between the target and the background in the infrared image are low. It brings great difficulties to the precise detection of the target by the infrared imaging detection system. Therefore, how to accurately detect targets from infrared images with low contrast...

AI Technical Summary

Problems solved by technology

For example, the bottom stagnation filtering method based on the time domain contour can suppress background clutter in the time domain and detect moving infrared weak targets, but it is less effective for infrared sequence images with drastic background changes and slow target movement

[0004] The background suppression method of infrared weak and small targets based on the maximum median / mean value filter processes the image in the air domain, which is easily affected by the size of the window. False contours are generated around the target point, and the positioning accuracy is also poor; based on the morphological Top-hat transformation method of weak target background suppression, the suppression effect of this method on the background clutter is very closely related to the selection of structural elements, because in the image Both the target and the noise appear randomly, and it is very difficult to find suitable structural elements without prior knowledge

[0005] Using the adaptive linear prediction method of the two-dimensional least mean square error filter (TDLMS), the least mean square error criterion is used in the prediction process. When the original infrared image contains strongly correlated clutter, the two-dimensional least mean square filter can use the clutter Correlation of the wave, the clutter component is predicted and eliminated from the input signal of the filter, and the target signal component is obtained in the residual of the filter. This method may lead to a decrease in the signal-to-noise ratio for the non-stationary background

[0006] The current infrared image background suppression methods are difficult to effectively suppress complex infrared backgrounds with high gray levels and sharp fluctuations.

Images