Processing Of Infrared Images With Low SNR Small Target

The paper studies the processing of infrared small target image under complex back grounds with low SNR, which includes infrared small target detection, segmentation and pretreatment.Mean shift is a very good Statistical clustering algorithm with many advantages in the computer vision area as image segmentation and visual target tracking, which is developed in recent years, but few scholars study it in the infrared small target detection area. The paper studies the algorithm and found its advantages in the processing of the infrared small target image under complex backgrounds. The main achievements of this paper are as follows:The third chapter studies the pretreatment for detecting infrared small target. The scene model of infrared small target image is established, the characteristics of image pretreatment are analyzed under this condition, and a pretreatment method based on mean shift filtering and high-pass filtering is presented. The method trades-off the result of noise-filtering, background-suppression, and image-enhancement, combines the different pretreatment methods to preprocess the infrared small target image. The experimental results indicate that the method can effectively preprocess the infrared small target image under complex backgrounds.In the fourth chapter, a new infrared small target detection method based on energy accumulation and mean shift clustering is presented by studying the detection of weak and small target in infrared image. Firstly, accumulating the energy of the infrared image sequence in the sliding window with a setting size can remove the random noise in the infrared images to increase the SNR. Then suppress the background by using the difference of adaptive mean shift filtering results with two different bandwidths. Target image and background image can be got by use of different bandwidth clustering of mean shift algorithm. the noise can be excluded at the same time. And then segmenting the regions and getting the true target.The fifth chapter studies the segmentation method of infrared flying target image, which is the most important problem is the optimization of mean shift-based image segmentation method for real time infrared image processing. Firstly, the initial searching points and bandwidth matrix, which are used for clustering, are optimized by the priori knowledge on infrared image data. Then the local modes are searched by the mean shift cluster algorithm. After the adaptive filtering, the final result of flying infrared object segmentation can be got through a global optimization criterion for mode merging.At last, we have summarized the main work in the paper.