Infrared Image Segmentation Combining Mutual Information and Genetic Algorithm

Infrared image segmentation is a very important preparatory work in the target recognition and tracking. Because infrared image has the features of low contrast, high noise and blurring, the segmentation is difficult to carry out. A novel image segmentation method, which utilizes the technique of mutual information and genetic algorithm, is proposed. The initial threshold is chosen by using Otsu algorithm firstly. Then considering the internal relationship between the original and segmented images, mutual information is used to determine the optimal threshold, which applies the method of image registration to image segmentation. The genetic algorithm is introduced into image segmentation to optimize the procedure of seeking threshold using the characteristic of quick seeking capacity. Experimental results show that the presented method can quickly obtain effective target region, which facilitates the target recognition and tracking in the next step.