Infrared Ship Target Segmentation Based on Active Contour Model

Abstract

Infrared ship image has low contrast, weak boundary and uneven gray-scale distribution, leading to difficult ship target segmentation. Chen-Vese model, a classic active contour model, does not rely on image boundary information, which can better segment images with weak or discontinuous edges. Moreover, it has certain noise resistance, but correct segmentation result is impossible for infrared images with uneven grayscales. In view of this, this paper first uses top-hat transform or bottom-hat transform to preprocess the infrared image, increase the image contrast, so that the gray value tends to be uniform in the target and the background. Then, the improved Chen-Vese model is used to test the ship target. Experimental results show that the new method can quickly and effectively detect infrared ship targets, which is superior to the Chen-Vese model in terms of curve evolution speed and noise resistance.