Infrared image segmentation through combined marker based watershed

To segment the infrared dim target efficiently, a new marker based watershed method through combination of the property of the infrared image, the kernel method, the mathematical morphological operations and the distance transformation is proposed. Firstly, the dim target is enhanced through the property of the infrared image, the kernel method and the linear extension. Secondly, the markers of the targets are obtained by applying the image binarisation and the morphological operations. Thirdly, the background of the image is separated according to the targets through the distance transformation and watershed transformation. Fourthly, the logical OR operation is applied on the markers and the separated background image to form the final marker image. After the gradient calculation of the original image, the final watershed segmentation is performed on the gradient image guided by the final marker image. Because of the appropriately combining of the property of the infrared image, the kernel method, the mathematical morphological operations and the distance transformation, the dim targets are well enhanced and the markers are efficiently extracted. Therefore, the proposed algorithm can efficiently segment the infrared dim targets. Experimental results verified the effective performance of the proposed method.