Study On The Remote Sensing Image Segmentation Algorithms Based On Mathematical Morphology

In the process of image processing, people often interested in some parts of the image, in the same time, different targets in the images have different characteristics. In order to identify and analyze the target areas of the image, it is necessary to separate them from the image. Remote sensing image is usually contains high resolution, rich information, complex target structure, but also includes a lot of noise. Due to these characteristics of remote sensing image, making the traditional mathematical morphology image segmentation algorithm is not entirely applied, therefore, to some extent, which hinders the promotion and application of the technology in the field of remote sensing image segmentation.According to the characteristics of remote sensing image and mathematical morphology in image segmentation, this paper puts forward a improved algorithm of watershed segmentation based on mathematical morphology. By experimental verification, the algorithm is efficient、accurate、fast and so on, which can be extract the target areas of continuous and closed. The traditional watershed segmentation algorithm is more sensitive to noise, which will produces the over-segmentation phenomena if this algorithm directly applies to the remote sensing images, the effect of image segmentation is not ideal. In this paper, the main work includes the following aspects.First, this paper introduces basic operation which contains expansion and corrosion of the mathematical morphology, it is based on binary morphology and extends to the grayscale morphology. Through the analysis and comparison of several kinds of gradient operator, the multi-scale morphological gradient operator is proposed in this paper can get ideal gradient image.Secondly, the selection of threshold is usually artificial set when using the extended least transform to tag on the gradient image, which is often with certain blindness. This paper uses the two-dimensional Otsu adaptive threshold segmentation algorithm to obtain the best threshold, it is effective to avoid the artificial intervention. Experiments show that the proposed algorithm can effectively restrains the interference of noise, marks the main outline of the image and keeps the complete information.Finally, the article puts forward mark watershed segmentation algorithm to overcome the deficiency of traditional watershed algorithm segmentation. The input of the original image is filtered by morphology method, extracting markers on the gradient image, and the tag as a compulsory minimum modifies gradient image. According to the results of the experiment, this method can effectively solves the problem of the over-segmentation of watershed algorithm and achieves the good segmentation effect.