Dynamic local thresholding technique for sea ice classification

Many methods of sea ice classification are based upon global thresholding techniques. In thresholding, points are selected according to the histogram of an image and are used to achieve a representation of that image in a limited number of gray levels, or classifications. In this paper we present our work, in which we developed and implemented a local thresholding technique adapted for sea ice classification. Our method offers both accuracy and flexibility to standard global thresholding methods. Because the local threshold is determined by examining small areas of the image, local gray-level changes which may not be distinguished by a global thresholding technique can be detected through local inspection. This leads to the selection of a more appropriate threshold level for each area of the image. In addition, the procedure of threshold selection used requires no user interaction whatsoever.