Iterative satellite image segmentation by fuzzy hit-or-miss and homogeneity index

Object-based segmentation is the first essential step for image processing applications. Recently, satellite image segmentation techniques have been developed, but not enough to preserve the significant information contained in the small regions of an image. The proposed method is to partition the image into homogeneous regions by using a fuzzy hit-or-miss operator with an inherent spatial transformation, which enables the preservation of the small regions. In the algorithm proposed here, an iterative segmentation technique is formulated as consequential processes. Then, each time in iterating, hypothesis testing is used to evaluate the quality of the segmented regions with a homogeneity index. The segmentation algorithm is unsupervised and employs few parameters, most of which can be calculated from the input data. This comparative study indicates that the new iterative segmentation algorithm provides acceptable results as seen in the tested examples of synthetics and satellite images.