ABSTRACT
Feature extraction is an essential pre-processing step to pattern recognition and machine learning problems. It transforms an image of containing redundant data into a reduced representation set of features. Many morphological approaches for feature extraction or image segmentation have been proposed. A morphological tool for image segmentation uses the watershed transformation [Vincent and Soille 1991; Gauch 1999]. Applying watershed transformation on the gradient image generates an initial segmentation, but resulting in over-segmentation, so that region merging needs to be performed as the post-processing. A morphological clustering method is illustrated in Yun, Park, and Lee [1998]. A morphology based supervised segmentation is introduced in Gu and Lee [1998]. They created the searching area where the real object resides by applying the morphological operators to the initial boundary. Video segmentation using mathematical morphology can be referred to Wang [1998].
