Adaptive clustering for segmentation of multi-sensor images

Abstract

Automatic detection or recognition of specific objects from a sequence of images acquired under varying conditions and from different modalities requires careful segmentation. We present here the use of an adaptive neuro-fuzzy clustering technique for image segmentation and boundary detection for better extraction of the dominant features in images of the same scene acquired by synthetic aperture radar (SAR) and electro-optic (EO) sensors. The advantage of such adaptive segmentation is clearer identification of objects that appear differently due to different sensor characteristics.