ECAS-II: A Hybrid Algorithm for the Construction of Multidimensional Image Segmenters

Abstract

In this paper, we present a hybrid algorithm that combines evolutionary methods with cellular automata for the segmentation of multidimensional images, with a focus on hyperspectral images. The algorithm enables the automatic generation of cellular automata transition rules by utilizing a training set composed of specifically designed synthetic RGB images. This approach simplifies the process significantly, as it circumvents the need for adequately labeled hyperspectral images. Furthermore, by adjusting the parameters of the synthetic RGB images used in training, the algorithm can produce a variety of high-dimensional segmentations. The proposed method has been evaluated on both synthetic and real hyperspectral images, demonstrating highly competitive segmentation results compared to other techniques reported in the literature.