This PhD Thesis deals with the segmentation of hyperspectral images from the point of view of Lattice Computing. We have introduced the application of Associative Morphological Memories as a tool to detect strong lattice independence, which has been proven equivalent to affine independence. Therefore, sets of strong lattice independent vectors found using our algorithms correspond to the vertices of convex sets that cover most of the data. Unmixing the data relative to these endmembers provides a collection of abundance images which can be assumed either as unsupervised segmentations of the images or as features extracted from the hyperspectral image pixels. Besides, we have applied this feature extraction to propose a content based image retrieval approach based on the image spectral characterization provided by the endmembers. Finally, we extended our ideas to the proposal of Morphological Cellular Automata whose dynamics are guided by the morphological/lattice independence properties of the image pixels. Our works have also explored the applicability of Evolution Strategies to the endmember induction from the hyperspectral image data.