Approche Evolutionaire Base Automate Cellulaire Pour La Segmentation Des Images Medicales

Résumé: Segmentation isa crucial step for different medical image applications and computer-aided diagnosis (CAD). Delimitation of the various structures present in images is needed as a pre-processing stage for advanced medical images applications. In conventional edge detectors, the basic concept is to calculate the first or thesecond-order of derivative of image intensities, which is limited when the images are textured. The objective of this thesis is to develop a new approach for accurate, robust and computationally efficient brain MRI images edge detection. In this thesis, we propose a novel procedure to extract the textures from brain MR images, which use the quantum information aspect for the local binary pattern (LBP) descriptor. Then, we investigate this model for three applications: a salt and pepper noise reduction for the Canny-deriche algorithm, as second application a cellular automaton (CA) based algorithm in combination with the new texture descriptor for edge detection task and finally, the model combined with Penguin search optimization algorithm (PeSOA) for brain tumor edge detection. The different experimental results demonstrate that each of these proposed approaches produces good results, in comparison to the appropriate methods devoted to each application. Keywords: Image segmentation, Edge detection, Local Classifier, Quantum Information, Magnetic Resonance Imaging (MRI), Noise Reduction, Cellular Automaton (CA), evolutionary algorithm

Publié dans la revue: