Markov random field image segmentation using cellular neural network

摘要: Markovian approaches to early vision processes need a huge amount of computing power. These algorithms can usually be implemented on parallel structures. With the Cellular Neural Networks (CNN), new image processing tool is coming into consideration. Its VLSI implementation takes place single analog chip containing several thousands cells. Herein we use CNN UM architecture for statistical segmentation. The Modified Metropolis Dynamics (MMD) method raw CNN. We are able implement (pseudo) random field generator using one layer (one memory/cell) introduce whole pseudostochastic segmentation process in 8 memories/cell. simple arithmetic functions (addition, multiplication), equality-test between neighboring pixels and very nonlinear output (step, jigsaw). this architecture, real execute relaxation algorithm about 100 iterations 1 ms. In proposed solution unsupervised. have developed pixel-level estimation model. turns original smooth one. Then two gray-level values every pixel: smoothed used estimating probability distribution region label at given pixel. Using conventional first-order Markov Random Field (MRF) model, some misclassification errors remained boundaries, because difficulties case low SNR. By greater neighborhood, problem has been avoided. our experiments, simulation system with fixed-point integer precision 16 bits. Our results show that even constrained conditions value-representations (the interval (-64,+64), accuracy 0.002) result an effective acceptable