Öz-uzay daǧilihm enerjili markov rasgele alanlariyla nesne yeri saptama

Aim of the work is detecting and segmenting objects (such as license plate or human eyes) from images with Markov Random Fields (MRFs). MRF is a region based segmentation method that works iteratively in Bayes framework and uses simulated annealing as optimization method. We focused to model the observation energy, that should be minimized, by Gaussian Distributions. Although, simple segmentation problems (such as thesholding) can be modeled by uni-variate Gaussian distributions, complex textures regions (such as license plate or human eye) need multi-dimensional distributions. We suggested to use Eigen-Density distributions to model observation energies for complex textured regions.