Evaluation of monte carlo methods for the analysis of material technology related microscopy images

OData support
Supervisor:
Dr. Csorba Kristóf
Department of Automation and Applied Informatics

The main goal of this thesis is to reconsider and expand the partly existing ReversibleJump Markov Chain Monte Carlo (RJMCMC) half automate optimalization component in cv4sensorhub framework, introduce it’s theoretical background, optimize it’s code and run, implementing debugging (manual modify, querying probabilities) and other supporting functions (graph coloring) with the protection of test driven developement.

This technique first of all is written for the analysis of material technology related microscopy images, but as a reusable component this works with other images as well. If finding contours is not enough to separate objects on an image (for example particales on a marble microscopy image), it is possible to oversegment the image and merging the smaller pieces into bigger ones with Monte Carlo method to represent the particle. The base of these merges can be the similarity of eighbouring areas, the convexity of the combined area (the ratio of it’s area and it’s convex hull), the difference of the avarage colors, ratio of the common orders.

Downloads

Please sign in to download the files of this thesis.