A Graph-Based Approach to the Segmentation of Images with Mould Filled Foam Matrices

In the paper a method of MST (minimum spanning tree) graph-based image segmentation supported by parallel GPUprocessing has been presented. It has been applied to images of foam matrices filled with mould fungi of Mucor genus. The aim of the segmentation was to extract the profiles of the matrix objects from image background. Parallel computing methods were used to the initial image filtering, graph edge array building and sorting. The results obtained with graph based method both with and without parallel processing have been compared with conventional analysis in MATLAB environment. It has been verified that GPU processing clearly reduce execution time of the algorithm.