Segmentation of blood vessels by region growing

Programskim alatom MATLAB konstruirano je programsko rješenje za segmentaciju žila sa slike očne pozadine. Segmentacija se provodi postupkom izrastanja regije počevši od izvora žile prema krajevima. Za pojačavanje žila unutar slike koristi se tenzorsko glasovanje kojim se pojačavaju segmenti u kojima se nalaze žile. Izrastanje regije bazira se na višerazinskoj metodi filtracije. Nakon filtracije računaju se svojstvene vrijednosti i svojstveni vektori pojedinih piksela te se određuje pripadnost pojedinih piksela žilama. Za testiranje su korištene umjetno generirane žile zbog nedostatka ručne segmentacije od strane stručnog liječnika Najveći problem ove metode je u detekciji grananja i križanja žila.With MATLAB software tool a software solution for the segmentation of vessels from retina images was built. The segmentation is performed with region growing method from the source of the vessel towards ends. To increase the vessels within the images a tensor voting was used to amplify segments containing vessels. Region growing method is based on multiscale filtration method. From a filtration eigenvalues and eigenvectors were calculated for each pixel and the segmentation of vessels was made. For testing purposes an artificially generated vessels were created because of the lack of manual segmentation made by an expert doctor. Main problem of this method is in detection of the vessel branching and crossing