Determining the similarity of raster images with 3D nested convex hulls

V diplomskem delu smo izdelali metodo, ki s pomočjo 3D vgnezdenih izbočenih lupin ugotavlja podobnost med slikami. Najprej smo podali opis štirih že znanih metod za ugotavljanje podobnosti slik. Implementirana metoda prejme sivinsko rastrsko sliko, ki jo redči s Sobelovo zaznavo robov in nato sestavi 3D vgnezdene izbočene lupine te slike. Na podlagi teh izračuna cenilko, katere vrednost uporabimo za ugotavljanje podobnosti slik. Metodo smo testirali nad štirimi slikami in ovrednotili ustreznost rezultatov.In the thesis, a method that determines the similarity between images with the help of nested 3D convex hulls was developed. Four already known methods for determining the similarity of images were described. The implemented method receives a grayscale raster image, thins it with the Sobel edge detection, and then constructs 3D nested convex hulls on the obtained points. Based on these hulls, the method calculates an estimation function, the value of which is used to determine the similarity of the images. The method was tested on four images and the relevance of the results was discussed at the end