| Abstract [eng] |
Data collection in the form of digital images is inseparable in many fields of our day life. Images are gathered and archived in almost every field. Because of the improvement of the technical equipment, images, image retrieval, collection, and analysis have become an integral part of biomedical, natural and social sciences, and engineering. Image analysis is already becoming an important new tool and information retrieval for image extraction equipment. This area is facing new challenges and opportunities. The goal of the research is to find new statistical decisions for spatial information in digital images. Spatial information defines the physical meaning of objects and the relationship between objects. Much attention is paid to the influence of spatial correlation in the Bayes discriminant functions by introducing spatial dependence in prior probability evaluation. Neighborhood schemes and their influence to classification methods are analyzed. Methods of analyzing eye fundus images are investigated in this work. Methods are used for measuring the width of eye fundus blood vessels and for vessel classification to the arteries and veins. Blood vessel profile extraction method, based on spatial function, is used for vessel width measurements. The measurement method is adaptive to different image sizes. New proposed features are used in blood vessels classification. Spatial feature normalization is applied in order to reduce the effect of uneven lighting of images. |
