Medical Image Recognition Based on Fractal Features

The article presents a new method of feature extraction from an image represented as a digital bit-mapped display in grey scale. The method is based on an idea of fractal model of nature and processes carried out in it. The results of three recognition experiments of different kinds of medical images were included. Fractal and traditional features were evaluated, by the Fisher coefficient, in each experiment. Extracted features were recognised by the use of: sigmoidal neural network, neuro-fuzzy system, k-nearest neighbour method and the nearest mode algorithm. The authors presented the discussion of obtained results, tips concerning fractal features extracting, the comparison of effectiveness and the properties of recognition the method used.