ON THE ACCURACY OF SELECTED IMAGE TEXTURE SEGMENTATION METHODS

Abstract

Texture segmentation methods are widely used in many image analysis tasks. However, accuracy of these methods is seldom discussed in the literature. This paper evaluates two texture segmentation techniques: based on multilayer perceptron and network of synchronized oscillators. Their accuracy of estimation of artificial image object parameters is assessed. The evaluation is performed for optical images, acquired using scanner, digital and TV camera. Evaluation results are presented along with discussion and conclusions.