NOISY TEXTURES DETECTION USING NEURAL NETWORK CLASSIFICATION

Abstract

The possibility of detecting texture areas of the image distorted by an additive uncorrelated noise has been analyzed with help of the classifier based on neural network learning. We investigated the following sets of input parameters (attributes): relative local variance, Harris detector, simple textures detector, detector based on discrete cosine transform. It is shown that detection performance depends on the number of input characters and on type of used input parameters. Issues of training the neural network and the application of a trained classifier for image processing with different textures are discussed.

Keywords

• текстура, виявлення, зображення, перешкоди, нейросеть, навчання.