Regresar

A method to classify digital images by means of statistics of a wavelet decomposition

Abstract:

There is a wide variety of methods for the analysis of textures and the extraction of image characteristics based on wavelet decomposition. The objective of this paper is to find a vector of characteristics of each image and to determine a classification among them using principal component analysis. The procedures presented here are for classifying test images. After fragmenting the image, a wavelet decomposition of the fragmented images is performed both in scale and in orientation. To characterize the images, statistics of marginal and joint distributions based on local neighborhoods of spatial, orientation and scale type are used.