Extracción de características en mamografía digitalizada utilizando filtrado quincunx

Abstract

In this paper, an Adaptive Wavelet Transform is applied on a Quincunx grid (Red-Black Wavelet Transform) with the image denoising purpose. The Quincunx filtering is applied on digitalized mammography images in order to detect microcalcifications. From the images, features like the gradient magnitude and features of local contrast and normalized local contrast with differents windows sizes, are extracted. These features model the microcalcifications and health tissue. The model applies a sequential selection method based on a General Regression Neural Network (GRNN) to obtain the best features. Images segmentation is carried out by means of unsupervised k-means algorithm; obtain images with information about microcalcifications and mamma tissue. The results are compare applied the classic Wavelet Orthogonal filtering. © Springer-Verlag Berlin Heidelberg 2007.