VarMet: a novel method for detection of image singularities
Thor Ole Gulsrud, Kjersti Engan, Jostein Herredsvela · 2006
Abstract: The aim of this study is to detect singularities in digital images based on a novel method for feature extraction; VarMet. The multiscale median transform (MMT) is known as a tool for extracting singularities in an image. In this work an extension of the MMT is proposed. The VarMet scheme relies on computing statistical features based on the coefficients obtained from the MMT of the input image. Some experiments are presented from which it can be concluded that the VarMet scheme provides very good enhancement of singularities in the test images. Key–Words: image singularities, feature extraction, object recognition, multiscale median transform (MMT), VarMet, microcalcifications 1