Detection of edges in analytical images using wavelet maxima

Two-dimensional element distributions generated by surface analysis methods like scanning Secondary Ion Mass Spectrometry (SIMS) are characterized by Poisson statistics of small integer values, especially when the concentrati ons of the measured element are very low. This usually leads to rather nois y and blurred images containing objects which do not usually have sharp edg es or may have noise induced boundaries. As a result, traditional edge dete ction techniques become difficult and yield rather poor results. This paper reports the application of a novel edge detection based on the wavelet-tra nsform for images of chemical content. The algorithm is able to detect edge s of images even at very low signal-to-noise ratios (SNR) while for the mos t part preserving the shape of the region boundaries and not blurring them. The methodology is discussed and experimental results using both simulated and real images are presented. Copyright (C) 2000 by Marcel Dekker, Inc.