COMPARISON OF WAVELET FILTERING WITH ESTABLISHED TECHNIQUES FOR DE-NOISING OF EPMA IMAGES

As a result of the small number of detected photons in each pixel of t he image, images from Electron Probe Micro Analysis (EPMA) are charact erised by Poisson statistic of small integer values. Hence follow low signal-to-noise ratios requiring the application of de-noising techniq ues both to enhance the visual impression of the picture and to suppor t further image processing methods. The wavelet transform, a new and v ery versatile technique developed during the last decade, has found a lot of different applications such as image compression, human vision, earthquake prediction, signal deconvolution or singularity detection and, as recently described, de-noising. In this paper the wavelet de-n oising algorithm proposed by Donoho and Johnstone is compared to sever al well-known digital smoothing operators. The overall ranking shows t hat besides its numerous successful applications in various other area s the wavelet transform can also be applied for de-noising purposes in analytical imaging yielding comparable results to the best filters ye t in use in chemometrics.