Gaussian-Weighted Moving-Window Robust Automatic Threshold Selection
A multi-scale, moving-window method for local thresholding based on Robust Automatic Threshold Selection (RATS) is developed. Using a model for the noise response of the optimal edge detector in this context, the reliability of thresholds computed at different scales is determined. The threshold computed at the smallest scale at which the reliability is sufficient is used. The performance on 2-D images is evaluated on synthetic an natural images in the presence of varying background and noise. Results show the method deals better with these problems than earlier versions of RATS at most noise levels.
KeywordsVision Graph Global Threshold Global Thresholding Edge Strength Faint Object
Unable to display preview. Download preview PDF.
- 4.Trier, D., Jain, A.K.: Goal-directed evalution of binarization methods. IEEE Trans. Image Proc. 17(12), 1191–1201 (1995)Google Scholar
- 5.Wilkinson, M.H.F.: Rapid automatic segmentation of fluorescent and phasecontrast images of bacteria. In: Slavik, J. (ed.) Fluorescence Microscopy and Fluorescent Probes, pp. 261–266. Plenum Press, New York (1996)Google Scholar
- 6.Wilkinson, M.H.F.: Automated and manual segmentation techniques in image analysis of microbes. In: Wilkinson, M.H.F., Schut, F. (eds.) Digital Image Analysis of Microbes, pp. 135–171. John Wiley and Sons, Ltd, Chichester (1998)Google Scholar