M. Haimerl, J. Moldenhauer, T. Beth, and U. Mende (Germany)
Medical Image Processing, Locally Adaptive Filtering, Image Enhancement, Ultrasound Images
In this article we introduce locally adaptive stochastic pro cesses as an alternative for nonlinear diffusion filtering in the context of ultrasound images. We realise these pro cesses by stochastic matrices which are defined by renor malised weights based on reliability measures or feature distances in local neighbourhoods. We analyse their prop erties and compare them with the requirementsfor diffusion scale spaces, e.g. stability and causality. We propose vari ants of these stochastic processes based on the optimisation of local diversity measures. We discuss the generation of appropriate and robust features and we show that specific realisations of the proposed stochastic processes represent approximationsfor other nonlinear filters like median filters or morphological operators.
Important Links:
Go Back
