Noise Reduction of Medical Volume Data Using Level Sets

摘要

A new method is developed here to reduce the impulse noise in the preprocessing of medical volume rendering. First the medical volume data is decomposed into upper level sets and lower level sets, and then each connected component of each level set is given a measure of volumes. In general, impulse noise is composed of connected components with extremely small volume. A threshold is selected and the connected components with volume less than the threshold are removed. The medical volume data is reconstructed from processed level sets and the noise is reduced. The method is characterized by shape invariance and contrast invariance. Finally, the performance of our method is demonstrated with an experiment.