Nonlinear Image Processing Methods

Image processing methods are often applied to nuclear medicine images to improve their smoothness or to sharpen detail boundaries. These processing methods are either linear or nonlinear. Nonlinear mnethods are those in which individual data points are compared with some standard or fit to determine their suitability for modification or acceptance. Nonlinear methods improve images by modification or removal of improbable values by rejecting data from smoothing areas because of size disagreement or by rejecting spatial frequency components that are too small. Problems in spatial frequency domain methods are discussed, and two-dimensional leastsquares bounding is compared with linear methods and with a two-dimensional median smoothing method.