Abstract
Quantitative and objective measurements from the medical RI-images have been prevented by the inferiority in picture quality, namely, noise and blur caused by the statistical nature of γ-ray emission of radioisotope and the limited resolution characteristics of the imaging devices. The improvement of the picture quality is desired. The actual desire for the picture quality improvement includes (1) reduction of noise and (2) deblurring of the unsharp boundaries. Generally these two requirements are contradictory in themselves, and linear filtering techniques will not give us satisfactory results.
When we observe the sample variance of the image density value within a small local area, it can be derived that the variance value increases if this local area contains edges of image. The local variance value is, therefore, found to be a good measure for the existence of edge. Utilizing the above mentioned, nonlinear filters, which smoothen the object regions and sharpen the boundaries of the regions, have been constructed. Simulation experiment and application to the actual RI-image, namely, ECG-gated RI-angiocardiogram, are presented to illustrate the satisfactory performance of these filters.
