Application of image processing techniques to gamma-angiography.

Different image processing techniques have been tested and compared on data derived from gamma-angiography images to detect the boundary of the left ventricle. The method involves a preprocessing step, followed by the edge detection itself. The best preprocessing is a nonlinear "variant" filtering, where each pixel is replaced by the average of the 3 X 3 neighborhood having the smallest variance. The edge detector giving the best contour is a Sobel operator. A second-order high-pass Butterworth filter also provides a good segmentation.