The nuclear medicine physician uses visual interpretation of a movie-like display of the beating human heart to detect wall motion abnormalities which might be related to impaired cardiac function. The present work is directed toward extracting more information from the heart motion study, and presenting it in a useful manner. A spatially adaptive smoothing routine using a quadtree image representation gives an improvement in mean squared error compared to the "S9" smoother commonly used for nuclear medicine studies. Functional images show the two-dimensional distribution of parameters of the heart motion. The most popular, the first harmonic phase functional image, formed from the first Fourier harmonic fit to each pixel time-activity curve, is subject to significant artifacts which make a simple interpretation of it difficult. A multi-harmonic approximation is more accurate and offers a wealth of unique parameters with which to construct more directly meaningful functional images.