The optimal frequency characteristic of Butterworth-Wiener filters (BWF) for improving the image quality of 201Tl myocardial SPECT was determined by a phantom experiment. Thirty two projection images of the phantom containing 11.1 MBq of 201Tl with 4 different cold lesions were collected during a 180 degree arc of a gamma camera. A set of the projection images were processed with each of 27 different BWFs, and SPECT images were reconstructed by Shepp-Logan filtered backprojection. The SPECT images were evaluated for their diagnostic ability to visibly detect the cold lesions by four nuclear medicine physicians. The lesion contrasts were used as an adjunctive tool to determine the optimum filter. The optimal combination of the parameters determining BWF characteristics (for the data of about 100 count/pixel at the myocardium) is: 1) cutoff of 0.25/pixel, 2) FWHM of 3 pixels, 3) noise/signal ratio of 0.02. FWHM and noise/signal ratio affected lesion contrast much less than cutoff frequency. Clinical myocardial SPECT images processed with the optimal BWF showed less noise and sharper delineation of the myocardium.