The distortions inherent in PET images of the human heart due to finite image resolution and cardiac motion limit the capability to evaluate physiology quantitatively. A method based on a simple geometrical model of region of interest representations in physical space has been developed to minimize these distortions. In this paper, simulation studies have been performed to evaluate the noise characteristics of the method. This study demonstrates that unbiased estimates of kinetic model parameters which describe myocardial physiology can be measured with an accuracy of 7%-15% for scale-related parameters and 4%-16% for shape-related parameters of kinetic models in studies with the equivalent of 1 million events. Application of the techniques developed in this paper for the measurement of myocardial blood flow in eight dogs (14 independent flow states) shows a strong correlation with microsphere determined blood flow in the same animals (slope = 1.022, intercept = -0.18, r = 0.96).