Functional images used in positron emission tomography (PET) have the advantage of presenting simultaneously the anatomical and functional information in cross-sectional body slices. However, the nonlinearity in parameter estimation, when combined with the finite image resolution, can cause systematic errors or biases in the estimated functional parameters. The effect of this error on blood flow images, which are commonly used in PET, is investigated in this study. Computer-simulated brain and heart phantoms of realistic configurations are used to examine the effect of various factors, such as imaging resolution, estimation nonlinearity, and structure configuration. The nonlinearity characteristics of six commonly used blood flow estimation techniques are simulated. Results show that structure boundaries on parametric images between tissues of different blood flows do not usually coincide with the true anatomical boundaries and would thus cause an apparent change in the cross-sectional size of the structures. The regional blood flow values as obtained from the blood flow images are usually lower than the true values. The severity of these effects is dependent on the characteristics of the flow estimation technique, the image resolution, and the size and shape of the structure. Although image resolution is a major factor in causing errors in the parametric images, its improvement, within the range examined in the present study [from 1.5 to 0.5 cm full width at half maximum (FWHM)], is not found to reduce drastically the underestimation of blood flow in brain phantom. The effect on boundary shift, however, is found to be in proportion to the FWHM of image resolution. Implications of these effects on generation, interpretation, and comparison of parametric/functional images are discussed.