We present a comparative experimental phantom study of fiber-based and noncontact fluorescence tomography with respect to quantitation and localization of reconstructed fluorescent inclusions in turbid media such as tissue. Noncontact acquisition is usually considered potentially superior to fiber-based techniques because of the availability of a large number of detector readouts through a CCD. Our results indicate, however, that noncontact acquisition itself might improve the quality of reconstructions significantly, even without increasing the number of detectors and thus keeping the inverse problem moderately complex.