The determination of the galaxy luminosity function is an active and fundamental field in observational cosmology. In this paper we propose a cost effective way of measuring galaxy luminosity functions at faint magnitudes. Our technique employs the use of galaxy redshifts estimated from their multicolor photometry (Connolly et al. 1995). Associated with the redshift estimate is a well defined error distribution. We have derived a variant of Lynden-Bell's (1971) C--method that considers, for each galaxy, the probability distribution in absolute magnitude resultant from the redshift error. This technique is tested through simulations and potential biases are quantified. We then apply the technique to a sample of galaxies with multicolor photometric data at moderately faint (B=23) limits, and compare the results to a subset of these data with spectroscopic redshifts. We find that the luminosity function derived from the photometric redshifts is consistent with that determined from spectroscopic redshifts.