A novel star detection approach based on color filter array detectors that simultaneously measures both the centroid position and spectral properties of the observed stars is proposed.This additional color information can be used to supplement the geometric measurements that form the basis of most modern star-matching routines used in star trackers. We present a candidate model for star detection, based on bivariate Gaussian intensity function with separate color magnitude parameters (i.e., red, green, and blue). The model is fit to the observed point spread function using a nonlinear least squares algorithm. We consider several parameterizations to our model and explore the performance of these alternate schemes using laboratory and night-sky testing. These studies demonstrate that consistent color measurements are possible with well-designed optics.