Based on Poisson variability, the strength of cross-correlation of two spike trains was quantified by a cross-correlation index (CCI). Discriminant analysis with CCI was used to quantify the contribution of cross-correlation in stimulus feature discrimination in an ensemble of retinal ganglion cells. Compared with simulated Poisson cell pair, it was found that cross-correlation carried useful information for stimulus feature discrimination and with the cross-correlation included in the analysis, the color and the intensity discrimination ability of a cell pair increase 4.2% and 2.7%, respectively.