A coherent optical data-processing system for identifying and counting blood cells has been developed. Optical Weiner-Kolmogorov (minimum mean-square error) linear filtering is employed to estimate the density per unit area of a given cell type present in a sample blood smear. The Weiner filter is designed to recognize the characteristics of an ensemble of cells of a type rather than specific individual cells. In this way the system is made insensitive to minor variations in shape, size, or orientation of the cells. A Poisson shot-noise model for blood cells as a spatial random process is employed to calculate the far-field scattering from the cells. Results of experiments in which the percentage of reticulated (immature) red cells present in a smear has been determined are presented. The identification and counting accuracies compare favorably with those obtained by laboratory technicians with a microscope. The potential time saving of instantaneous estimation for the entire smear is significant.