An integrated method for detection and discrimination of a signal pulse in two-dimensional cluttered background noise is presented in this paper. The method is used to estimate the magnitude (detection) and the two-dimensional position (centroid) of the signal using a statistical model for the correlated background noise. The optimal discriminant can use measurements from multiple spectral bands and prior information about the signal shape as well as derived discriminants such as the velocity of the signal pulse. Preprocessing techniques include background suppression in multiple spectral bands with a generalization of temporal and spatial filtering. The coefficients of the optimal filter used for detection and discrimination depend on the statistical characteristics of the signal and the background (averages and cross correlations). Illustrative two-dimensional scenes and representative numerical results are presented to show how this integrated approach (which uses all discriminants) compares with a piece-meal approach to discrimination.