Modern communication systems are often interferencelimited. By modeling the co-channel interference asspatially colored, temporally white Gaussian noise, itis straightforward to incorporate interference rejectionin the metric of a sequence estimator. In general, estimates of both the channels and the spatial color of theco-channel interference and the noise are needed. In thiswork, a structured model for the spatial noise covariancematrix is proposed and maximum likelihood estimates ofthe parameters are derived. The choice of model orderis also addressed. Simulation results show large gains due to the use of these structured estimates compared with the conventional, unstructured, approach.