This work addresses distributed processing methods to precoder selection in wireless systems. To explore the spatial correlation of the medium, an additional step of nonuniform power distribution over data streams is proposed. The min-sum algorithm in factor graphs is then applied to reach a near-optimal solution. The system capacity is optimized in a distributed manner assuming that it is simply the sum of individual sum rates. Also, spatially-correlated MIMO channel matrices drawn from measured data are considered. Evaluations on the potential of weighted precoder are provided and its performance is compared to the case with uniform power distribution. Simulation results are presented and discussed. As expected, the min-sum algorithm obtains gain in system capacity over a baseline greedy approach.