In large scale sensor networks the cross-correlations between nodes are rarely known. Failing to account for this detail in the design of distributed estimation schemes may lead to statistical inconsistencies and may even cause some estimators to diverge at various nodes within the network. Among the approaches that have been proposed to address this problem, covariance intersection is perhaps the most well-known. Other techniques rely on Chernoff fusion, of which covariance intersection is a special case. In a recent work we have introduced a technique, particles intersection, which is the application of Chernoff fusion using the empirical probability densities of two or more particle filters. This approach, however, may in some cases suffer from an excessive computational overhead. This issue is alleviated in this work by turning to a different yet related information fusion technique, a convex combination of probability densities. The utility of the new approach is demonstrated through simulations of cooperative robots localization, where it is compared with particles intersection and another state-of-the-art technique for distributed particle filtering. The results show a decisive advantage for the new technique both in terms of estimation accuracy and computational load. In addition, we provide an account of known techniques for tuning of the fusion parameters, and also suggest a new one based on the optimization of Chernoff information.