In this paper, we propose algorithms to jointly optimize the multiple multi-antenna relays which assist the multi-point to multi-point communication in wireless networks. Assuming that perfect channel state information (CSI) is available for all channels, we design the multiple-input-multiple-output (MIMO) relays with three different methods; (1) maximize the minimum of the signal-to-inteference-plus-noise ratios (SINRs) of all destinations satisfying the transmit power constraint of each MIMO relay, (2) minimize the sum of the powers of the relays while fulfilling the SINR requirements for all destinations, and (3) minimize the total interference and noise power received at each destination under the conditions that the desired signal is preserved and the transmit power of each relay is kept below its power budget. It is shown that the former two problems are non-convex but they can be solved accurately and efficiently using the standard semidefinite relaxation and randomization techniques. However, the last problem can be reformulated into a second-order cone programming (SOCP) problem. Computer simulations verify the improved performance of the jointly optimized MIMO relays when compared to zero-forcing (ZF) relays.