Variance reduction techniques are necessary for the Monte Carlo (MC) calculations in which obtaining a detailed flux distribution for a large and complex model is required. A new method for generating mesh weight window (WW) parameters is presented, named global weight window generator (GWWG). In the method, MC calculation is performed to generate the importance of each mesh voxel in the user-set region, and then the corresponding WW parameters are obtained. The importance is related to whether the simulated particles can be uniformly transported to each mesh voxel, which is called particles density uniformity. The GWWG method does not rely on user's design experience and can significantly improve the calculation efficiency. The tests for this method have been conducted on ITER reference neutronics models. In the calculation of ITER C-Lite model, the FoM was improved comparing to the analog FoM by the factor of 637.4.