Monte Carlo Simulations (MCS) are associated with long execution times, which is one of the major issues preventing their use in routine clinical practice. To accelerate MCS, variance reduction techniques (VRT) can be used. Within this context, the Woodcock Tracking method allows improving particle navigation through a CT image. However, this method presents only a very small efficiency gain compared to the standard navigation method. Indeed, particle transport is oversampled even if a small part of the phantom contains a high density region. We address this issue by proposing a new VRT called Super Voxel Woodcock (SVW) which combines both standard and Woodcock tracking navigation methods by introducing the super voxel concept. This method consists in grouping the voxels of the volume in a super voxel grid (pre-processing step) by associating to each of the super voxels a local maximal density which later serves in the interaction distances’ sampling. The proposed methodology was compared to the performance of other methods within the context of a low dose rate brachytherapy application using 109 photon particles. All the methods were implemented within the GGEMS platform running on NVIDIA GTX1050. The SVW efficiency gain as a function of the result uncertainty and the simulation time was 3.7 with respect to the standard navigation and 2.7 with respect to the classical Woodcock tracking. Results show a real improvement of the SVW compared to standard navigation and Woodcock tracking methods without introducing any approximation in the simulations.