The general answer is yes. See, for instance, Pearson et al. Curr Biol. 2008 18(13):982-6; Sherwood and Pearson PLoS One. 2010 5(12):e15217; and Byrne et al. Psychol Rev. 2007 114(2):340-75.

Synaptic strength (if I understand your question correctly), especially in relation to deliberate practice...that is more difficult to figure out. I'm not aware of any particular research on that topic (and it would be hellishly difficult to do). Whether (and if so, how much) visualization-gained improvements are transferable to real-world skills is also controversial.

I'll indulge in one paragraph of guesswork here. Extrapyramidal centers (such as spine, basal ganglia, cerebellum) appear excluded during visualization exercises. Say you are visualizing a martial arts kata, and say that visualization does produce potentiation. Even in this case, all of the changes would be limited to the premotor area and the primary motor cortex - areas that are critical for actual movement execution (especially cerebellum, balance centers in the brainstem and spinal centers) would be unaffected. Worse, the changes in the high-level centers would be made without corrective input. When kata is then attempted in real life, these idealized neural plans might slam nose-first into unexpected feedback responses - therefore making things worse, not better. For this reason, I would personally eschew visualization as a training modality in any actual physical skill.