When using vwrays with rcontrib to render a image, the pixels sample, just like the option -ps in rpict, can be implemented to reduce the computing time? (This idea may be a little bittle crazy 
!)
The image-plane sampling in rpict is fairly complicated, and you wouldn’t want to use it in a pure Monte Carlo sampling context such as rcontrib. You really need all those samples to reduce the variance in your final result, which will likely have to average each pixel of quite a few samples using dctimestep and/or pfilt.