In this paper we present a very fast pipelined parallel ray casting algorithm for volume rendering. Our algorithm is based on an extended space leaping method which minimizes the traversal of data and image space by using run-length encoding and line drawing algorithms. We propose a more advanced space leaping method which allows the efficient implementation of parallel forward projection by merging the run-lengths for the line drawing. We shall show that the whole algorithm is sharply speed up by reducing the time taken to project the run-lengths onto the image screen, and by exploiting the pipelined parallelism in our space leaping method. Also, we shall show the experimental result of the parallel ray casting algorithm implemented on our Computational Grid portal environment.