A shear-warp algorithm with run-length-encoded volume, one of the fastest CPU-based speed acceleration techniques developed so far for direct volume rendering, has various defects such as increase in memory consumption as well as preprocessing time and deterioration in image quality. This paper provides two kinds of new optimization techniques which can solve such defects without degrading rendering speed. One optimization technique concentrates on enhancing image quality while decreasing memory consumption without reducing rendering speed by making direct access to the memory space where the initially loaded volume data is stored. The other concentrates on decreasing preprocessing time and hence accelerating rendering speed by creating only one run-length-encoded volume and combining non-photorealistic rendering techniques with shear-warp algorithm. We shall show a novel result that both optimization techniques efficiently decrease the memory consumption and preprocessing time while enhancing rendering speed and image quality simultaneously.