For non-focal blur, it is more useful to use a single ray per pixel. The additional ray depth can be used for true motion blur by offsetting the rays relative to the pixel instead of the camera position. This is done by adding a ray angle to each ray and then using the angle plus ray distance to determine which ray the ray will be in, given its position in the frame. This process is annoying and slow for any camera position-based blur, so instead we'll use a ray simply for inter-ray filtering and operate the blur by ray distance, normally using data from the final frame to determine ray position.
Finally, normal effects such as focal blur, motion blur, bokeh, and so forth are easily accomplished by specifying additional mesh data, where the rays in the different meshes are offset relative to the camera position. For example, for motion blur, we're just going to shift the mesh positions forward from where the last frame's rays were at, effectively going back in time.
So far, there are really three options, and for now they all use one ray per pixel to determine the ray color. The one ray per pixel solution also happens to be the fastest method. However, it is also the hardest to understand for someone unfamiliar with the concept of ray tracing. The other two methods, using multiple rays per pixel or using multiple rays per mesh, have advantages and disadvantages. In general, the multiple ray per mesh approach allows for better transparency effects, while the multiple ray per pixel requires a run-time mesh update, so you might be able to draw the scene so it's always only fully opaque.
If you're doing motion blur and you have a single ray hitting a single point (for example, a single object), then you can't really do much with that. In most cases the first method is the only way to get a quality result without expert-level knowledge.
Using multiple rays per pixel is useful for generating anti-aliasing (since standard AA won't work) or for special effects such as focal blur, motion blur, and so forth, with each additional mesh specified in the camera representing a slightly different camera position. d2c66b5586