4
$\begingroup$

Given a 3D space and a light source, I use ray casting to determine the shadows generated in the scene. If I'm to store the shadow data (boundary of the shadow, object which cast the shadow, etc.) what is an efficient way to go about it?

Edit:

Its basically for an application where I want to be able to move the shadow around manually to see the change in the 3d objects position. Say I apply a force to the shadow of a ball, I can actually see the ball move and also the change in the shadow. So I need to be able to set up colliders for the shadow and hence need a way to be able to store the boundary of the shadow and information about the object that cast it to carry out the actions.

Improve this question
$\endgroup$
6
  • $\begingroup$ In my understanding, raycasting isnt very good for boundary information. Depthmaps also dont dont benefit from raycasting. Deepshadow maps might benefit from raycasting. Otherwise you would just be better of baking lightning to model uv maps or some kind of voxel tree. Possibly per vertex/face. The question is a bit hard to swallow, could you specify it a bit. Best is not defined, efficient in this case is not terribly defined either $\endgroup$
    – joojaa
    Commented Oct 19, 2015 at 6:15
  • $\begingroup$ Are you asking for something geometric such as Franklin Crow's Projected Shadow Volumes? $\endgroup$
    – Simon F
    Commented Oct 19, 2015 at 12:19
  • $\begingroup$ Please fill in the "etc" and tell us the purpose of storing the shadow data. $\endgroup$
    – user1703
    Commented Oct 19, 2015 at 21:07
  • $\begingroup$ I apologize for being so vague.. Its basically for an application where I want to be able to move the shadow around manually to see the change in the 3d objects position. Say I apply a force to the shadow of a ball, I can actually see the ball move and also the change in the shadow. So I need to be able to set up colliders for the shadow and hence need a way to be able to store the boundary of the shadow and information about the object that cast it to carry out the actions. $\endgroup$
    – ichigo1191
    Commented Oct 20, 2015 at 1:03
  • $\begingroup$ Can I infer that "etc" is void ? What other information is relevant ? $\endgroup$
    – user1703
    Commented Oct 20, 2015 at 8:49

2 Answers 2

Reset to default
3
$\begingroup$

If the goal is mouse interactivity, you can do with a simple "shadow buffer", i.e. an image that holds the identity of the occluding object (if any) on every rendered pixel. You will compute this map during the casting process as you shoot the rays to the source.

In case there are several occluding objects, it is up to you to choose which one to consider (or keep a linked list of all occluding objects). If there are several sources, you can keep one map per source.

Improve this answer
$\endgroup$
4
  • $\begingroup$ Why not trace on click? There sa really neat trick if there is less than 32-96 objects. $\endgroup$
    – joojaa
    Commented Oct 20, 2015 at 14:48
  • $\begingroup$ @joojaa: no idea what you mean. $\endgroup$
    – user1703
    Commented Oct 20, 2015 at 15:55
  • $\begingroup$ Why make a buffer of rays? If your only interested in one ray, shoot the ray when your needing it. Its fast enough. Why shoot a shadow buffer with rays just use a separate pass and use depth maps + index color. Again you can do this on demand. If you have 3 color channels you can use a bit mask for surfaces hit tahtway you can store ALL objects in pass of ray if you must etc. Too many open variables in question to nake any meaningful impact. $\endgroup$
    – joojaa
    Commented Oct 20, 2015 at 16:21
  • $\begingroup$ @joojaa: I guess this comment belongs to the OP. I am answering the question "how to store ?". $\endgroup$
    – user1703
    Commented Oct 20, 2015 at 17:05
0
$\begingroup$

You might want to look into signed distance field textures. They are very efficient and high quality for monochromatic images like shadows. They work by storing a clamped, signed distance from the pixel to the shape. Working this way, the distances play nicely with the bilinear interpolation of texture sampling, so you can have very nice results for very low resolution. You can also give them a soft edge to have a soft shadow look.

Improve this answer
$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.