| Author | Message | ||
     Puppet Kite Kid Rating: N/A |
Creating and Using a Virtual T-Square for Determining the "Best 3D Effect" for a Stereoscopic 3D CGI Scene: http://www.puppetkites.net/stereo3DCGI/tsquare1.htm Let me know if anyone goes through this and tries it. It seems to be working very well, and is an alternative to using something like the "1/30th stereo base rule" which only works well when you have both a near point and a point at infinity, which is rare, actually, and especially so in stereo3D CGI. You can read my opening statement, which includes sort of a disclaimer, but this indeed is why movies like Polar Express worked so well. Of course, you don't have to get obsessive about it, either, but it might be a great way just to check up on your 3D effect every now and then ;-) PKK | ||
| Puppet Kite Kid Rating: N/A |
Okay, I have done my homework... :-) Get ready :-) Now that I have studied what "perfect 3D" really looks like, I can really "see it", now, and it is amazing to observe. I am testing the following theory over and over and it works. Now, I know exactly why Polar Express worked so well as far as stereo effect is concerned :-) Actually, contrary to what I originally guessed, the "1/30 rule" is a visual rule, based not on cameras, but on our eyes. I won't go into the details, because I am not a mathematician, but I have searched the internet and found some great mathematical explanations by the *experts*. I can find some links, if needed. Abram K once explained it in terms of complex math, angles, degrees, etc, and although I can't follow or compute the math, I do basically understand it. So, up to this point, cameras are irrelevant... the "optimum amount of 3D effect" is based on a "1/30 rule" that's based on the way we see, not our cameras. So, indeed, as I _did_ guess, the amount of "optimum stereo effect" is a given, and it is a constant, therefor all camera formats, focal lengths, etc, can use the original "1/30" rule as a guideline. So calculating this "optimal amount of 3D effect" in the field turns out to be as simple as it can be, but it will require very different settings for all focal lengths, stereo bases, distances to near points and far points, etc, but actual eyeballing this measurement is easier than you'd think, too. Just look at the distance between the two perspectives of your farthest point in your frame and then look at the distance between the two perspectives of your nearest point. The difference in those two measurements needs to equal 1/30th of your frame width. Even if you disagree about the 1/30 rule, or if you want to stray away from what our eyes see as "perfect", you can adjust your optimum measurement to fit your preferences. I am practicing with my "virtual T-Square", which does this measurement very simply and extremely quickly with CGI, but I have to use the edges of my viewfinder to do this with my cameras. I will try to work up some more examples, but again, here is a PNG overlay you can use to check your images: http://www.puppetkites.net/stereo3DCGI/deviationprintouttrans.png PKK | ||
| Puppet Kite Kid Rating: N/A |
If this demo doesn't clearly prove my theory, nothing will :-) Here are three examples of what I call "an optimum amount of stereo effect" as strictly defined by the "1/30 rule", which indeed produces, by definition, an "orthoscopic" view, but these images are using a variety of focal lengths and stereo bases, including a couple of that might be considered "extreme" examples. The third image will potentially create some controversy, but I can still stand fast to my theory. I also included the "T-Square" that I use to do the "eyeball" measurements, with no math involved whatsoever, so people can visualize the idea better and perhaps apply it to what they see through their viewfinder. The stereo bases are "1x to the nearest point", are "approximate", and the near point is at the nearest point of the cube and the far point is near the back side of the cube. The T-Square is also always placed at that point, as you can see in these images when you look at them stereoscopically. The cube itself was never "resized" in any of these examples. I did some cropping, so the original "1/30th of the raw image frame" measurement of the deviation might be different, but I didn't check... just don't worry about that. The method is always the same... subtract the far point separation from the near point separate and make it 1/30th of the raw frame size, no matter what you have to do, i.e., move the cameras, change the stereo base or the FL or any combination that gives you that result, which is what this demo shows. Parallel images: Lens: 48mm Stereo Base: 1/10 http://www.puppetkites.net/stereo3DCGI/orthocube2_P.jpg Lens: 135mm Stereo Base: 1/15 http://www.puppetkites.net/stereo3DCGI/orthocube3_P.jpg And now for a potentially controversial one :-) This is an extreme usage of stereo base and FL, which results in a severe perspective distortion, but the same "optimum amount of stereo effect" as defined by the original 1/30 rule still applies, identically to the other images, so by definition, this is still an "orthoscopic sound" view"! To prove that point without the math, simply notice that it is very "pleasing" or "comfortable" to look at this image, even though the image perspective is very "strange". Again, "orthoscopic" pertains to "how" we see, not "what" we see. Lens: 24mm (Look at this stereo base!) Stereo Base: 1/5 http://www.puppetkites.net/stereo3DCGI/orthocube4_P.jpg PKK |
