Preserved Topic: Multi-Pass Comping #5 (Page 1 of 1)

Finally here.

Directory listing because I'm too lazy to deal with HTML right now. Currently 10 renders totaling around 450k of high jpegs. Another render or two available upon request. Speaking of more renders, I will probably add a caustics pass when I come up with a cheap solution that I'm happy with.

Z-depth was requested last time, so I've got two in there. Just wait until you see how evil they are.

I'm not completely happy with the sky backdrop. Chances are you won't be, either. Few other things that I don't want to taint your mind with.

Multi-Pass 5

It's going to be awhile before I get the chance to play with these. I've already "wasted" most of the day with this and the dishes are calling my name.

edit:
Got to looking at that sky reflection render and decided that I'm very not happy with it. I'll render a few alternative sky reflection renders and drop another note.


[This message has been edited by warjournal (edited 02-16-2004).]

Well, it seems interresting.
Just in case, and once you're done with the dishes, could you add a Z-depth render of the pier ? thanks.

The sky reflection render is ambra.jpg ? If you can make a render with the bump map used/generated on the water surface, it should be ok enough to replace it directly within Photoshop.

Damn you, Poi. You reminded me of a trick that I've used plenty of times in the past. If I had remembered it, I could have saved myself a lot of time in setting up the scene and rendering reflections and refraction. Wish I had more time to talk about this right now.

Added some basic HTML with two new renders.
Multi-Pass #5.

edit:
http://www.1000skies.com


[This message has been edited by warjournal (edited 02-16-2004).]

warjournal:
Oh yeah, those skydomes will really do the trick.

Well now you got me interested in Stroker Normal Maps. I know what a Normal Map (well I though I did) at least from articles on game engine techniques. But don't have a clue how one could use it in Photoshop, or what the values are represented in the render.

Do you happen to have a link on their use or a tutorial on there generation. What render or plug-in you using to generate it, maybe I can find a online manual.

J. Stuart J.

[This message has been edited by jstuartj (edited 02-16-2004).]

jstuartj: I'm not 100% sure, but it seems that the wavesnxyz.jpg image is just a normal map render of the surface of the water. And that it can be used as a Displace Map in Photoshop, at least that's how I'm using it with rather good results.

Yep, that what I imagined for it's use. At least that what I planned on using it for. I guess you could cycle it, positive and negative to generate an animation of moving waves.

J. Stuart J.

Normals

In the 3d world, a face points in a given direction. That direction is called the normal. The normal tells the renderer how the given surface reflects the light (incident?). If you can manipulate the normal independant of it's surface, you can do some funky tricks. For example, a surface can reflect light as if it were facing one way when in fact the surface is facing another way alltogher. Just by manipulating normals, you can make objects disappear or psedo-fake a watery surface.

In the 2d world, there is R, G, and B. So, the 3 different channels can be used to encode the 3d information. Each RGB channel represents the "bend" of a single 3d axis.
We get something like this:
Red = x
Green = y
Blue = z
When you slap them all together, you get a colourful image that looks very similiar to a D-Map. As a matter of fact, it's not entirely inaccurate to think of a D-Map as a 2d normal map or vice versa. :snicker:

In 3d, there are several views. Two commons ones are camera and world. For a proper normal render, I believe it is done from the camera view. But don't hold me to that because I'm not entirely sure. From what I've read and the fiddling that I've done, that is the case. Someday I'll do some proper varification.

Now, wanting to do my own normal renders, I ran into some problems. The main problem being that I can renders normals in world, but not in camera. When done in world space, the full 180 degrees is not used. So I had to hack together my own version of normals. I call these "stroker normals". Stroker normals use the full 180 for each axis and are in world space. So far the only real disadvantage of stroker normals is that some "resolution" is lost (around 17% total or 50% of B). That lost resolution goes back to how camera normals are rendered.

Egads.

Anyways, I originally got the idea for this from Raytracing in Photoshop. An old favorite that might explain some of what I'm babbling about.

As it turns out, some folks have recently been doing some similiar normal work with PRMan 11. They do a "bent normal" render, which is normal data that is modified with occlusion data. I'm pretty close to being able to do bent normals renders with BMRT. However, I'm thinking about switching to Pixie for various reasons.

Jstuart, normal maps are a bit different when used in a game engine. It's mostly the same kind of data, but it's used as a hyper bump map in local space. One of the first poplular articles explaining them: Creating Normals Map with C4D. Should also provide some better understanding of regular normal maps. I'm sure you will be able to see the connection between this article and the Raytracing in PS article.

edit:
Several edits. Started yet another and forgot what I wanted to add/change.
[This message has been edited by warjournal (edited 02-16-2004).]

[This message has been edited by warjournal (edited 02-16-2004).]

[This message has been edited by warjournal (edited 02-16-2004).]

Thanks, useful info there. Say if you happen to have time, could you render a pass for just the water plane. Currently there are some problem with trying to add depth of feild. as the pier tends to blur it the background. It would be a lot easyer to control the depth of field.

J. Stuart J.

Jstuart, I'm not quite catching your drift for another render. The "stroker normal" render of the surface should get you there, but I could be wrong because I don't understand your request. Be a bit specific on how you are slapping things together?

All of this normal talk is really getting to me. I've been digging into them and several surrounding things for quite some time. As you can imagine, I'm seriously getting the bug to finally talk about these things - and you know me when I get the bug.

Chances are I will be starting another thread just for normal chat and my recent travels with them. It will be a healthy dose of material. Now, most of it will be 3d oriented and probably belongs in the 3d forums, but here just plain feels right. Besides, maybe some of you folks can help bring it more to the PS side of things and do some cool extrapolating.

Sorry, I wasn't as clear as I thought, I am no loinger on the normal map. What I could use is an ambient render of the oceans surface Being 100% and Reflective. Exactly the same a ambra.jpg, but with out the pier.

Because of the way gaussian blur and some other filters work. You can't blur the ambra.jpg with out data getting pulled from pier, which spreading into the sky on blur. Sure you can mask the pier but it still has an effect on the blured image.

Not a big problem, as one can cut a mask or clone to replace the effective area easyly enough. I though if might help in future multipart renders if I brough up issues that caused problems.

Don't have much time to play with it anymore week anyhow, and I have another project that I have half done and should really on this weekend. Perhap I will render it in multipass once I finish modeling it.

J. Stuart J.


[This message has been edited by jstuartj (edited 02-17-2004).]

Gotcha. I was kind of apprehensive about providing the requested render. Why? Because of the normal render and the sky domes available at that link. Using both of those, you should be able to come up with a very fine backdrop of your own that doesn't have the pier in it. But I'm in the mood to be a nice guy.

Multi-Pass #5.

Excellant point about the Gaussian bleeding.
Anybody out there with PS CS care to tell if Lens Blur does this correctly?
Here is that link showing the deal with Lens Blur: Part II.
It certainly looks like Lens Blur doesn't bleed, but I would appreciate some confirmation.

Thanks, I have CS and will give lens blur a try. I only played with it a short while and so far I haven't been impressed. I found it diffucult to get the effect I desired, slowwwwww (Expecially on print resolution images.) , and I felt I could get a more controled effect simply blurring through severl soft edge masks. If it in fact it doesn't suffer from the blur bleed effect it could prove useful.

So far it appears to be one of those wow-wee wiz bang marketing features, much like the annoying filter gallery. Fun toys to play with but mostly fluff.


[This message has been edited by jstuartj (edited 02-18-2004).]

Ok, I just PS-CS Lens Blur a quick test. My first impression was wrong. Lens blur does seam useful, expecially for 3D post work. I imagine it could save some time when rendering with a z-buffer would be faster the rendering depth of field, only to get it wrong and have to re-render.

Here's what I did.

I hacked out a simple composite with floats containing the letter "A" in focus. I then loaded the depth map "zsurface.jpg", and "alpier.jpg" each into seperate alpha channels. I then created a custom depth mask by inverting the "zsurface", loading the "alpier" as a selection, I filled it with 100% Black as the Lens Blur filter treats black areas as nearest to the camera. (You also can also invert the map from with in the Lens Blur filter.)

I then applied the Len Blur to the composite image. I selecting my custom alpha channel as the depth map. With the depth buffer active, Lens Blur allows you to clicked on a preview image, in the location. It then samples the depth data at that point and applys the bluring accordingly.

I select the outer most tip of the pier, as you can see, the pier wasn't effected, but the letter "A" on the floats beyond were. So lens blur does work well, It requires that you set-up your depth data correctly, which means setting anything you don't want to effect to black. The only real problem I found is the depth of field on the multipart render is very shallow, making it hard to control the fall off point. But setting the pier to black solved that problem.

Creating a truely useful depth mask for a photograph could be tricky by hand. But in 16+ years of professional retouch work. I have only been asked to create this effect 10-15 times. So I don't know how much use I'll get out of it.

I could be intresting to combine the two techniques. Creating a virtural set ruffly based on a photograph and then generating a depth map off of the virtural set. It would ease the creation of more realistic depth maps, and allow some truely intresing depth of field effects. Worth playing with in the future. I now wonder if it's recordable so it sould be applied as a action in a batch processes of still frame.

My example of the lens blur test is here.

www.texturehunter.com/img/temp/LensBlurTest.jpg

[This message has been edited by jstuartj (edited 02-18-2004).]

Once I got everything into my PSD, I went to start adjusting and realized I had started my MODE in greyscale... DOH!

Wasted an inordinate amount of time trying to change to RGB without flattening the image and then decided to take a break. Scratched my head, closed the graphic, then diddled with some other stuff,

came back here.... And...

WOOSH!

That's the sound of all this stuff going over my head.

Anyone got a Multi-Pass Filter?
[snicker]

______________
Is This Thing On?

Bleah...