Question: Errors When Using Gradient Displacement Map

[josh_dannenberg]

I'm having trouble getting a gradient displacement map to work correctly. When I apply the map, the result has bizarre stripes or ripples in the displaced geometry. I'm not sure if the problem is the gradient jpeg file or vray's interpretation of it. In any event, if someone knows how to fix the problem, that would be great. The images below show the files I used, as well as the desired result. I'm also attaching a .zip with all of the files included, in case you want to replicate the problem.

Zip File:
http://www.studio-office.com/ASGVIS/WOBBLE.zip

Desired Result (variable orientation in glass at left):


Rendering:


Displacement Map:

[dalomar_cg]

Re: Question: Errors When Using Gradient Displacement Map

I do know what the problem is...basically, 8 bit color files are only so accurate. So it won't take much to see the inaccuracies inherent within the format. My suggestion would be to use a more accurate format such as 16 bit or better yet 32 bit formats (hdr or exr would be the best). That might help you get where you're looking to go.

[dalomar_cg]

Re: Question: Errors When Using Gradient Displacement Map

Ummm....I'm starting to think now that its just an issue with the image itself. If you look very closely, you can actually see the banding in the image that you're seeing on the surface. Thats most likely a jpeg compression issue, so either A) use the highest quality compression (ie close to none) that you can. or B) use another format that doesn't use compression or has lossless compression, bmp or png respectively.

I'm not saying that my previous suggestion doesn't play a part, but I think my two suggestions tell two sides of the same story.

I'm still testing the first 8bit-32bit idea right now (crazy accurate displacement mesh can take a while )

[josh_dannenberg]

Re: Question: Errors When Using Gradient Displacement Map

Yes, the original jpeg i used did seem to have some banding. I just did a few tests with 2 fresh gradients (1 generated in P-shop, the other exported from Illustrator), both using 16- and 32-bit variations, and the results still aren't great.

But, when you say "crazy accurate displacement mesh can take a while...", will different render engines give different results? I'm using the ir-map/qmc-medium settings right now.

Also, do you think there's a better way to make the gradients other than P-shop or Illustrator?

Thanks for the help!

[dalomar_cg]

Re: Question: Errors When Using Gradient Displacement Map

But, when you say "crazy accurate displacement mesh can take a while..."

That was just me... in order to make sure that I got the mesh as absolutely smooth as possible I set the maximum edge length to .25 pixels, and then proceeded to render at 1600x1200. Definitely not the most efficient settings, but that's not the norm anyway.

Also, do you think there's a better way to make the gradients other than P-shop or Illustrator?

I'm not really sure. I actually wasn't too happy with the gradients from ps. You could see in the displaced surface that the 8 bit gradient wasn't constantly linear. I'm not sure if there's a limitation with that, but it didn't seam as precise as I though it would have been.

Ultimately, since this is meant for smooth reflections, I'm not 100% sure that displacement would be the best way to go with this. When it comes down to it, an image will always have, at one scale or another, little steps within the information. Maybe with a larger image and/or a more precise mesh you can get smooth those steps out, but I don't think its going to really be as smooth as reorienting the plane as you need it. However, since this shot is from farther away, you could most likely find some settings that would be satisfactory. Its just a matter of playing around with it to find out what those are.

[josh_dannenberg]

Re: Question: Errors When Using Gradient Displacement Map

Ultimately, since this is meant for smooth reflections, I'm not 100% sure that displacement would be the best way to go with this.

I think you're right, Damien. So I tried working with the procedural settings in both the bump and displacement options and had some success using the checker option, replacing the black-and-white colors with various noise settings. This was good in that it allows me to work with a sizable grid, but bad in that the checker setting produces a diagonal pattern:

A B
B A

...where the A's and B's continue in diagonal lines. Do you know if there's a way to randomize the noise within the checker, so that there is no legible pattern, something more like:

A1 B1
B2 A2

Thanks again for your help.