normal from displacement map

[jasonhuang1115]

Hi,

I have a 32-bpc displacement map with value explicitly bound between -1 and 1. Is there a way to convert that to a normal map so I can only apply the normal map to get the rendering result that matches the displaced result?


Thanks,
Jason

[Vizioen]

I think you can use software like bitmap 2 material for this: https://www.substance3d.com/products/bitmap2material

[jasonhuang1115]

Thanks for the input. I have used bitmap 2 material briefly in the past but not sure if that would fit the purpose. I would assume it's mainly for generating a set of PBR textures from an input image.
I am looking for a way/workflow that would convert an existing displacement map to a normal map so I can use the resulting normal to achieve a matched look from the displacement.

[Neilg]

Does using a vray hdri loader to bring in the exr and bump2normal shader work?

[Vizioen]

I'm not sure if a normal map is what you want/need in this case. The way I see it (someone please correct me if I'm wrong), displacement only goes in two directions, up and down, unless it is a vector diplacement map, but I don't think it is. A normal map describes the surface in 3 directions so a regular displacement that only describes 2 directions would generate a normal map that in general would be just as good as a bump. Or not?

Edit: Or is a normal, bump on steroids?

[jasonhuang1115]

Does using a vray hdri loader to bring in the exr and bump2normal shader work?

The idea is to verify the normal map rendering result in vray to ensure it matches the displaced result, but then I would like to take the converted normal map out to a game engine. Can you bake out a normal map with this bump2normal shader?

[jasonhuang1115]

I'm not sure if a normal map is what you want/need in this case. The way I see it (someone please correct me if I'm wrong), displacement only goes in two directions, up and down, unless it is a vector diplacement map, but I don't think it is. A normal map describes the surface in 3 directions so a regular displacement that only describes 2 directions would generate a normal map that in general would be just as good as a bump. Or not?

Edit: Or is a normal, bump on steriods?

That aligns with my understanding of disp, bump, and normal. So the question is how would you correctly covert the "strength" from the disp map to the normal map? From the tests I have done so far, I can increase the amount of normal strength in vray, but it ends up breaking the pattern of details existed in the original disp map.

[joconnell]

Could you render a plane and use a samplerinfo element as your normal?

[jasonhuang1115]

Could you render a plane and use a samplerinfo element as your normal?

Sorry John, not quite following here. Are you referring to baking the normal map out or?

[joconnell]

Yep - orthographic camera above a plane with the aspect ratio of your map, apply your displacement map and then render a frame with the samplerinfo normal element - admittedly it's filtering the output but it'll be driven faithfully by the displace map at least. Alternatively take your plane and give it enough subdivs to match the res of your map, apply the displace and export it as an obj, export a non displaced plane too (no need for high subdivs on this one). Bring them into xnormal and do a high to low bake - https://xnormal.net/

[jasonhuang1115]

Yep - orthographic camera above a plane with the aspect ratio of your map, apply your displacement map and then render a frame with the samplerinfo normal element - admittedly it's filtering the output but it'll be driven faithfully by the displace map at least. Alternatively take your plane and give it enough subdivs to match the res of your map, apply the displace and export it as an obj, export a non displaced plane too (no need for high subdivs on this one). Bring them into xnormal and do a high to low bake - https://xnormal.net/

Haven't done that but worth a try. Regarding the high to low bake approach, I have done similar, if not the same, in Mudbox where I subdivide the object to about 45 mil polys (one more lv up to 120+ mil poly would probably crash the graphics card I have) , sculpt the subdivided object with the disp map, and extract the normal map. This is where I got the discrepancy comparing the result of rendering with the extracted normal and with the original disp map. I guess the details you can extract is up to the poly count you can afford? The disp map I have is for a face that has lots of details of pores and wrinkles.