Real World Diffuse, Specular, NormalMap and Light

[CiroC]

Thanks dubcat. 22€ is not that a bad price for a filter. I will follow closely this thread since this was something I really wanted to do after reading that article.

[dubcat]

Thought I could share the "camera flash"/linear results I got after running around like an idiot in the thunderstorm. There is too much wind to capture any vegetation right now.

The pictures you see are Albedo / Normal Picture / Specular
I'm kinda amazed by how much spec these old bricks have

[dubcat]

I've been thinking about building a normalmap scanner box. My idea right now it to buy a cardboard box and glue some LED stripes on each side. Cut a hole in the box to put my camera lens inside, and then have a switch to turn each LED stripe on/off.
I did a test today with my little 110 lumen flashlight, the flashlight was hand held, crappy result, but it gave me some feedback on how I'm going to build the box.

The Illidan test is with the ceiling lamp on, I wanted to see how other light sources affected the result.

[Juraj]

Heh I like you a lot :- ).

Basically you want to build a cheaper scanner like the QUIXEL guys did for their Megascans.

[dubcat]

Basically you want to build a cheaper scanner like the QUIXEL guys did for their Megascans.

Yes, except I don't have to travel to Pakistan and drive into the desert at night :)

I've been playing around with light intensity this evening, it has always been a pain in the ass. Is the intensity too strong ? Is the exposure wrong ?
It began with me measuring my nifty little LED flashlight. The opening is 2.5cm, 110 Lumen / 9350 Candela.
I took my ColorChecker, flashlight, camera rig and turned the lights off. I placed the ColorChecker 40cm away from the flashlight and used 100 ISO / 8 shutter / 8 aperture.
I went into 3dsmax and made a Disk light, 1.25 radius, 110 Lumen, same camera settings, and guess what. It looked like a nuclear bomb had gone off.

I faced the flashlight towards the camera and made a 14 stop HDR. In 3dsmax I made a 2,5cm plane and used the HDR as light source. A 13x9 plane with 149 RGB, 40 cm away from the light, that's the same as the ColorChecker. After some interactive rendering tests, the material intensity was 0.017.
The equivalent Disk light is 34 Lumen, and not 110 as the package suggest.

I've also matched the cone direction of the light.
Now I can capture the specular maps with the flashlight, and match them in 3dsmax with the same light setup! Woho

I'm sharing the HDR here if you guys want a calibrated flashlight. Just remember to change the intensity to 0.017 and direction to 0.5

[dubcat]

Until now, all the pictures I have posted have only been for reference. Ugly vignetting, only a little spot of specular. Useless for texture work.
But there is a little secret to overcome these problems, it's called "Lighten".
I have taken some pictures of a chimney that looks like it's been painted with shit.

This is how it works:
1. You move your light source around the room and take multiple pictures. You will get better result the more pictures you take.
2. Copy all the pictures into the same document in Photoshop and set "Blend Mode" to "Lighten" on all the layers.

Right now there are some errors in the maps, this is because it's impossible to hold the light in the correct angle in all the photos. I hope this problem will go away when I have built my cardboard box scanner.

I tried a quick render test to see how the maps reacted to sunlight and a spotlight. The workflow is not production ready, but it's a start. The normal map scanner will take it to the next level! (It's using bump map, displacement doesn't work after last daily)

[pionier]

Yeah, cross polarized is so nice, it change the way I thnik about textures and blow my mind when saw it first time three years ago, here is the link to topic about it at cgfeedback: http://www.cgfeedback.com/cgfeedback/showthread.php?t=1854

I also attached my tests, shoot on canon 40d with m42 lens

[dubcat]

Yeah, cross polarized is so nice, it change the way I thnik about textures and blow my mind when saw it first time three years ago

Have you had any luck converting the specular over to reflection ? As you guys have seen, when we subtract the 100% spec 50% diffuse from the 50% diffuse, we get an inverted color 100% specular map. I had a look at ICTs reference pictures the other day, and after some testing I found out that they have applied "-100" Saturation" and then "-1.05 Exposure" to the 100% specular pass. But the reflection map that they have created from those photos are around 214 RGB Median. Maybe they have used the high pass technique that I posted a guide on. I haven't had enough time to play around with it yet.

[dubcat]

I began porting Emily 2 over to Corona because I need a good base shader for my future head scans. My vacation is over so I only had an hour to work on this today, the reflection element still need a lot of work. I can share the shader when I feel it is matching the reference.

[JakubCech]

Hey Dubcat, lately I have been running trough all your posts as you seem to be endless repository of knowledge and scientific tests.

The first picture is about 50% diffuse and 0% specular
The middle picture is around 38-50% diffuse and 76-100% specular
The last picture is 0% diffuse and 100% specular

According to this, viewing the picture with the fur, in the first picture, there is a visible reflection on the floor although specular should be 0%, but its reflecting - is it reflecting diffuse component? I am not a expert in this polarization stuff.
Also for your normal map creation from 4 angles lighting setup, how can I do it? Say I want to create normal map of my rattan chair and take 4 photos, what next?
Thanks.

[dubcat]

According to this, viewing the picture with the fur, in the first picture, there is a visible reflection on the floor although specular should be 0%, but its reflecting - is it reflecting diffuse component?

The fur is lying on my glass sofa table.

Also for your normal map creation from 4 angles lighting setup, how can I do it? Say I want to create normal map of my rattan chair and take 4 photos, what next?

You want a really bright and large surface light.

Something like this (Picture from Wojciech Piwowarczyk, he did some kickass leaves textures)



After that, head over to http://www.zarria.net/nrmphoto/nrmphoto.html

EDIT:
I want to add that I don't use this method anymore. After I discontinued my scanner box I changed over to photogrammetry.

[dubcat]

I saw someone on History Channel used this program in a Tomb in Egypt.
http://www.hpl.hp.com/research/ptm

You can use one/two snooker balls or place your flash in fixed positions.
The guys on History Channel took pictures in an umbrella pattern on two axis.

I did a quick test in Corona. I placed my snooker ball in the center to make my life easier. I only did 15 renders, from what I have read you want 32+.



Here is the result from only 15 badly placed lights.
I need to go and buy some snooker balls, this could be great for fabrics and other small details.
You only need a snooker ball, flash and camera.

[romullus]

Why do you need a snooker ball? Does software calculates light positions from highlights on that ball?
Quality of resulting map is truly impressive, but boy... that really comes at the cost :/

[dubcat]

Why do you need a snooker ball? Does software calculates light positions from highlights on that ball?

Yes. You place a point on the ball, adjust the radius and then click Calculate Position. They recommend two balls for best results.

[Juraj]

I want to add that I don't use this method anymore. After I discontinued my scanner box I changed over to photogrammetry.

Can you elaborate ?

I started to build the scanner, and my brother is doing the software part, I pretty much want the megascans 'ghetto' version. Is this something you moved away from ?

I need it only for micro-detail (fabrics), where I am afraid photogrammetry would simply suck, together with much harder shadow removal.