3D shadow scanner

Hello everyone!

First I want to express my gratitude and respect to Fiend3d for the wonderful tutorial that inspired me to write this article. I have always experienced envy in half with delight, watching how the physical world turns into a set of equations in the mind of a mathematician. nine0007 I also want to warn against attempts to take my exercises for ready-made technology. It's just a fun experiment. If someone finds it useful, it's not my fault :)

So, let's get down to business! Discussing how to make a three-dimensional object out of a flat picture, I remembered the depth channel (Z-buffer). All we need is to get a photograph of the object with the illumination uniformly decreasing in depth. And preferably without perspective distortions. Most of us have such a camera on our desktop (or at least on the next table). This is a regular flatbed scanner. nine0003

But of course, not everything is so simple.
Problem #1: objects have their own color. And dark spots differ little from shadows. I solved this problem simply. I took an object made of plaster :) Here it is:

By the way - I had to wash it a little :) Of course, if you want to take something not from plaster - there is always a can of paint :) I scanned with low resolution, but 16 bits per channel to preserve the nuances in the shadows and highlights. By the way, keep in mind: the location of the shadows depends on the direction of movement of the scanner carriage. nine0003

A bit of Photoshop next. I took the red channel (channels/split channels) and recorded it as a separate file (this is better than a simple "discoloration" - desaturate). Yes, I almost forgot: it's worth "cutting" the object and filling the background with black. And it's not bad to additionally highlight the most protruding places (don't grumble, it took no more than 15 minutes): Explain to 3DS Max what to do about it. The astute reader is already screaming: "Display! Display!" - And he's right! :) But I don't want to use a display card when rendering, it's more convenient to work with geometry. So - the Displace modifier, which I applied to a regular Plane object. In order not to suffer with the fit, the map was brought to the proportions of the object 3x4 (1500x2000px):

Oops! Somehow I imagined it differently... So problem No. 2: the regular grid divides the object into squares , as if "pixels". If we increase its frequency, we quickly fly out for a reasonable number of polygons. In addition, the picture is slightly "noisy". Obviously smoothing is needed. We remove micro-noises directly in Displace (Image/blur), macro-irregularities using the Relax modifier:

"Let me!" - the indignant reader will say, - " Where are the small details? What were you fighting for??"

Yes, and this is problem №3 . It is relatively easy to solve. From the same picture we make a bumping map, of course normal bump. (I used a small free program Сrazybump )

Now it would be nice to remove all unnecessary things.And what are these nasty stains along the contour ? the shape of the object. The scanner cannot look around the corner, and where the scanned map ends, crap begins. So we just cut off these parts completely. No need to suffer and aim. Use the opacity map. Remember, I asked you to "cut" the picture first? You should have kept the alpha channel!0003

Doesn't look too much like plaster, you say? Well, it's time to include illumin -shine:

and finally in movements

Here are the final layouts for modifiers and material settings:

Fucked 2x1800/ATLON64/2GB, Win2000SP4, Autodesk 3DSP4 , V-Ray 1.53RC1

I hope I managed to entertain the audience a little :) thanks for your time!
Mikhail Naumov a. k.a. Mic Nau

526 0 850 31

19

2007-12-05

Good lesson. i liked it :)

2007-12-05

But how to completely make a model in all 360 degrees?

2007-12-05

yes the lesson is not bad... I get good stucco molding in the typical way... But it's not a 3d scanner... and unfortunately Displace can't be converted into a model...

2007-12- 05

if the semitrich model, then you can apply Miror. if not, then it’s okay to connect the two sides somehow +) ... the sense is crooked somehow ....) I didn’t get it right :(

2007-12-05

I liked the lesson! Only I’m too lazy to repeat it :)

2007-12-06

convert to geometry)) (yes, not in 3d max) but you can)) we will not disclose in which software specifically BUT)) 2dimson 3d )) unfortunately I don't think that having 1 mona picture will make the whole model ))

2007-12-08

1) Gygygy .. The display is excellent and in max it is distilled into geometry))) I will not say "connect the Brain" BUT. 2) The article is funny. As an option for the lazy - will do. Unfortunately, the free program CrazyBump - gives a normal map that is remotely reminiscent of the truth. And to the same regret, most people for some reason do not see this. \u003d \ Relief - and hurt yourself. And the fact that you put it in a brazen stupid photo on the bump and get + - the same Nonsense .. no one understands. 3) The author forgot to mention the intricacies of scanning. I'll clarify - Fotik will not work - you need a scanet with the ability to "Volumetric" scanning. Do not be afraid of this word, it has nothing to do with 3D. Simply Previously, scanners had a focus point that was located just above the surface of the glass, and everything further - Blurred. Now almost all scanners work like cameras - ie. take a picture coherently deep into a long distance. those. like volume. and not a flat sheet.. And the second and most important thing is that you need the same random-dimensional and unidirectional light. That is why they use a scannet. The effect of depth is obtained due to the attenuation (scattering) of light and, accordingly, the decrease in its intensity. Those. - When scanning, the room must be Dark. so that extraneous light does not spoil everything. Last post: In any case, the result is unimportant. And in terms of time and crap .. hmm .. a normal, not armless modeler during the same time can easily sketch such a model in 3d. in addition, the quality of the Form will be higher .. and the detailing .. 15 minutes in the brush (again, for the less experienced) P / s / why I say all this so confidently - I myself did Exactly the same experiment about 5 years ago. )) if not more. nine0003

2007-12-09

2 Shiva: CrazyBump really gives normal bmp far from the truth. But "in a brazen stupid photo on the bump" - still worse - checked :) The author did not mention the intricacies of scanning quite deliberately, so as not to clutter up the article. But about the fact that the room should be dark - yes, I confess, I just forgot. Probably because it's always dark here :) And finally, about "lightly sketching" - if it's not a secret, how long did it take, for example, for this model: http://img192.imageshack.us/my.php?image=demon7xs.jpg ? :) And really - thanks for the qualified answer! It was not in vain that I wrote at the beginning that this is "just a fun experiment." The actual application of such technologies (even if everything were perfect) is very limited... At best, put the figurine on a shelf in the interior, and even then, do not fly close to it... Thanks to everyone who took the time to read to the end :) Experiments, however, continue... but in other directions

2007-12-09

I read somewhere the way of such entertainment. There, the normal map was obtained by photographing the object alternately illuminated from three sides along the coordinate axes. The resulting pictures are thrown on the channels. further clear.

2007-12-10

Misha, thanks for an interesting test! :)

2007-12-12

2 Mic Nau :)) I answer - this model took . . hmm .. so not to lie - 15 minutes of a lunch break. During these 15 minutes, it was First a classic Alien with a big head, then a zombie from the Iron Maden group, and only then this Elk ..) and then another 20 minutes dabbled with retopology in Silo - redrawing the grid neatly. 2 Taras Burlin: yeah, there is a way. much better result. But, there are some drawbacks - shadows .. That is. illuminating from one side, we get shadows cast by protrusions .. But .. if this is a small bas-relief (modeling) - then it turns out almost 90% original.. if you want a display, if you want a Normal map.. and Very accurate and high quality.. eh.. if it were not for the shadows.. =\

2007-12-15

time of work activity.

2007-12-15

2 Fiend3d :)

2007-12-22

why in the original the horse's mane is flat, and as a result - some kind of convex?

2008-01-15

dimson3d I think just take a photo from both sides, i.e. from both sides and all

2008-01-17

this is a good way to make a symmetrical model and poor quality, but this horse is not symmetrical at all. I would make polygons, the details of the wings and muzzle - bump or shake with brushes.

2008-05-26

But they didn’t try to scan their physiognomy

2008-05-26

[smile=04] the first thing that came to mind... I was too lazy to make up

2004 cool if you need to make a bas-relief.