I would like to make a series of normal or even bump maps for common NZ coins. Most of the time if I use something like a flatbed scanner, the resulting image is fairly useless because of the bands of reflectivity and generally poor image... would need to be repainted by hand... not so simple when trying to convincingly replicate the fur of the kiwi on the $1, or the Queen's head on the other side.

...simpler method ideas please?

The following idea is not exactly simpler, but definately a fun project.

If you go buy a red, green and blue lightbulb you can create a really high detailed accurate normalmap from 3 photographs.

Most people doing textures and image maps use a high rez cameras set it on a tripod, and take hanging from a wall with soft diffuse lighting, why a camera is that you can change exposure setting you are capturing and store it in raw to then take it into a software package to edit and remove glare like the shin of human skin or the sss out of the skin awell.

i know a guy at rhythm and hues that takes the photos of the actors for the 3d department ill see what his software it either way its not a cheap package. There are work arounds in the form of photoshop scripts/ actions but they just as much work and trouble, as doing it from scratch most of the time.

well hopes this helps a little ive flicked off a email post what he says ok and for what its work you should probably add the colour maps as well?
laters
S


ps crazy bump is 99 us and does a great job at giving you a diffuse, spec, AO, Normal and Reflection maps with about 10 or so clicks from a colour photo.

Thats not as fun as my way

Lol

I like lightbulbs... be interesting to try out and see what kind of results it gives.

Nice answers, thanks.

If you do try it, just make sure you get the lights in the right places

I vaguely remember doing coin related work before and ended up just using images i googled -

once you desat it and play with the levels a little you should be pretty close - a little bit of photoshop work wouldnt hurt.

can't you use a regular lightbulb and add colour in photoshop?


what if you paint the coins with some matte substance before taking photo would that help kill the reflection?

im not sure how well that would work, if at all (adding colour in photoshop). But i havent tried it.

Actually - I'd been thinking exactly the same thing. The main thing which kills the clarity is the extreme reflection & shine, so painting a thin layer of white could be useful.
I'd also thought it would be too difficult to get an accurate mix of real bulb colours... so something like a greyscale white from each of the three angles, have each one filtered to the UV colours as layers and multiplied / added over each other would probably work.

...now I just gotta put it into practise and see.

A crude experiment confirms that the Laz's light bulb method (along with link_choi's modifications) actually works!

I just set a camera up on a tripod facing downwards and physically moved a florescent bulb around each of three shots to simulate uv normal lights from different angles. I got blurry results from the lack of proper nightlight settings...



Then, I estimated what the three UV colours should be for each angle based on a normal map I made of a cone. (Note that different 3D applications have uv normal colours from different directions).



I made a three layer Gimp/Photoshop file of each colour, and applied the black and white coins as transparency masks of the coloured layers... I didn't do any real merge method; just "normal" layers stacked to make something looking like a normal map (this would make the top layer dominant - I was being lazy)



Given all my inaccuracies, applying the resulting map as a texture turned out better than I'd expected...





I reckon proper studio lighting of the coin, along with some real analysis on what the official normal map colours should be (e.g. not just my guess of them), with a more definitive merge method could indeed make very convincing results.

Thanks for your pointers everyone.

Hey great work Lancer!

With a good camera, it's possible to pull an incredible amount of 'physically' accurate detail with this method. Is it a production-ready technique? Depends, once all is in place, you could whip through many assets very quickly. But most of the time being physically accurate is not really that important. If you got large enough images with tonnes of detail it wouldnt be hard to turn the normal map into a displacement map (there is software that does this pretty well)

What we really need is a camera that is able to capture depth information. Now that would be kickarsey.

That's possibly the coolest thing I've seen all day. Well done on not having to mess around with coloured lights.

To calculate depth, all you need is an array of finely spaced cameras and a PhD is awesomeness or something similar. People are writing papers on this stuff. It can be done with two cameras by calculating paralax, but you get errors where one camera can see something the other can't. That's where the array of cameras and the awesomeness come into play. Nuke's Ocula does a pretty amazing job just with two cameras, don't know how it corrects for the errors though.

Yea, imagebased modelling apps have been around for donkeys years, this kind of tracking is only getting better and better. There is motioncapture software that allows you to track a person, without markers with standard handycams, thats crazy cool.

BTW. I hold a PhD in Awesomeness from IUEC (Inconceivable University of Extraodinary Coolness)