| |
The technique I've been using to make minifig torsos can be extended to heads,
but I have one critical problem that I haven't been able to figure out yet:
Before tracing an image, I need an image that is facing the camera head-on.
Normally to get a picture to trace, I just take an existing image and skew and
stretch it into place so it's facing me directly. This really won't work for
photos of heads, which have varying depth because they're mapped to a cylinder.
All I really need to make a patterned head is to have an image with an aspect of
26x16 (26 LDU wide, 16 high) that shows the face centered, and I can trace it
and break it up into sections as needed to create the necessary shapes. (The
image would probably need to be distorted first so the parts closer to the sides
were stretched more.) I can find some photos online which I believe are too
small to use easily, but they might be a good starting point if I can figure out
how to correct for their perspective. Example here:
http://www.bricklink.com/PL/3626bpb271.jpg
Notice the face on the left is turned slightly on the Y axis, and the face on
the right is turned even more. Both are pictured from an elevated position. Like
most Bricklink photos I think the size is too small to use my technique on well,
but if I find something with a higher resolution it will still need correction.
Is there any software that would help me correct the perspective in these kinds
of photos so I can use them? The only alternatives I can think of are getting a
better camera and taking my own photographs, painstakingly lining up each one,
or else trying to write my own image distortion tool. Neither is a particularly
exciting prospect.
|
|
| |
In lugnet.cad.dat.parts, Lee Gaiteri wrote:
> The technique I've been using to make minifig torsos can be extended to heads,
> but I have one critical problem that I haven't been able to figure out yet:
> Before tracing an image, I need an image that is facing the camera head-on.
> Normally to get a picture to trace, I just take an existing image and skew and
> stretch it into place so it's facing me directly. This really won't work for
> photos of heads, which have varying depth because they're mapped to a cylinder.
>
> All I really need to make a patterned head is to have an image with an aspect of
> 26x16 (26 LDU wide, 16 high) that shows the face centered, and I can trace it
> and break it up into sections as needed to create the necessary shapes. (The
> image would probably need to be distorted first so the parts closer to the sides
> were stretched more.) I can find some photos online which I believe are too
> small to use easily, but they might be a good starting point if I can figure out
> how to correct for their perspective. Example here:
>
> http://www.bricklink.com/PL/3626bpb271.jpg
>
> Notice the face on the left is turned slightly on the Y axis, and the face on
> the right is turned even more. Both are pictured from an elevated position. Like
> most Bricklink photos I think the size is too small to use my technique on well,
> but if I find something with a higher resolution it will still need correction.
>
> Is there any software that would help me correct the perspective in these kinds
> of photos so I can use them? The only alternatives I can think of are getting a
> better camera and taking my own photographs, painstakingly lining up each one,
> or else trying to write my own image distortion tool. Neither is a particularly
> exciting prospect.
Here are some ideas:
1. If you know the exact physical dimensions/characteristics of your digital
camera (focal length, aperture, diode size, etc.) I bet you can find a tool or
collection of algorithms online that you can use to convert the image into some
intermediary format.
2. POV-Ray is a good tool to "reverse-engineer" photographs IIRC. It requires
lots of technical knowledge however.
3. Locate and purchase an orthographic camera of some kind. As far as I am
aware, however, these only exist for industrial usage. So you may have trouble
finding one or affording it if you do happen to find one.
Those are the only ideas I could come up with.
-Mike
|
|
| |
> Is there any software that would help me correct the perspective in these kinds
> of photos so I can use them? The only alternatives I can think of are getting a
> better camera and taking my own photographs, painstakingly lining up each one,
I think that making good photos is the best way. You need a camera with good
optical zoom, make your photos directly in front of head, and from a distance to
minimize field of view. The photo you get is not very different from a
projection on a plane, so after patterning the whole face you can use SlicerPro
to map the pattern onto the head.
One drawback of most cameras is that their macro position generally work best in
wide angle setting. Since you want a narrow angle to minimize distortion, you'll
end up in a relatively tiny head in your image. But with huge resolution of
modern camera, the head after cropping has enough pixels for patterning
purposes...
Philo
|
|
| |
In lugnet.cad.dat.parts, Michael Horvath wrote:
> 3. Locate and purchase an orthographic camera of some kind. As far as I am
> aware, however, these only exist for industrial usage. So you may have trouble
> finding one or affording it if you do happen to find one.
FYI, it's called a "telecentric lens". Wikipedia:
http://en.wikipedia.org/wiki/Telecentric_lens
Not sure these are commonly available to the hobbyist market, though.
-Mike
|
|
|
