Stereo Anaglyph Images - How to...
pleased to hear that you enjoyed the anaglyph
images! I must admit that at times it is difficult to get the images to behave
as well as one would want, as much depends on the dark/light tonal range of the
original colour masters as to how well they translate into a red and cyan range
of monotones. This almost always varies every time, so not consistent in output,
as you may well have noted from my
samples.
Anyway, for what it’s worth. I will set out the procedure for you to try out for yourself (Apple Macintosh keys named here, substitute 'CTRL', instead of 'COM', for PC):
Anyway, for what it’s worth. I will set out the procedure for you to try out for yourself (Apple Macintosh keys named here, substitute 'CTRL', instead of 'COM', for PC):
First of all, the Strata
masters:
EARLY VERSIONS OF STRATA Set up a new camera (lock the ‘roll’) and
compose your scene using all the camera attributes as required.
‘Replicate’ this camera in one plane only (X OR Z but not all), as seen
from the ‘Top’ view. You need to set the TWO cameras side by side, both
angling in slightly, pointing at the same centre spot of your object.
The angular separation will exaggerate the perspective so the cameras do
not want to be too far apart. It would help to identify them using a
custom colour texture, RED for LEFT, CYAN for RIGHT, so you know which
image-half you are processing later. Once you have set these attributes,
you could ‘Group’ the cameras together so if you have to move them about
they will not lose their settings.
Each camera image is rendered separately using ‘Raytracing’ at ‘Finest’,
‘Extra Smooth’, and ‘Expert: Pixel Detail Block Size = 1”. This ensures
better photorealistic image resolution, especially for bit maps. It is
also better to use pixel bitmaps four times larger than an object
measured in pixels, and scale to fit, to intensify detail.
You will be rendering each camera image separately in full colour, so it
is important to remember which camera produced which image when saving
the rendered result, so ‘Save As’: “Object123-L”,“Object123-R”, etc.
STRATA 3D CX This is much more elegant in operation. Set up your camera
as before, and now go to the ‘Object Properties’ Palette and change
Camera type from ‘Normal’ to ‘Stereo’. You will notice that you can also
set ‘Convergence Distance’ and ‘Eye Separation’ there. The focussing is,
in principle, the same as with the two cameras before, but this time you
drag out the camera focus point on screen to your object centre axis.
You could also match the convergence distance to the distance of camera
to object, but convergence does not seem to make a vast difference to
the result, the default seems OK. I have not covered all the focal
lengths that might be employed yet, and I suspect it will improve image
perspective by tweaking that carefully.
The eye separation is what is says, so it pays to be cautious here,
unless you wish to simulate the view that a horse, or larger mammal, may
have of the world! So far the default again has proved reliable enough,
but you could accurately measure your own eye separation in mm and use
that dimension.
Rendering options are exactly the same as before, but this will render
all in one, resulting in both Left and Right images being rendered in
the one file. There is a choice in the ‘Render Image’ window, and I have
‘Side by Side’ and ‘Left First’ selected here.
It is important not to have any pure white highlights, so keep the
lighting even if possible. Contrast can be increased in post-production
if it proves necessary.
POST-RENDERING PROCESSING This uses Adobe Photoshop (any version). Use
‘Open File’ to bring your Strata PICT files into the application.
On the RIGHT image (select right-hand half if a CX file), use ‘Com-A’
(Apple Key+A...But you knew that!), then ‘Com-C’, then ‘Com-N’, then
‘Com-V’.
That was ‘SELECT ALL’, ‘COPY’, ‘NEW File’, ‘PASTE’. You will notice that
it produces a new file exactly to the size of the copy. (Better never to
work on the originals!). Click on the ‘Layer’ name and type in ‘RIGHT’
(Press return).
Select the LEFT image and ‘COPY’, but this time ‘PASTE’ directly on top
of the new file, and name it ‘LEFT’.
On LAYER ONE (LEFT and RED), Select ‘HUE/SATURATION’, and ‘COLORIZE’
with HUE = 0, SATURATION = 100, and LIGHTNESS anything down to -50
(max). Now TURN OFF this layer with the eye icon.
‘SELECT’ Layer Two.
On LAYER TWO (RIGHT and CYAN), Select ‘HUE/SATURATION’, and ‘COLORIZE’
with HUE = (235?), SATURATION = 100, and LIGHTNESS anything down to -50
(max).
‘SELECT’ Layer One.
Then TURN ON Layer One again and set Layer Filter Attributes to
‘LIGHTEN’.
You should now see both images together! Using the cruciform ‘MOVE’ tool
tweak the alignment so there is perfect vertical alignment on eg: the
horizon line, and perfect lateral alignment on objects or edges nearest
to your viewpoint.
You can use the ‘CROP’ tool now, to trim off any misaligned raw edges.
Depending on the exact colour value of your glasses, you may need to
tweak each image now independently. This is the HARD part! The CYAN
value can be altered (235?) in Hue/Saturation again, when Layer One is
selected (maybe while turning off Layer Two), so that the whole Cyan
image looks as black as possible when viewed through your RED lens (this
is very obvious in practice, there is a clear cut-off point either side
of correct).
There is a lot of variation in lens spectral ranges, I have noted. Some
really smart plastic-lensed 3D glasses I had mailed from the USA behave
differently from my ordinary card mounted gels, so I imagine yours may
differ also.
Do the same for the RED. The LIGHTNESS value on each layer is crucial
here too! Anything down to -50 could work. It all depends on the Gamma
value (Light/Dark Balance) of the original(s), as to how the tonal
balance plays out at this end. You may find that using ‘CURVES’ (COM-M)
on the originals to lighten shadows, and darken highlights slightly
before importing to layers will help in this. (You can ‘SAVE’ a curve
value, and ‘LOAD’ back into curves for reuse).
If you overdo any part of this, just backtrack through 'HISTORY' to get
back to the original unadjusted layer, it may be easier than
accumulating too many overlaid tweaks.
'FLATTEN' the layers.
Adjusting ‘BRIGHTNESS and CONTRAST’ also useful here after ‘FLATTEN
IMAGE’ eg: Brightness = -5. Contrast = +10.
Usually saved as a 'PICT' file for storage, JPEG copy only for Web use.
Recently, I have had some success with the FINAL flattened image by
using ‘CURVES’ and ‘AUTO’ which has produced a virtually monochrome base
image with good 3D separation (getting rid of the pink or violet cast).
You can see an example at:
http://www.creativegraphicarts.btinternet.co.uk/3D-Page37.html
....then click on the ‘STEREO’ icon. It WILL download, but is a large
image file!
So that is about all the truth of it so far, though I am constantly
working on the theory of the process, it is mostly the application of
art!
Some recommended sites for some stunning (real) 3D images are:
http://marsrovers.jpl.nasa.gov/gallery/3d/
and
tx0=Image&tx1=&col=mmg&qp=&qs=&qc=&ws=1&nh=12&lk=1&vf=0&ql=a&op0=%2B&fl0
=ContentType%3A&ty0=p&op1=%2B&fl1=category%3A&ty1=p&op2=%2B&fl2=showcase
%3A&ty2=p&tx2=SEMU775V9ED&showcase=Mars+Express" target="NewWindow">http://search.esa.int/queryIG.html?rf=3&searchType=general&tipo=Image&
tx0=Image&tx1=&col=mmg&qp=&qs=&qc=&ws=1&nh=12&lk=1&vf=0&ql=a&op0=%2B&fl0
=ContentType%3A&ty0=p&op1=%2B&fl1=category%3A&ty1=p&op2=%2B&fl2=showcase
%3A&ty2=p&tx2=SEMU775V9ED&showcase=Mars+Express
Hope this works for you. Let me know if any of the above points need
further clarification. Being able to see virtual models of one’s
engineering concepts in 3D space is for me the best! It makes all the
difference to design thinking, and is still technologically in the early
days of development for general access, though may well become the norm
for interactive play, and tv viewing in the future.
Regards,
Jan Kaliciak Creative Graphic Arts
www.creativegraphicarts.btinternet.co.uk /
EARLY VERSIONS OF STRATA Set up a new camera (lock the ‘roll’) and
compose your scene using all the camera attributes as required.
‘Replicate’ this camera in one plane only (X OR Z but not all), as seen
from the ‘Top’ view. You need to set the TWO cameras side by side, both
angling in slightly, pointing at the same centre spot of your object.
The angular separation will exaggerate the perspective so the cameras do
not want to be too far apart. It would help to identify them using a
custom colour texture, RED for LEFT, CYAN for RIGHT, so you know which
image-half you are processing later. Once you have set these attributes,
you could ‘Group’ the cameras together so if you have to move them about
they will not lose their settings.
Each camera image is rendered separately using ‘Raytracing’ at ‘Finest’,
‘Extra Smooth’, and ‘Expert: Pixel Detail Block Size = 1”. This ensures
better photorealistic image resolution, especially for bit maps. It is
also better to use pixel bitmaps four times larger than an object
measured in pixels, and scale to fit, to intensify detail.
You will be rendering each camera image separately in full colour, so it
is important to remember which camera produced which image when saving
the rendered result, so ‘Save As’: “Object123-L”,“Object123-R”, etc.
STRATA 3D CX This is much more elegant in operation. Set up your camera
as before, and now go to the ‘Object Properties’ Palette and change
Camera type from ‘Normal’ to ‘Stereo’. You will notice that you can also
set ‘Convergence Distance’ and ‘Eye Separation’ there. The focussing is,
in principle, the same as with the two cameras before, but this time you
drag out the camera focus point on screen to your object centre axis.
You could also match the convergence distance to the distance of camera
to object, but convergence does not seem to make a vast difference to
the result, the default seems OK. I have not covered all the focal
lengths that might be employed yet, and I suspect it will improve image
perspective by tweaking that carefully.
The eye separation is what is says, so it pays to be cautious here,
unless you wish to simulate the view that a horse, or larger mammal, may
have of the world! So far the default again has proved reliable enough,
but you could accurately measure your own eye separation in mm and use
that dimension.
Rendering options are exactly the same as before, but this will render
all in one, resulting in both Left and Right images being rendered in
the one file. There is a choice in the ‘Render Image’ window, and I have
‘Side by Side’ and ‘Left First’ selected here.
It is important not to have any pure white highlights, so keep the
lighting even if possible. Contrast can be increased in post-production
if it proves necessary.
POST-RENDERING PROCESSING This uses Adobe Photoshop (any version). Use
‘Open File’ to bring your Strata PICT files into the application.
On the RIGHT image (select right-hand half if a CX file), use ‘Com-A’
(Apple Key+A...But you knew that!), then ‘Com-C’, then ‘Com-N’, then
‘Com-V’.
That was ‘SELECT ALL’, ‘COPY’, ‘NEW File’, ‘PASTE’. You will notice that
it produces a new file exactly to the size of the copy. (Better never to
work on the originals!). Click on the ‘Layer’ name and type in ‘RIGHT’
(Press return).
Select the LEFT image and ‘COPY’, but this time ‘PASTE’ directly on top
of the new file, and name it ‘LEFT’.
On LAYER ONE (LEFT and RED), Select ‘HUE/SATURATION’, and ‘COLORIZE’
with HUE = 0, SATURATION = 100, and LIGHTNESS anything down to -50
(max). Now TURN OFF this layer with the eye icon.
‘SELECT’ Layer Two.
On LAYER TWO (RIGHT and CYAN), Select ‘HUE/SATURATION’, and ‘COLORIZE’
with HUE = (235?), SATURATION = 100, and LIGHTNESS anything down to -50
(max).
‘SELECT’ Layer One.
Then TURN ON Layer One again and set Layer Filter Attributes to
‘LIGHTEN’.
You should now see both images together! Using the cruciform ‘MOVE’ tool
tweak the alignment so there is perfect vertical alignment on eg: the
horizon line, and perfect lateral alignment on objects or edges nearest
to your viewpoint.
You can use the ‘CROP’ tool now, to trim off any misaligned raw edges.
Depending on the exact colour value of your glasses, you may need to
tweak each image now independently. This is the HARD part! The CYAN
value can be altered (235?) in Hue/Saturation again, when Layer One is
selected (maybe while turning off Layer Two), so that the whole Cyan
image looks as black as possible when viewed through your RED lens (this
is very obvious in practice, there is a clear cut-off point either side
of correct).
There is a lot of variation in lens spectral ranges, I have noted. Some
really smart plastic-lensed 3D glasses I had mailed from the USA behave
differently from my ordinary card mounted gels, so I imagine yours may
differ also.
Do the same for the RED. The LIGHTNESS value on each layer is crucial
here too! Anything down to -50 could work. It all depends on the Gamma
value (Light/Dark Balance) of the original(s), as to how the tonal
balance plays out at this end. You may find that using ‘CURVES’ (COM-M)
on the originals to lighten shadows, and darken highlights slightly
before importing to layers will help in this. (You can ‘SAVE’ a curve
value, and ‘LOAD’ back into curves for reuse).
If you overdo any part of this, just backtrack through 'HISTORY' to get
back to the original unadjusted layer, it may be easier than
accumulating too many overlaid tweaks.
'FLATTEN' the layers.
Adjusting ‘BRIGHTNESS and CONTRAST’ also useful here after ‘FLATTEN
IMAGE’ eg: Brightness = -5. Contrast = +10.
Usually saved as a 'PICT' file for storage, JPEG copy only for Web use.
Recently, I have had some success with the FINAL flattened image by
using ‘CURVES’ and ‘AUTO’ which has produced a virtually monochrome base
image with good 3D separation (getting rid of the pink or violet cast).
You can see an example at:
http://www.creativegraphicarts.btinternet.co.uk/3D-Page37.html
....then click on the ‘STEREO’ icon. It WILL download, but is a large
image file!
So that is about all the truth of it so far, though I am constantly
working on the theory of the process, it is mostly the application of
art!
Some recommended sites for some stunning (real) 3D images are:
http://marsrovers.jpl.nasa.gov/gallery/3d/
and
tx0=Image&tx1=&col=mmg&qp=&qs=&qc=&ws=1&nh=12&lk=1&vf=0&ql=a&op0=%2B&fl0
=ContentType%3A&ty0=p&op1=%2B&fl1=category%3A&ty1=p&op2=%2B&fl2=showcase
%3A&ty2=p&tx2=SEMU775V9ED&showcase=Mars+Express" target="NewWindow">http://search.esa.int/queryIG.html?rf=3&searchType=general&tipo=Image&
tx0=Image&tx1=&col=mmg&qp=&qs=&qc=&ws=1&nh=12&lk=1&vf=0&ql=a&op0=%2B&fl0
=ContentType%3A&ty0=p&op1=%2B&fl1=category%3A&ty1=p&op2=%2B&fl2=showcase
%3A&ty2=p&tx2=SEMU775V9ED&showcase=Mars+Express
Hope this works for you. Let me know if any of the above points need
further clarification. Being able to see virtual models of one’s
engineering concepts in 3D space is for me the best! It makes all the
difference to design thinking, and is still technologically in the early
days of development for general access, though may well become the norm
for interactive play, and tv viewing in the future.
Regards,
Jan Kaliciak Creative Graphic Arts
www.creativegraphicarts.btinternet.co.uk /
Posted: Thu - March 17, 2005 at 12:17 PM
Quick Links
Find
|
|
Categories
Calendar
| Sun | Mon | Tue | Wed | Thu | Fri | Sat |
More Fuzziness
Archives
XML/RSS Feed
Site Meter
iBlogger Web-Ring
Statistics
Total entries in this blog:
Total entries in this category:
Published On: Jan 19, 2006 02:04 PM
Total entries in this category:
Published On: Jan 19, 2006 02:04 PM