(continuing a discussion hosted on the RealVIZ forums)
Color Shifting Issue
(images and text generously contributed by Jim Scott)
 |
 |
(1)
CubicFaceDemo.jpg |
(2)
VRScreenDump.jpg |
IMAGES 1 & 2 are examples of a PhotoShop image (cubic face) & the same area of the VR (screen dump) that show actual matching colors! These examples relate to the discussion of matching source image colors with the resultant VR.
The key here was that the dynamic range of the source images did not overwhelm the camera's native capability. 180 degrees from this view, in the SAME cubic, this was not the case and it took prolonged color correction to get the VR to match the original images' color balance. There were large windows in this "180 degree" view in the same VR that completely threw the color for a "loop" because the dynamic range was seriously exceeded. This example was a good test piece because all other variables were constant -- same camera, lens, software processing, etc. Only the dynamic range changed.
This is not to say the color always matches perfectly in a VR when the dynamic range of the source images is within the camera's range, but there is not the washed out, funky, disappointing differential that often occurs when the dynamic range is exceeded.
Dynamic Range Discussion
(images and text generously contributed by Jim Scott)
 |
 |
 |
(3)
AgVineYd.jpg |
(4)
AgVineYdOpt.jpg |
(5)
AgVineYdOptSHO.jpg |
Bonus images for you, as a faithful participant in the
Stitcher forum! ;->
These are HDR image examples from my experiments with the Optipix (by
Reindeer
Graphics -- review)
& SHO (by Applied
Science Fiction) plug-ins:
#3 is what the camera would have taken as the "correct"
exposure.
#4 is the resultant image when processed using just the
Optipix filter.
#5 is the same image also run through the Optipix/SHO
plugin.
#4 is the sum of all (9) exposures -- the "bi-cone"
-- an equal number of
lighter and darker exposures on either side of the "correct"
exposure.
With #5 I used a darker weighting for the initial Optipix
processing, which
means the final image came out on the dark side (more on this below).
When
processed with the SHO plug-in it brought up the image's dark tones, and
a
more pleasing result for the eye.
There is no other processing (such as sharpening) used here -- just a
straight dynamic range comparison -- which, to me at least, clearly
demonstrates the value of an HDR approach to taking VRs.
I think you can reliably get the images required for HDR down to 4 or
5 from
the apparent HDR standard of 9-10. But these 4-5 images still need to
run
the much of the gamut of the exposure possibilites -- i.e. if 9 images
represents the full exposure range with #1 being the
darkest; #5 being the
camera's "correct" exposure; and #9 being the
lightest -- I would in
general pick #1,2,4,5,7 for the Optipix plug-in
to process, with #7 used
twice (this formula was used for #5 -- AgVineYdOptSHO.jpg).
This skewed use of exposures results in an underexposed image -- but using
the SHO filter the dark quarter tones are brought up nicely -- with better
results than using the average of all nine images with the Optipix plug-in
by itself... and far fewer images required to boot.