Re: poor anaglyph quality on the WEB

[P3D Jonathan Gross <catalyst@xxxxxxxxxxxx>]

It is possible to obtain high quality electronic stereoscopic images and
post them on the Web, but most authors do not seem to understand how
easily a good image can be ruined.

Primarily, you need to consider resolution, both spatial and chromatic.

Stereo images with spatial resolutions of less than 1024 X 7?? are
likely to be perceived as fuzzy.  Even 800 horizontal pixels is really
insufficient to support the parallax required for a high quality
stereoscopic image.  Especially important are the edges of features
which have only a small amount of parallax.  Also static images need
more resolution than moving images because the visual system is
apparently more forgiving of blur when it senses motion.

It is sometimes said that chromatic resolution can compensate for the
lack of spatial resolution.  If this were not so to some degree, then TV
would be totally unacceptable (the images, not content, which is another
issue).  While that may be true with monoscopic images, with
stereoscopic images you really need both spatial and chromatic
resolution to create the illusion of stereopsis.

Often, JPEG images are saved with very high compression ratios.  What
many people do not understand is that although JPEG maintains the
spatial dimensions of an image, the way it achieves high compression
ratios is by reducing the number of and subtle differences in the color
of the image.  Again, while that may be acceptable for monoscopic
images, subtle differences in color help the eye identify boundaries and
shading which are important to the quality of a stereoscopic image.

Also, successive generations of JPEG images degrade.  Lets say you start
with a high quality 1024X768X24 bit image pair.  When you save that
image with JPEG compression there will be a loss of quality that depends
on the amount of compression applied.  Now lets say that I retrieve that
pair of images, align them horizontally and vertically with my image
processing tools, and then I save it again using JPEG.  There will now
be a loss of quality that multiplies the first degradation.  

Getting a good image in the first place is quite a trick.  Film
generally has a much higher spatial resolution and dynamic range (light
to dark) than electronic images.  That is why it is important to scan
images at 36bits per pixel or more, even if the image will only be
stored with 24 bits of color.  After scanning, the image can be
adaptively squeezed into 24 bits in a way that preserves most of the
detail.  There are valid technological reasons for Kodaks Photo-CD
image format, and many of them have to do with preserving the details
caught by films broad dynamic range.

For a more specific explanation of scanning and why I scan 35mm images
at 2,700DPI with 36bits of color, see
http://www.skypoint.com/~catalyst/stereoscopic.html#services

There are solutions to the problems associated with representing and
displaying high quality stereoscopic images electronically, and most of
them were addressed by NEOTEK in the creation of the .neo stereoscopic
image format.  Just so there is no misunderstanding, I am NOT suggesting
that .neo become a standard for the stereoscopic community, but I am
suggesting that some other stereo image formats, while very clever, do
not directly address these issues.

Jon Gross
http://www.skypoint.com/~catalyst/stereoscopic.htm


------------------------------