Re: 256 web page

[P3D Jonathan Gross <catalyst@xxxxxxxxxxxx>]

> > The point being that it wouldn't matter how many images you put
> > on a page, they're all getting quantized down to the same 8-bit
> > palette.  At least this is how it works in Windows.  
> 
> That's why it was suggested that only one image per page might
> be beneficial.  The palette would be chosen for that picture rather
> than having the palette "chosen" to try and match a number of pictures.
> 
> Better matching would result.  Obviously, for a stereo pair, it'd make
> little difference if each image of the pair were separate because they'd
> both want the same palette.
> 
> Mike K.

Actually, I think there are other ways to avoid jagged palette.  The
problem is not so much with the publisher of the images, but with the
viewer.  Windows "tries" to reproduce colors well, but is constrained to
some extent by defaults designed for legacy (small) system
configurations.  There are two independent variables at work here (if
you dont consider the monitor calibration):  

1)  The format of the image posted on the web, and 
2)  the color depth of the Video Adapter in the viewers system.

The most common formats for images in HTML are GIF & JPEG.   The most
common depths for Video systems are 8, 16, and 24 or 32 bits.  16 bits
is dubbed "high color", and 24 or 32 bits are dubbed "true color".

Images stored in GIF (Graphic interchange format) are inherently 8 bit
(256 color palette), BUT the author of the image has some options
relative to  selecting  what those 256 colors are.  The choices are:

1)  uniform palette (equal numbers of colors across the color spectrum), 
2)  standard palette (predetermined by convention), or 
3)  adaptive palette (optimized for the particular colors in the
specific image). 

In addition, each of these choices may optionally use dithering. 
Dithering is a technique where small amounts of two or more colors are
displayed adjacent to each other in a way that causes the eye to "mix"
them psychologically into a third perceived color.  This is the
technique used by most paint-deposition color printers.  It works OK,
but results is a loss of spatial resolution.  Trading off spatial
resolution for chromatic resolution is an aesthetic choice heavily
influenced by the characteristics of the display device and individual
preferences.  Less I digress too far afield

GIF images should normally be produced with adaptive palettes
understanding that the color will NOT show up well on systems with only
256 color (8-bit) Video adapters.  The images will, however show up well
on systems with displays operating in 16bit or higher modes.  Here the
choice is really to accommodate the lowest common denominator, or to
provide better quality images (knowing that every month the cost of the
target systems continues to drop).  Guess you can tell where I am on
that issue.

JPEG (Joint Photographic Expert Group) formatted images are ALWAYS
adaptive and are stored with 24-bit of color for each pixel.  With JPEG
the chromatic error is a function of the level of compression that the
author selects.  Here, the trade-off is color veracity verses files
size, but NOT spatial resolution.  That means if you are willing to
accept a larger file size, you can cast a image with truer colors.  If
you want very small file sizes, then the spatial resolution remains the
same, but the viewers system is able to "reconstruct" the image with as
large a palette as possible for their particular system's capability.

With that explanation, it should be clear that photographic images
should almost always be posted on the net in JPEG format, and simple
graphic images (or where space and time are at a premium) should be in
GIF format with an optimized pallet.  If you publish with these
techniques, you give the viewer the maximum flexibility to view your
images either quickly, or with high quality, at their discretion.

OK, so what about putting only one image or set of images on an web
page?  You can do it that way, but you can also travel East but going
West, but for a much longer period of time

If the viewer has enough video memory in the video adapter's frame
buffer to display in true-color (24 or 32-bits/pixel), you can have as
many images on a page as you like because every color will be
represented to the best quality available by the viewing system (i.e.
there is no problem).

In high-color mode (16 bits) there still is a limitation of only 256
colors PER WINDOW, but each window can have its own optimized color
palette.  Solution:  If you want to see an image or set of images in the
best light possible (my family hates my puns), use your browser to put
the images in a separate window.  In Netscape for example, if you select
the graphic using the RIGHT mouse button, and use the "View Image"
option, you will see the image in a separate window with an optimized
pallet for that image (if you followed my advise above on how to post
them).

I hope all the background information was worthwhile.  I felt as if I
could not respond to the issue without describing my assumptions first. 
Here is my summary:

1)  When possible, post photographic images on the web in JPEG format. 
Even dramatic levels of image compression have a negligible effect on
perceived quality.
2)  If you can, use your browser on a system in true-color mode (24 or
32 bits per pixel); you then need to do nothing to see images at the
best quality possible for the particular trade-offs made by the author
of the image.
3)  If you can only use High-color mode (16 bits), view each image in a
separate window.
4)  Please come visit me if you make it to NSA97;  I will be happy to
continue this discussion and demonstrate the concepts on my computer.

Jon Gross


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

End of PHOTO-3D Digest 2135
***************************
***************************
 Trouble? Send e-mail to 
 wier@xxxxxxxxxxxxxxx 
 To unsubscribe select one of the following,
 place it in the BODY of a message and send it to:
 listserv@xxxxxxxxxxxxxxx 
   unsubscribe photo-3d
   unsubscribe sell-3d
   unsubscribe mc68hc11
   unsubscribe overland-trails
   unsubscribe icom
 ***************************