I've been holding off weighing in on this subject area because I'm not
expert but I may be able to shed some light. I just ran some experiments and
I put the images into a directory on our website:
. Our webmaster helped out by converting all the
tifs to jpegs for viewing (uh-oh). I've asked him to convert them back, but
that may take some time (tomorrow?). By the way, this image was created by
our dry film process called Digital PIC. The latent image was scanned
without processing, an interesting technology. You can check out our Digital
ICE, Digital ROC and Digital GEM products as well as Digital PIC at
www.asf.com.
There are several components to jpeg. One, the image is converted from RGB
to luminance and chrominance components (b&w information, color information)
we refer to it as y + u + v. You can keep the full resolution of y, u and v
called 4,4,4. Or compress (I think it just decimates) u and v two to one,
4,2,2. Or compress u and v four to one, 4,1,1. Then you can add hoffman
compression. So far even at 4,1,1 with hoffman you have very few or no
_noticeable_ artifacts. From there you add additional compression. I'm not
sure what this is called, but this is where the artifacts start showing up.
The jpeg compression program divides the image into 16x16 bit cells. That is
why the artifacts look like little squares all over the image and why thing
like text start looking goofy (a technical term).
Now, how Photoshop implements the standard is the question. I use Photoshop
5 and it uses a scale of 0 to 10. Photoshop 6 uses a scale of 0-12. Exactly
which compression techniques come into play at which level I don't know. I
do know that if you take a tif image, save it out as a jpeg at level 10 in
Photoshop 5, close both images, open them both again and do an apply image
of the two with difference the pixel values I found were 0,0,0 or a 1,1,1 a
few 0,0,2 (1 out of 255 that is what, 0.39% and that is at only one pixel
location most had no change, 0,0,0) on the difference between the two
images. If you do an adjust levels and put the arrows around the
information, you can see what detail is lost, pretty interesting, mainly
noise.
When I saved a 1267 by 1974 pixel image as a tif it was 7,335KB. As a
compressed tif, 4,814KB. As a jpeg the first time at photoshop quality 10,
2,475KB, second time 2,475KB, third time 2,476KB, fourth time 2,477KB. Each
time I saved it I checked the difference between the images. There were the
same small differences at first, but there was no difference between the
third and fourth image. Saving it at level 0, the file was 131KB and the
difference between it and the tiff were as high as 71 out of 255.
One other point. The image I started with was a 16 bit image. I had to
convert to 8 bits before I could start since jpeg can't handle 16 bit images
(another reason to use another format). It was 18,803KB as a compressed (LZW
lossless) 16 bit image. It was 14,662 as an uncompressed 16 bit image. I'm
not sure why the file increased in size when I compressed it.
Your results will vary depending on the image you use. I hope this data is
useful. Your conclusions will vary depending on your needs obviously.
Good luck.
Jack Phipps
Applied Science Fiction
On Mon, 2 Apr 2001 14:37:09 +0100 Alan Tyson
(AlanTyson@bknowl.freeserve.co.uk) wrote:
> Is it perhaps that jpeg is specially suited to
> 'photographic' images, and not to areas of single solid
> colours with sharp edges like Henk's image? The latter is
> rare in photographic images, and lossless gif does an
> excellent compression job on that sort of thing anyway.
