[Lowfer] Watering hole lowfers
JD
listread at lwca.org
Tue Mar 5 20:51:14 EST 2013
>>> What is your technique of getting those HUGE capture graphics down to
>>> the 30 Kb ?
Magic! :-)
Actually, it's a combination of techniques along with a helpful graphics
program called LView. LView Pro 2006 is one I bought to be able to process
images better in 32-bit Windows, but the older 1995 shareware version, LView
Pro 1.B/16, is actually easier to use for compressing graphics of this sort.
The first thing I do to keep final file sizes down is something you might
not expect: I let Argo save everything in Windows BMP format. If you save a
picture in JPG, then process it in some way and run it through JPEG
compression again, the artifacts arising from the first compression show up
as added "detail" (really, noise) in the image, and each subsequent
compression tries to account for them as if they were legitimate detail.
This runs the size up a little, and degrades image quality as well.
Therefore, I keep everything in uncompressed form until the very end.
When prepping a capture for posting, my next step is usually to crop heavily
to get rid of any detail on screen that isn't needed to tell the story.
Once in a while, you may have seen some images where I omitted large parts
of the frequency scale of an Argo capture, for instance. Solid blocks of
color or white space don't count so badly against you in JPEG compression,
but lots of adjacent high-contrast pixels do, so I generally eliminate as
many extraneous details as I safely can.
(Sometimes I even do a separate version of the image in which I can soften
the focus, then use MS Paint to help me replace less important areas of the
waterfall with the softened version, while leaving the portion containing
the trace at its original sharpness. It helps to double the image sizes
first, do the replacement, then shrink back to the original size. The
resulting reduction of spatial frequency response means a lot of the
Discrete Cosine Transform coefficients become insignificant and don't have
to be accounted for, enabling the compressed file to be smaller. This
softening and replacement can be a time consuming process, though, so I
don't do it very often.)
The final trick is to set the JPEG I/O quality option of the LView software
for greater compression/lower image fidelity. Doesn't sound pleasant, but
if you do it only one time as the very last step, the damage is minimized.
LView uses a scale of 20 to 95 for image quality, in steps of one unit, with
a default of 75. (Many JPEG graphics programs have such a setting, though
some may use different scales for degree of compression; while others only
offer a few quality grades.) I keep the original bitmap open, and re-save
it in JPG format with progressively lower quality settings until I achieve
the desired result. Depending on the size of the image and the amount of
detail in it, I generally only have to go down to 50-60 units to make a
typical capture end up at 32 kB or less. Sometimes I have to go much
farther, and other times (very small images) I can leave it at the default
setting.
The latest image is one where I wanted to show large areas around the
desired traces, including spurious signals, to illustrate the effect the
increasing static levels had on everything simultaneously. That precluded
doing much more cropping, or any softening of detail at all. The only
helpful manipulation was a bit of gamma correction to darken the background,
as I had left Argo's sensitivity relatively high in AGC mode. That not only
normalized the appearance, it also reduced the difference between adjoining
pixels in less bright areas of the image, enhancing compression efficiency
slightly. However, that still didn't allow a whole lot of reduction, so
this time it was mainly brute force--the first time I've ever had to go all
the way down to 20 units on the compression quality setting. But since the
image had remained uncompressed in all steps prior to the end, I think it
didn't turn out too badly.
Hope that helps answer the question without being too boring.
John
More information about the Lowfer
mailing list