From: "Tim" <tjp314@pacbell.net>
Date: November 10, 2005 3:35:29 PM MST
To: Astro_IIDC@yahoogroups.com
Subject: Re: Stacking and Sharpness estimation.
Thanks for this, Milton! I need to go back and look at some of my better videos and see
whether I might be able to pick them off at a more sensible CI value.
Oh, and I've had the bimodal histograms a few times. I thought I might be losing my
mind, so it's good to know what that indicates (basically what I thought).
-Tim.
--- In Astro_IIDC@yahoogroups.com, Milton Aupperle <milton@o...> wrote:
Hi Folks;
One thing I'm working on is a pictorial guide to interpreting the
Histogram sharpness curves, what it means with image sharpness and
some idea as to how to set your cutoff value. Here is a quick verbal
run down of it. I don't have the images put together yet, but here's
more or less the text for it.
Note that in all the cases described below, the base assumption for
estimating sharpness is that some frames will be sharper than others.
If the image is out of focus and consistently blurry to start with,
it may just give you the "Type 1" or "Type 2" histograms. Also this
applies to stacking for Planetary, Lunar or Solar imaging where
sharpness is important. For DSO's or longer exposures you normally
stack just to suppress noise and sharpness of each is of much less
concern.
You get an indication of how consistent the frame sharpness is is by
looking at the Min - Max value range, the STD68 or STD96 standard
deviation values and how "skewed the curve is. The larger the Min -
Max range, the more variance in frame sharpness and the higher you
want to increase the Confidence Interval Slider. The Skew of the
curves is indicated by comparing at the Mean and Median values and if
they are the exact same, the histogram is not skewed. Visually a
skewed histogram has a long tail or long nose and the peak of the
hump is shifted to towards the low or the high end of the histogram.
The skew direction tells you whether you have more sharp frames than
unsharp frames. Generally you want to use a Confidence Interval that
selects (ie hilights in red) the sharpness bars that are to the right
of the central hump.
Basically you will encounter the following main types of histograms
(they will grade into each other):
1) Single bar histogram, which means all frames are almost the same
sharpness. That's virtually impossible because of turbulence and the
fact that image gain will also be interpreted to a small degree as an
indication of sharpness. In this case you can accept a very high (i.e.
100% Confidence Interval) of the frames, depending on how many frames
you want to use in your stack.
2) A single unskewed symmetric "hump" with a narrow range of
sharpness values and a small STD96 value of probably 0.05 and lower.
This is the curve you get on good imaging nights or if you take lots
of sample under consistent skies. With this you should select a
Confidence Interval percentage that is to the right of the central
hump in around the 50 to 75% range, depending on how many frames you
want to use in your stack.
3) Skewed hump with longer left tails or longer right noses. The
direction of skewing indicates whether you have a larger percentage
of sharp frames or a larger percentage of unsharp frames. If the
Median is greater than the Mean value then it's skewed to the right
(has a longer left tail) and this means it has more sharper frames
than unsharp frames. If it's skewed to the left, then it has more
unsharp frames than sharp frames. With this type of curve you want to
be more restrictive in how many frames are used and probably want to
go with a 65 to 80% Confidence Interval. If it's skewed to the right,
I tend to increase the Confidence Interval value and if it's skewed
to the left, I decrease it.
4) The double or "bimodal" hump which may have almost no "Skew" at
all, high STD96 value > 0.1 and a large range in the Min Max range.
It means means you have a percentage of really blurred frames (lower
left hump) and a percentage sharper frames (upper right hump). In
this case you'll want to use CI percentage that selects frames that
are to the right of the upper right hump, usually that means it's in
the 70 to 80% range.
5) Undulating hump, with multiple peaks, wide range of values and
very wide STD96 value. I generally throw these out as just "bad" and
don't process them.
Hope that helps someone.
Milton J. Aupperle
President
ASC - Aupperle Services and Contracting
Mac Software (Drivers, Components and Application) Specialist
#1005 - 815 14th Avenue. S.W.
Calgary Alberta Canada T2R0N5
1-(403)-229-9456
milton@o...
www.outcastsoft.com