Re: [APML] Process this Raw Image

[<pleiades2004@pleiades-astrophoto.com>]

Hi Chuck,

> This is certainly a noiseless image. The available dynamic range is
largely
> unused: all the information is hold by a small segment between 0.0007 and
> 0.0124 in the normalized [0,1] range. This is also quite oftenly found in
> CCD astroimages. However, this image has virtually infinite SNR regardless
> of dynamic range usage, which is not precisely what happens with
astrophotos
> of any kind.

>I want to make sure we're not confused here. The image is only 10 bits
>so I would think you'd want to normalize over 1024 levels. That would
>make it take up a larger portion of the available dynamic range. It's
>true that none of the channels take up the entire range.

Thanks! No confusion. The numbers I've given are *normalized*, so they are
independent on bit depth. In the program I've used, histograms are
internally generated with 16-bit resolution, then the user can choose an
arbitrary output resolution for representation purposes (plotting values are
interpolated from the original 16-bit histogram). However, a normalized real
range from 0 to 1 is used to specify the values of control data (background
and highlights clipping values and midtones balance for a simple levels
adjustment, for example). This effectively gives complete independence on
bit depth.

To illustrate this, here is the histogram of your raw image plotted with
1000 discrete values and zoomed 64x horizontally:

http://pleiades-astrophoto.com/tmp/pi-test-images/raw-1000levels64x.jpg

The small triangles enclose the effective dynamic range, which once
normalized is of course [0.0007, 0.0124] again. This means that there are
only 12 discrete levels with nonzero pixel counts in the range [0,1000]. In
the 16-bit range there *should* be 767 nonzero levels. Here is a portion of
the raw histogram plotted with the full 16-bit resolution and horizontally
zoomed 400 times:

http://pleiades-astrophoto.com/tmp/pi-test-images/raw-16bit400x.jpg

I guess this file has been generated and/or processed with Photoshop. I know
this because every odd level in the 16-bit histogram has a zero pixel count,
which is shown as alternating gaps in the graphic (note that there are very
small dots for each odd level just at the x-axis, indicating zero pixels --- 
zoom the figure 2x to see them). For performance reasons (?) Adobe Photoshop
works with 15 pixels, not 16, in order to handle signed pixel sample values.
Existing information at the odd levels is simply lost; it cannot be kept by
redistribution of those pixels within even levels, since this would cause a
dramatic change in the image. So your raw image is described by, roughly,
384 unique nonzero levels.

Take care,

Juan

P.S.: 2048x4000 ... with the same SNR ??? Wow! These images must be
absolutely stunning!

-- 
-------------------------------------------------
Juan Conejero, Pleiades Astrophoto
juan.conejero_at_pleiades-astrophoto.com
http://pleiades-astrophoto.com/
-------------------------------------------------

> Here is my try:
>
> FITS format, 16-bit unsigned integers (1.23 MB):
> http://pleiades-astrophoto.com/tmp/pi-test-images/200406161520-pi.fit
>
> JPEG format (316 KB):
> http://pleiades-astrophoto.com/tmp/pi-test-images/200406161520-pi.jpg
>
> Processed with the latest beta of our PixInsight. Only a single
> histograms
> transform and a wavelet transform (à trous algorithm with deringing).
>
> This is the raw 16-bit histogram shown with 64x horizontal
> magnification.
> The stretch points are enclosing the nonzero segment. The histogram is
> filled with gray because of gaps, i.e. unused steps:
> http://pleiades-astrophoto.com/tmp/pi-test-images/raw-16bit-
> histogram.jpg
>
> And this is the whole histogram after processing, shown in the 12-bit
> range
> (no gaps at all):
> http://pleiades-astrophoto.com/tmp/pi-test-images/final-12bit-
> histogram.jpg
>
> An illustrative example.
>
> Regards,
>
> Juan
>
> -- 
> -------------------------------------------------
> Juan Conejero, Pleiades Astrophoto
> juan.conejero_at_pleiades-astrophoto.com
> http://pleiades-astrophoto.com/
> -------------------------------------------------


_______________________________________________
Astro-Photo mailing list
Astro-Photo@seds.org
http://seds.org/mailman/listinfo/astro-photo