[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]
Re: [APML] OT: DSS image processed: Barnard 142/143 Complex("E" Nebula)
Hi Ray,
I've made two crops of a selected area of the image at full size, before and
after all of the processing:
http://www.pleiades-astrophoto.com/pixinsight/examples/B142_Crop_Before.jpg
http://www.pleiades-astrophoto.com/pixinsight/examples/B142_Crop_After.jpg
As you can see, there is no data loss. In fact, there is much more
perceptible data in the second image! Star colors are more differentiated:
you can easily view the redder stars surrounding the dust clouds. On the
other hand, brightness gradients on the clouds are also more defined, due to
applied noise suppression and large-scale processing methods. Therefore, if
you take into account only the *significant* data of the image, you'll see
that the applied processing has recovered a lot of data which was invisible
in the initial image. Really, do you think there are more things to see into
the original?
I think astronomical image processing has more things to do than enhancing
the data and simulating that the image has not been processed at the same
time. Astronomical image processing must say *something* to the spectator,
leading to some emotional stimulus inside him/her. Representing the data
collected at the telescope by just levels and curves adjustments doesn't
suffice in many occasions, IMO, since our eyes *won't notice* all the
existing data in the image.
Let me explain this with an example.
Here you have two images again. The object is M1, from the POSS II. The
first image is the original RGB combination with levels and midtones
adjustments. The second one is the same data processed basically with
wavelets. All the processing work has been made by Juan Conejero:
http://www.pleiades-astrophoto.com/pixinsight/examples/M1_JC_Original.jpg
http://www.pleiades-astrophoto.com/pixinsight/examples/M1_JC_Final.jpg
Download both images and save them on a directory of your hard disk. Then,
look at the images with the Windows' viewer, or some similar program
allowing you to change between both images quickly.
If you take a look at the first image, and then you press the PgDown key to
view the second one, you'll think: "Wow, the second image has a lot more
detail!". But try pressing PgDown again and look for a second time the
"unprocessed" image. Now you'll see more detail on the unprocessed image!
Your eyes are now, after viewing the processed image, identifying subtle
features near the noise level in the original image which were not being
perceived before processing. This demonstrates that many features in the
image must not only be perceptible to the spectator, but they must cause a
*visual impact*. This concept is tremendously important! Why do you need to
apply unsharp mask to an image?? Is this process bringing out details not
present in the original image??
As a pianist, I don't pursue achieving a naturalist's or reporter's work. I
try to achieve a personal interpretation of the object that appears on the
image, but, first of anything, RESPECTING the object and its nature.
At this point, processing technics are essential. Think on astrophotography
as a sum of forces. The sum of all of the forces is the visual impact. But
these forces are in a fragile equilibrium because they are opposing each
other: luminance vs. chrominance, small scale details vs. large scale ones,
color balance, static vs. dynamic objects, etc. The more perfected a
processing technique you have, the bigger visual impact you can give to the
image, since the forces will increase, but the center of gravity will stay
stable. Or maybe you want to do an unstable image intentionally, perhaps
enhancing only small scale detail!
But if you raise the visual impact of the image, noise will increase too!
Therefore it is absolutely necessary to make a noise reduction to increase
the visual impact of the image. Without noise reduction, you cannot exploit
all of the expressive resources from the objects. This means that after
processing there can be some areas on the image with no noise and also with
no small scale detail, but this doesn't mean that you are losing detail!
These are two different concepts!
Now see this third M1 image:
http://www.pleiades-astrophoto.com/pixinsight/examples/M1_JC_LargeScale.jpg
I've processed this one over Juan's result to enhance large scale features
of the object. If you raise the contrast of small details, why not to raise
the contrast of large ones?? I think this is a more complete processing,
since small scale and large scale features in the image cause different
emotions to the spectator. But this would be another independent thread! :-)
Well, that's all, for now.
Take care,
Vicent.
At 13:12 11/10/2004 +0100, you wrote:
> > Vicent Peris has processed one square degree of DSS data for the B 142
> > region, AKA the "E" nebula.
> > See it at:
> >
> > http://pleiades-astrophoto.com/pixinsight/DSS_Gallery/index.html
> >
> > I think Vicent's results are very impressive. Hope you like it too.
>
>It's impressive, and a great illustration of the functionality of the
>software...but as an aesthetic observation, maybe it's been over-smoothed
>in the wavelets/noise reduction department? It reminds me of "space-art",
>painted with an airbrush...to me it just lacks reality, somehow.
>
>There's a story, which this image inspires me to tell. A few years ago, I
>assisted a pilot research program of employing deconvolution and
>wavelet-based noise reduction on digital mammography x-ray images. We
>processed the images to "improve" them to our eyes, and spent a long time
>optimising this. But the litmus test was a set of clinical trials with
>radiologists at University College Hospital, Galway. The radiologists
>evaluated the test images under various headings. The tests were blind
>(the "before" and "after" images were presented randomly), and our team
>repeated them several weeks later to check for internal consistency
>(radiologists have excellent memories for image details: you have to allow
>a long time to pass before presenting them with the same set of data
>again, as otherwise they'll just remember their previous judgements).
>
>The upshot of all this was that what _we_ considered to be better images
>(with rather aggressive processing) was not what _they_ considered to be
>better. They were accustomed to seeing certain levels of noise, grain and
>high-frequency artefacts, and our first, over-processed versions threw
>them. They felt less secure in evaluating the possible presence of
>dangerous micro-calcifications, or whatever. So we scaled back our
>processing to a level that they were much happier with, and then there
>were genuine statistical improvements in their clinical diagnoses.
>
>The moral of the story? I think it might be that if one presents an
>over-processed astrophoto to someone who is very used to looking at
>astronomical photos/images, it will provoke an uncertain reaction. If one
>presents it to Joe Public, he'll probably think it is beautiful and admire
>its smooth noiselessness. There are cases when there is no option but to
>produce images which have that "airbrushed" look - it can be the only way
>to produce an acceptable image in the photon-starved X-ray domain, for
>example the Chandra satellite's images. But in the optical domain, it's
>really not my preference; when the processing is obvious and holds the
>eye's attention, it looks somewhat unnatural. The trick to image
>processing is to make the viewer think that there was no processing. It's
>really hard to do this, of course.
>
>Ray "who finds image noise familiar, maybe even comforting" Butler
>
>>Regards,
>>Juan
>>___________________________________________________________
>>Juan Conejero, Pleiades Astrophoto Team
>>juan.conejero__at__pleiades-astrophoto.com
>>PixInsight Home Page: http://pleiades-astrophoto.com/pixinsight/
>>___________________________________________________________
>>
>>_______________________________________________
>>Astro-Photo mailing list
>>Astro-Photo@seds.org
>>http://seds.org/mailman/listinfo/astro-photo
>
>--
>Dr. Ray Butler (ray.butler@nuigalway.ie || ray@physics.nuigalway.ie)
>Lecturer, Dept. of Physics || Computational Astrophysics Laboratory
>National University of Ireland, Galway, University Road, Galway, Ireland.
>Tel: +353-91-524411 ext. 3788 FAX: +353-91-525700
>
>_______________________________________________
>Astro-Photo mailing list
>Astro-Photo@seds.org
>http://seds.org/mailman/listinfo/astro-photo
Vicent Peris
http://www.cepheusastrophoto.com/Index.html
_______________________________________________
Astro-Photo mailing list
Astro-Photo@seds.org
http://seds.org/mailman/listinfo/astro-photo