Author Topic: Objects at or beyond the limits of resolution  (Read 7645 times)

gsilsby

  • Newbie
  • *
  • Posts: 2
    • View Profile
Objects at or beyond the limits of resolution
« on: September 27, 2012, 12:53:16 am »
Need another identification option or two: Object is at the limit of resolution, object is beyond the limit of resolution.  Some of the objects are most likely to be galaxys but are so pixelated that no further information can be deduced.  This would tag those objects for longer exposure in the future if someone cared.

Graham Silsby

JeanTate

  • OotD posters
  • Hero Member
  • *****
  • Posts: 2976
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #1 on: September 27, 2012, 06:58:13 am »
Welcome to the Zoo, gsilsby!  :)

Would I be right in guessing that you're referring to the CANDELS (Hubble IJH) images, in the new Zoo? For example, AGZ0000drv (or perhaps even fainter):



If so, why not use the existing hashtag #fhb (faint hubble blob)?

As I understand it, the automated photometric pipeline - which is what gets to 'see' all these images first - has some very powerful routines in it, and if it says there's an object there, there really is. Of course, while it has the distilled wisdom and experience of thousands of astronomers and millions of images built into it, it's still quite dumb; for example, it sometimes thinks that things which we mere humans can tell - at a glance - are artifacts are worthy of further study  ::)  :P

What's intriguing about these 'at or beyond resolution' objects is that zooites, collectively, can 'see' their morphology (shape) really well. Well, they we (I'm a zooite too!) often can, but certainly not always.

And it's that collective pattern recognition ability which the astronomers who built the Zoo are hoping to tap into. So, just give it your best shot; call the classification as you see it ...

Budgieye

  • Hero Member
  • *****
  • Posts: 3860
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #2 on: September 27, 2012, 09:55:50 am »
These objects are probably half way across the universe.
We'll mop them up by Galaxy Zoo 42 (We're on 4 just now)

In the meantime, just classify it as best you can, maybe smooth and in between.

zutopian

  • Hero Member
  • *****
  • Posts: 2852
  • REOPEN this forum!!!
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #3 on: September 27, 2012, 10:48:20 am »
« Last Edit: September 27, 2012, 10:51:07 am by zutopian »

zutopian

  • Hero Member
  • *****
  • Posts: 2852
  • REOPEN this forum!!!
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #4 on: October 08, 2012, 01:52:36 am »
I suggest, that they should improve the image quality. It seems, that it is possible. I altered an image, which I posted.:
http://www.galaxyzooforum.org/index.php?topic=280414.msg617436#new

zutopian

  • Hero Member
  • *****
  • Posts: 2852
  • REOPEN this forum!!!
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #5 on: October 08, 2012, 02:25:26 am »
I altered another image. It is the one from the discussion in TALK and the blog post.:

Original

http://talk.galaxyzoo.org/objects/AGZ00004kg

Attachment:
Altered version



« Last Edit: October 08, 2012, 06:00:18 am by zutopian »

Freethesouls

  • Newbie
  • *
  • Posts: 22
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #6 on: October 11, 2012, 12:46:31 am »
I altered another image. It is the one from the discussion in TALK and the blog post.:

Original

http://talk.galaxyzoo.org/objects/AGZ00004kg

Attachment:
Altered version

Altered image may take up much time to do because these "FHB" or Faint Hubble Blob galaxies are download to the computer before the astronomers on the science team sees them, however, it seems like a good idea to try to make them clearer as possible. It would make sense to make them easier to classify

Tom Zolotor

zutopian

  • Hero Member
  • *****
  • Posts: 2852
  • REOPEN this forum!!!
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #7 on: October 22, 2012, 03:24:32 pm »
Original image:
http://talk.galaxyzoo.org/objects/AGZ0000xtc/discussions/DGZ10068kq

In the attachment there is the altered version by me.:
« Last Edit: October 22, 2012, 03:28:43 pm by zutopian »

Brooke

  • Astronomers
  • Newbie
  • *
  • Posts: 39
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #8 on: October 23, 2012, 02:51:39 pm »
Sorry, but I don't really see these as improvements. First of all, you can no longer see the noise in the form of the colored speckles in the dark parts of the image. But seeing the noise in the image is good for the science.

Say you have an image where the sky is totally black, with no noise at all:


How do you know you're seeing all the detail you can see if you can't see a little bit of the noise to know for sure what level of brightness is all noise and "zero" signal?
If you tune the image so that it displays just a bit of the noise, that allows you to judge:

Some possible features show up here; they're faint, but you can see them, and you can see the noise level well enough to judge whether you think they're real.

Of course, if you turn up the brightness even further, you might see even more, but it gets less and less helpful. Eventually you also lose the detail in the center of the galaxy:


It's like trying to adjust a photo from your camera taken at night. When you turn the brightness up so you can see the details of what the flash didn't capture, you start to get some noise in the image, and at some point you know you're seeing everything there is to see. If you keep going, the parts that the flash did illuminate get washed out. For a glossy photo you put up on your wall, maybe you want to have a perfectly black sky. But for the science images, you don't always want that.

The GZ team went through lots of discussion, in this case with the CANDELS team, on how to tune the parameters of the image "just right" so the images show enough of the noise to help users judge what's real and what's not, but aren't so bright that your ability to see features in the center get lost. The scaling decided on may not be perfect in every case, but for this example I actually think it's very good.

Second, in the altered image it looks to me like the image has been blurred -- so there's a bit of information lost compared to the original image. Smoothing is a technique that certainly has some excellent scientific applications, but I don't think it's obvious that it's an improvement here. I do appreciate that some of these galaxies are faint and small and thus appear pixellated, but just do your best with the classification and remember that the reason it looks that way is that it's billions of light-years away!

zutopian

  • Hero Member
  • *****
  • Posts: 2852
  • REOPEN this forum!!!
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #9 on: October 23, 2012, 09:34:49 pm »
Thank you for your reply!
There are some Candles images, which are pixelated, but others are not. There are Candles images, which don't have any background noise. So I think, that your argument concerning the value of background noise seems to be questionable. I don't understand, why there are images of different quality? There are images of the same object (but different image center), which have different qualities.: pixelated, with background noise, without background noise.: http://www.galaxyzooforum.org/index.php?topic=280414.0
Kevin had told in a TALK discussion, that the other object is a smooth galaxy, but I wasn't able to see a smooth galaxy. After altering the image, it looks to me indeed like a smooth galaxy! That's an obvious improvement in my opinion. I think, that pixelated images means poor image quality. It is like playing "Guess that object?", when trying to classify pixelated images.
Maybe someone else (zooite or GZ astronomer) is able to alter the pixelated images better than I did! I would be pleased about that.

« Last Edit: October 25, 2012, 07:02:06 am by zutopian »

JeanTate

  • OotD posters
  • Hero Member
  • *****
  • Posts: 2976
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #10 on: October 24, 2012, 05:41:31 pm »
Wish I'd read this thread before posting this:(
Quote from: JeanTate
I think that there's automatic image processing done, starting with the cleaned Hubble data, to produce the images presented in GZ4, for us to classify.

That processing automatically scales the image we are presented according to the 'size' of the object to be classified; the image scaling is done in the same way, for all objects. As a result, images of small objects become highly pixelated.

It's certainly possible to produce less pixelated images, from the same cleaned Hubble data; however, I suspect (I certainly do not know!) that the experimental design (for the GZ4 classifications) requires images to be processed and presented in the same way. Does that make sense? Who knows? Certainly not ordinary zooites!

zutopian

  • Hero Member
  • *****
  • Posts: 2852
  • REOPEN this forum!!!
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #11 on: October 25, 2012, 12:50:16 pm »
Here is an original GZ image, which is pixelated and has a blank background.:


http://talk.galaxyzoo.org/objects/AGZ000068f
« Last Edit: October 25, 2012, 02:04:33 pm by zutopian »

zutopian

  • Hero Member
  • *****
  • Posts: 2852
  • REOPEN this forum!!!
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #12 on: October 25, 2012, 02:02:39 pm »
(...)
I do appreciate that some of these galaxies are faint and small and thus appear pixellated, but just do your best with the classification and remember that the reason it looks that way is that it's billions of light-years away!

I had posted a pixelated image in Talk and a science team member had said #star.
If so, why is it pixelated, even though stars and not as far away as galaxies ?  ???
http://talk.galaxyzoo.org/objects/AGZ00003kb
« Last Edit: October 25, 2012, 02:07:15 pm by zutopian »

Brooke

  • Astronomers
  • Newbie
  • *
  • Posts: 39
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #13 on: October 25, 2012, 04:47:35 pm »
How a color image of a galaxy appears depends on a few things.

From a ground-based telescope, almost everything will be smoothed and blurred by the smearing effect of the atmosphere (an effect similar to that in zutopian's altered images).

From a space-based telescope like Hubble, that doesn't exist, so you're left with only the smearing caused by the instrument itself, which you can see the effect of by looking at an image of a star. Even Hubble can't resolve a star like the one posted about above, but they still don't appear as a single pixel -- they have a size. Some galaxies (in SDSS as well as in Hubble) are not much bigger than the appearance of a star; we'd call those barely resolved, which is not a precise scientific term but it conveys the point.

Also, the way a star is smeared (this is called the PSF of an image) can vary from image to image, because the images were taken under different conditions. Even from Hubble, the PSF can change slightly between filters and exposures because the telescope might be rotated differently from one exposure to the other, or slightly shifted (often deliberately). And the PSFs for different instruments on Hubble are different too.

The color Hubble images from CANDELS and used in the new GZ are made from three filters on two different instruments, so because the PSFs between colors don't perfectly match, a star can look a bit strange. Even if you take care to match the PSFs between filters in the post-processing of images, it is not perfect. And because a galaxy image is smeared out in the same way that the star is, that effect translates to galaxies too. These galaxies are so faint and small compared to the size of the pixels (and also compared to many SDSS galaxies, which are often hundreds of pixels in size, so there's no need to zoom in as far to see them) that the slight differences that remain even after PSF-matching can show up pretty clearly and make the image appear pixellated.

But, basically, the reason I don't think it's an obvious improvement to use a slightly blurred/smoothed version of the CANDELS images is that a) one of the key strengths of Hubble is that it's above the blurring effect of the atmosphere, so I'm not necessarily keen to take part of that away, and b) the first question we ask is how smooth a galaxy is, and if we smooth it before we even ask, we're potentially biasing that answer. I suppose we'll see what the classifications show...

Brooke

  • Astronomers
  • Newbie
  • *
  • Posts: 39
    • View Profile
Re: Objects at or beyond the limits of resolution
« Reply #14 on: October 25, 2012, 05:09:19 pm »
Here is an original GZ image, which is pixelated and has a blank background.:


http://talk.galaxyzoo.org/objects/AGZ000068f

Yes, there are two things going on here, besides what I mentioned above.

One (pixel size scale): the object is small on the sky, so it's zoomed in such that the original Hubble pixels are much bigger than the pixels on your monitor.

Two (brightness scale): the brightness that is called zero in that image is brighter than the noise level, so the background looks blank. The brightness scaling varies like the pixel size scaling for the same reason (you can't use the same setting for the bright and/or big galaxies as the faint and/or small ones and still have useful images for each).

Because of the number of galaxies in CANDELS both kinds of scaling were done automatically instead of individually for every single galaxy. We did check a lot of them -- I personally looked at well over 1,000 and I'm not the only one who spot-checked -- so the scaling works pretty well most of the time, but there are examples where one or both are a bit off. Both these oddities are more likely to show up on the forum or on Talk just because they're weird enough to be remarked upon, but mostly it works well.