Astronomer's Moon is difficult to believe (APOD 24 Aug 2007)

[astrogeek]

It is a common technique to freeze out the blurring and distortion due to Earth's atmosphere by taking many very short exposures and adding them together. However, I have a hard time believing that such a small telescope can capture such detail.

The theoretical limit of a telescope's angular detail is given by the Rayleigh criterion for diffraction. From any introductory physics or optics textbook:

angular resolution = 1.22 * (wavelength of light) / (diameter of telescope)

The answer comes out in radians. Using 550 nanometers for the wavelength (an estimate) and 230 mm for the telescope diameter (quoted on the image), I get 2.9x10^-6 radians. This converts to 0.6 arcseconds.

The image quotes a diameter for Ganymede of 1.7 arcseconds, only 3 times larger than the angular resolution. Even if the video capture can freeze out atmospheric blurring (of order 1 arcsecond) completely, the telescope is still limited by diffraction. The image shows far more detail than 3 resolution elements across the disk. I have a hard time believing that the image is real. It's more likely the "features" on the disk are noise amplified by the image processing method.

Anyone else have thoughts on this image?

[dodi]

If I look at the raw images of this video grabbed image (with hardly any significant detail on the planet Jupiter itself in a single raw frame) and subsequent stacks (each with clearly different markings on the surface) posted by the authors on Cloudy Nights forum, I really doubt most (if not all) of detail on the image.
A telescope of this size (even with firewire cameras - better than the usual webcam for video astronomy) will not be able to do more than perhaps a *subtle shading* on Ganymedes, which is not uncommon, but only in good to excellent condition without too much atmospheric turbulence. Most of the "details" seen in this particular image are most likely artefacts introduced by repetitive blurring and sharpening in image processing software.


Dodi

[Case]

2007-06-30 was a night of a Full Moon. Midsummer at 54° Latitude means it isn't really dark yet at 23:05 EEST (20:05 UT) for Minsk, Belarus.
Jupiter is in the South (Az 177°) at only 14° above the horizon.
At that time the Moon was also in the South (Az 148°) at about 2° above the horizon. Enough for an extra glow.

Hardly the best conditions for taking pictures of celestial objects.

[Wes]

I really doubt any of the detail shown is real, I have seen the Jupiter image that this was taken with and the detail on Jupiter is not at all on a level you would expect with the sub arc second detail on Ganymede.
Extreme enlargement as this image has and then subsequent smoothing can make artifacts look like real features

Wes

[astrogeek]

After posting the initial message, I wondered if Hubble had an image to compare. They do:

http://hubblesite.org/newscenter/archiv ... 5/image/c/

The detail in the Hubble image is nearly equivalent to the APOD image. Given Hubble is 10 times the diameter of the telescope in question and it is not looking through the atmosphere, I think we can safely call the APOD image a fake.

[ChrisO]

The camera specification, actually Fire-i 501b is the model number:
http://www.unibrain.com/download/pdfs/u ... -i_501.pdf

The CCD specification:
http://products.sel.sony.com/semi/PDF/ICX424AL.pdf

I have no astronomy experience, but this CCD seems more geared for industrial applications than for astronomy. Can any of you compare this CCD to those from cameras astronomers typically use? I am wondering if the dark current is giving noise that the image processing is amplifying. Since they were capturing at 30fps, you might have larger effects of dark current.

The camera does run in 8-bit and 12-bit modes, but I don't know if this just means that the internal ADC is 12-bits, so 8-bit mode gives you the upper bits anyways?

It would have been quite interesting to have a similar stack of frames taken in a dark spot of the sky near Ganymede that night, then apply the same image processing techniques. Also, it may be interesting to have done the whole process multiple times to see if you end up with similar results.

[stargazer_7000]

...guys, I believe Yuri at al. did not do that on purpose, which the term "fake" would imply.
I believe they just processed noise that was in their raws.
sure - trhe effort they took is great, and lots of their planetary images are really outstanding, take a look on their recent mercury image.
but concerning ganymed - it surely was an "over-enthusiastic" processing...

ps: here is the link to the mercury image FYI
http://www.cloudynights.com/ubbthreads/ ... 070728.jpg

BTW: what surprises me a bit, is, that the pros of NASA APOD did take this image as an APOD...there are other images of Yuri et al. being worth to be shown. I wonder why a professional astronomer does not realize the impossibility in that image...well, I guess we all are simply human...(-:
regards,
Dietmar
http://www.stargazer-observatory.com

[iamlucky13]

Remember, this is a composite of 400+ frames. There are relatively simple methods of using such redundant exposures to eliminate random noise. There is a more complicated process of resolving surface detail at a greater resolution than your telescope is capable of when you know with precise timing what portions of the surface are visible. This is how our current best pictures of Pluto were developed. The photographer here may have been doing something similar.

I don't know if it's possible to tease out as much detail as this APOD seems to show, but I thought I'd toss it out there.

[Case]

This is a composite of 400+ frames. There are relatively simple methods of using such redundant exposures to eliminate random noise.

One could select those 409 frames from the 10,000 to "match" the noise to the expected features, thereby enhancing the chance to see what one likes to see. If you make a specific selection from random data (noise), you will get something non-random that can match anything you'd like.

[NoelC]

What caught my eye was the sharpness of the edges of the satellite, where it meets black space. It's pretty round, and seems to have a relatively believable brightness falloff for a satellite with an atmosphere. It is likely wavelet and/or deconvolution processing parameters were chosen to accentuate this edge sharpness.

I doubt very seriously that anyone selected 409 frames by hand, but rather used software to statistically select the sharpest frames by comparison to the others. That's how Registax can be set to work, for example.

Regarding resolution of surface detail on a satellite of Jupiter... I see quite a bit of surface detail on Ganymede in the APOD of March 29. Certainly not in a completely different ballpark from that in the more recent APOD.

Personally I see no reason not to believe the detail shown on Ganymede is real.

P.S., Case, the image you point to is shown as a "Simulation" on CloudyNights.com. Have you any info how that image was created?

-Noel

[Case]

The image you point to is shown as a "Simulation" on CloudyNights.com. Have you any info how that image was created?

The linked image was posted for reference by Yuri on the board, which is linked from the APOD page. He didn't mention where it was from, but I do know that Starry Night Pro and Stellarium have simulated surfaces for planetary bodies when you zoom in on them. I suppose these apps use NASA imagery in combination with globe projections and phasing.

[OriEri]

What caught my eye was the sharpness of the edges of the satellite, where it meets black space. It's . . . It is likely wavelet and/or deconvolution processing parameters were chosen to accentuate this edge sharpness. . . . Personally I see no reason not to believe the detail shown on Ganymede is real.

That same wavelet and sharpening processing could generate artifacts of resolution higher than the diffraction limit of the instrument. I think you go out on a limb to say whether (or not) they are 'real' structures, meaning a higher angular resolution would reproduce them. There is a certain amount of information stored in the image limited by the MTF of the imaging system including the processing. While various processing can change the shape of that MTF, I do not think it is possible to add more information than is already there. As you use sharpening techniques to enhance some spatial frequencies of the image, you are going to take a hit at other spatial frequencies.

I agree that some sort of convolution processing had to be used to get that nice sharp edge, and I also believe that this same sharpening could also artifically sharpen real low frequency contrasts on the disk beyond reality.

[Nereid]

There's a discussion of this image in the BAUT astrophotography section.

The view, from folk who've actually done astrophotography with similar set-ups, seems to be that while the image is the result of some pretty heavy processing, it is not 'a fake'.