[X] My Assistant

[hubdel11]

the DIRS have posted 2 new mosaics on their site :
http://www.lpl.arizona.edu/%7Ekholso/Titan...3K_big_ster.jpg
http://www.lpl.arizona.edu/%7Ekholso/Titan...Apr15_gnom_.png

[dvandorn]

the DIRS have posted 2 new mosaics on their site :
http://www.lpl.arizona.edu/%7Ekholso/Titan...3K_big_ster.jpg
http://www.lpl.arizona.edu/%7Ekholso/Titan...Apr15_gnom_.png

I guess I've just been spoiled by the MERs and MGS, but I am disappointed in the quality of the images from Huygens. Why the very limited number of pixels per image -- was it a matter of the width of the comm pipe between Huygens and Cassini? Would it have cost that much more in terms of a weight penalty to build five or ten times more pixels into the camera system? Even in the late 80s, when Huygens was first developed, they could build higher-resolution cameras than they ended up using...

I know there was fog and haze (more than expected), I know we lost half of the images... but the image encoding from the source has introduced so many artifacts it's really hard to analyze the images to anywhere near their theoretical resolution.

Yes, there is a great deal of good information in the images and other data Huygens returned... but it's so tantalizingly close and yet so lacking in critical definition that I just want to tear my hair out... *sigh*...

-the other Doug

[Decepticon]

It didn't help that the Earth testing images where SHARP and pixless.

I saw the link before but can't find it now.

[deglr6328]

I guess I've just been spoiled by the MERs and MGS, but I am disappointed in the quality of the images from Huygens.  Why the very limited number of pixels per image -- was it a matter of the width of the comm pipe between Huygens and Cassini?  Would it have cost that much more in terms of a weight penalty to build five or ten times more pixels into the camera system?  Even in the late 80s, when Huygens was first developed, they could build higher-resolution cameras than they ended up using...

Yes it was the com link. Because Huygens was descending on a parachute, twisting and swaying in the wind, it was of course not possible to use a high-gain directional antenna and because it was a small probe it only had so much of a power budget to make RF transmissions with. So Huygens was restricted to a mere 10 watts of transmitter power at 8,192 bits per second. Thats just 1 KBps !! I think condsidering the amount of other data they had to transmit (something like 5 other science instrument packages with several individual instruments per package with each one producing data) the images we got back were really really good. The DISR team knew the bandwidth they had to work with and they squeezed every last drop of data through it that they could.

This problem of a trade off between directionality and power got me thinking though. Maybe someone here more knowledgeable can correct me. If we're ever to increase the science returns from these types of missions there must be a way around this problem somehow. Optical transmission is out of the question right away obviously because of the even higer limit on pointing accuracy and attenuation prblems associated with the atmosphere. But what about a phased array transmitter? The problem with using directional radio transmitters to increase the signal/noise ratio on a decending atmospheric probe is obvious - conventionally, you'd need to use a dish to concentrate the beam in a particualr diretion (just like cassini's high gain antenna) and you'd need to continually re-point this dish as you're falling and turning under the parachute. You would lose track of where to keep pointed after just a few seconds of this. But what if you had a transmitter on the orbiting data recieving spacecraft that sent a pure tone to the falling probe and a small fixed position directional reciever antenna (juat a whip antenna) on the probe? It would be easy to determine at least roughly where the signal was coming from as you were moving and rotating by simply analysing the strength of the received tone and when this information is coupled to a phased array transmitter you could continually re-point the radio beam in this direction instantly, without moving any physical antenna. Phased array (digitally controlled) techniques are fairly new, I wonder, has this ever been considered before?

Incidentally there are reports on Huygens data return that I simply do not understand. For instance it was stated in a press release that 474 Mbits of data was returned by huygens over a 3.75 Hr. time. Now, every single place where I have seen it listed, the data rate from huygens is said to be 8192 kiloBITS/sec (fixed rate). Even figuring generously at a 4 hour mission that only gives you a maximum of 14,400 seconds to work with for a max of 14 MBytes (or 112 MBits) of data so where this 474 number comes from I have no clue. Once again ESA is no help in figuring this one out as they've published several conflicting values of total data returned by Huygens.

[Phil Stooke]

I know nothing about this, so I'm only playing with numbers... but if we say 8000 bits/sec times 10000 sec we have 80 Mbits tansmitted.. at 6:1 compression we would have 480 Mbits when decompressed. (note your units are mixed up!) ESA could be more forthcoming but it's not too difficult to manipulate the numbers to make a reasonable guess. The images look much more than 6:1 compressed, and other data might be a lot less, but it could average about 6:1.

Phil

[Guest_BruceMoomaw_*]

Incidentally there are reports on Huygens data return that I simply do not understand. For instance it was stated in a press release that 474 Mbits of data was returned by huygens over a 3.75 Hr. time. Now, every single place where I have seen it listed, the data rate from huygens is said to be 8192 kiloBITS/sec (fixed rate). Even figuring generously at a 4 hour mission that only gives you a maximum of 14,400 seconds to work with for a max of 14 MBITS of data so where this 474 number comes from I have no clue. Once again ESA is no help in figuring this one out as they've published several conflicting values of total data returned by Huygens.

This has to do with the fact that this was actually Huygens' maximum data return rate -- initially it returned its data at several lower bit rates. It may take me a while to dig up the place where I saw the details on this (one of the ESA's web documents, I think).

[Decepticon]

Looking at the above images will they eventually clean up the lines and color the data?

The above mosaic looks very raw.

[deglr6328]

I know nothing about this, so I'm only playing with numbers... but if we say 8000 bits/sec times 10000 sec we have 80 Mbits tansmitted.. at 6:1 compression we would have 480 Mbits when decompressed.  (note your units are mixed up!)  ESA could be more forthcoming but it's not too difficult to manipulate the numbers to make a reasonable guess.  The images look much more than 6:1 compressed, and other data might be a lot less, but it could average about 6:1. 

Phil

oops! I of course meant 8,192 BITS/sec for huygens data rate not 8,192 KBits/sec!!

[alan]

It didn't help that the Earth testing images where SHARP and pixless.

I saw the link before but can't find it now.

Test images here
http://www.lpl.arizona.edu/~kholso/test_images.htm
The lowest altitude image from Huygens has so many artifacts because the frames in the upper upper part of the image had to be stretched so far just to show a few details. If they weren't stretched so much the images would look more like this.

[tedstryk]

In response to Phil's comment, that trick was used often by Galileo. Additionally, it is often said that Galileo returned 14,000 images. They leave out that the bulk of those were from the Earth/Moon flybies where it could transmit at a relatively high data rate. Also, for the rest of the images, 14,000 images includes scientifically useless opnavs, images so compressed that they are unrecognizable (often to find targets in the image to trasmit certain portions at a higher rate - which was then counted as another image). Very few images sent by Galileo from Jupiter are full 800x800 frames, and even those are compressed. So, while not technically fabricated (that is about the number of Galileo SSI EDRs you will find on the PDS), it is a bit Enron-ish.

[edstrick]

Note also that the lowest altitude frames from Huygens are mostly of nearly featureless dark plains/channel material. There almost isn't anything there to see.... without far more signal/noise than we ended up with after fiber optics relayed the image to the CCD, it was read out and the data was compressed.

Sigh.

[tty]

This problem of a trade off between directionality and power got me thinking though. Maybe someone here more knowledgeable can correct me. If we're ever to increase the science returns from these types of missions there must be a way around this problem somehow. Optical transmission is out of the question right away obviously because of the even higer limit on pointing accuracy and attenuation prblems associated with the atmosphere. But what about a phased array transmitter? The problem with using directional radio transmitters to increase the signal/noise ratio on a decending atmospheric probe is obvious - conventionally, you'd need to use a dish to concentrate the beam in a particualr diretion (just like cassini's high gain antenna) and you'd need to continually re-point this dish as you're falling and turning under the parachute. You would lose track of where to keep pointed after just a few seconds of this. But what if you had a transmitter on the orbiting data recieving spacecraft that sent a pure tone to the falling probe and a small fixed position directional reciever antenna (juat a whip antenna) on the probe? It would be easy to determine at least roughly where the signal was coming from as you were moving and rotating by simply analysing the strength of the received tone and when this information is coupled to a phased array transmitter you could continually re-point the radio beam in this direction instantly, without moving any physical antenna. Phased array (digitally controlled) techniques are fairly new, I wonder, has this ever been considered before?

I’m afraid a phased array is probably not a viable solution. A phased array is just that – an array of transmitting elements which steer the beam electronically by controlled interference between the transmitters. This means:
1. a fairly large number of transmitters is required
2. the complete antenna must be large in relation to the wavelength of the transmitted signal
3. the beam can only be steered within a half-hemisphere (theoretically that is, in practice a great deal less)

Typically this means that a phased array needs four separate antennas for all-round coverage (have a look at an AN/SPY-1 radar for an example). I suppose it might be implemented in a box-shaped probe (if this is possible from an aerodynamic point of view). The sides would also have to slope quite a bit if you need to have coverage near zenith. The whole arrangement (including the steering electronics) would be rather heavy and bulky though.

Incidentally phased arrays are not a new concept, but it was quite difficult to implement it effectively until the advent of solid state transmit/receive modules.

Now that we know that a landing on Titan is survivable I suggest that a better solution would be to store an uncompressed high-definition version of the data on-board and transmit it after landing, either through a high-gain antenna or over a longer period of time (which probably implies a radioisotope power source). You could still transmit a low-res version while descending in case the landing fails.

tty

[tty]

I might add that there is a method to increase the resolution of images taken after landing that would be fairly simple to implement. If it is possible to change the direction of the camera slightly between pictures (one or two pixels is enough) it is possible to calculate a composite image with considerably better resolution than the original ones. Unfortunately Huygens was apparently rock-steady after landing. Perhaps a small pyrotechnic charge that gave the probe a nudge after say 3 images on the surface had been transmitted would have been a good idea?

tty

[Phil Stooke]

tty mentions slight movement of Huygens to increase resolution in the landed images. Unfortunately there are lots of reasons this would not have been feasible.

This method is called super-resolution at JPL and (basically the same thing) drizzling at STScI. It was invented at NASA Ames. It involves having multiple views of a scene with small (subpixel) offsets. A new image is created by enlarging all the images by a factor of 2 or 3, registering them to the nearest new pixel in the enlarged image and combining. The increase in resolution goes with the square root of the number of images: 4 images, 2x resolution. In principle... actually it's not really as good as that, but it it still useful as it slightly increases resulution and also SNR.

Tim Parker at JPL does it in Photoshop, and having told me how he does it, I have too. Excellent results, for instance, on the Voyager images of saturnian satellites which were taken in multispectral sequences. Remove noise, enlarge the images, contrast enhance each one, sharpen each one individually, then register and merge.

The trouble with tty's idea is... are... (1) Huygens was not designed to survive landing, (2) extra weight and complexity, (3) you really need a few more than 2 images to help very much, so you would need several sequential charges.

Phil

[Phil Stooke]

Here is an example of the results achievable with super-resolution, in case you have not seen this kind of comparison before. It's part of a multispectral sequence of Dione from Voyager 1, taken as Dione transits across Saturn (the grey background). The color version is well known. There is very little color variation across Dione, making the sequence ideal for super-res (Io would be useless!)

The right image is one original frame, enlarged 400% to show pixel size. The left is a composite of the entire multispectral sequence, weighted to give more prominence to the better images. The black dot (a reseau mark on the original - thankfully NASA quit using them after Voyager as they served very little purpose... a leftover of the days when images were archived on film) has been removed with a patch from another image in the sequence. I should add that other special processing to remove artifacts has also been done.

I put this here because it follows from the Huygens thread but obviously it could be in icy moons as well, or instead.




Phil

[centsworth_II]

I see a lot of talk about how much image data could be transmitted from Huygens, but weren't the cameras designed to do a lot more than take images? To do spectrometry and particle studies? Is this being factored into the data stream? I am pleased by the Huygens images. A lot of the credit goes to the Titan landing area itself for being such an interesting place. But Huygens captured that area wonderfully. I do wish there was some method -- mirrors or panning ability to get a wider view after landing. But over all a great imaging accomplishment.

[tedstryk]

I see a lot of talk about how much image data could be transmitted from Huygens, but weren't the cameras designed to do a lot more than take images? To do spectrometry and particle studies? Is this being factored into the data stream? I am pleased by the Huygens images. A lot of the credit goes to the Titan landing area itself for being such an interesting place. But Huygens captured that area wonderfully. I do wish there was some method -- mirrors or panning ability to get a wider view after landing. But over all a great imaging accomplishment.

I have mixed feelings. On one hand, I think the DISR did a great job. But I wish that we had the Channel A images, which, when stacked on the Channel B images, could, in addition to gapfill, be used to reduce compression artifacts (in other words, if you have multiple images of an area at different heights and angles, it is easier to tell whether something near the limit of discernability is real or not). Any chance there will be any recovery via the VLBI data? I doubt it, but then again I thought they would never get anything out of the Galileo I24 2x2 binned image data, and I was wrong.

[centsworth_II]

I have mixed feelings.  On one hand, I think the DISR did a great job.  But I wish that we had the Channel A images...

It sure would have been great to have the channel A images. But the difference would be one of degree and not kind. There may have been some better resolution and more coverage with the additional images, but the essential information about types of features present on that area of Titan would remain the same. In other words, the additional images would not double the visual information we have about the landing area. Maybe they would add 10 or 20%, to the quality, and no substantial new information. At least that's the way I look at it to feel better.

[deglr6328]

I have mixed feelings.  On one hand, I think the DISR did a great job.  But I wish that we had the Channel A images, which, when stacked on the Channel B images, could, in addition to gapfill, be used to reduce compression artifacts (in other words, if you have multiple images of an area at different heights and angles, it is easier to tell whether something near the limit of discernability is real or not).  Any chance there will be any recovery via the VLBI data?  I doubt it, but then again I thought they would never get anything out of the Galileo I24 2x2 binned image data, and I was wrong.

Can you say more about the "Galileo 124 2X2binned image data" recovery via VLB aperture synthesis? I've never heard anything about this before and I'd like to add it to the Wikipedia article!

[Guest_BruceMoomaw_*]

One question-and-answer page on the DISR site ( http://www.lpl.arizona.edu/%7Ekholso/expert_bushroe.htm ), unfortunately, confirms that none of Channel A's data is recoverable -- Huygens' telemetry signal as received on Earth was so weak that no attempt was even made to lock onto any more than its carrier signal.

As for the DISR experiment's other data: all of its substantial amount of spectral data -- very important scientifically -- nevertheless took up so little bit space compared to the images that all of it was duplicated on both channels, and so none of it was lost whatsoever. However, the only release of any of it so far has been two graphs of the near-IR reflectance spectrum of the surface taken during the last 100 meters before landing, when the searchlight was illuminating Titan's surface in all wavelengths (as compared to sunlight, which only punches through the methane and haze in a few limited spectral windows, which is why the searchlight was carried in the first place). By far the better of the two can be seen at http://sci.esa.int/science-e/www/object/in...fobjectid=36529 -- it shows water ice mixed with a complex mixture of organics whose exact nature hasn't been announced yet. (As with most of Huygens' science return, the experimenters are being very close-mouthed until the "Nature" issue with Huygens' findings is published some time this month.)

One thing that was lost -- simply due to the unexpected violence with which Titan's stratospheric wind turbulence tossed the probe around -- was most of the attempted observations of the width of the solar aureole, which would have told us about the size of the haze particles. Also, Huygens shocked everyone by spinning the WRONG DIRECTION for most of its descent (maybe due to a jammed parachute-line swivel?), which by itself blocked most of the azimuthal data from the DISR Sun sensor -- making it much harder to interpret the precise direction in which the DISR instruments were pointed at any time. However, the latter is being slowly pieced together indirectly.

[scalbers]

the DIRS have posted 2 new mosaics on their site :
http://www.lpl.arizona.edu/%7Ekholso/Titan...3K_big_ster.jpg
http://www.lpl.arizona.edu/%7Ekholso/Titan...Apr15_gnom_.png

Glad to see the DISR team taking the time to get the geometry right. In addition to the "raw" version
shown in these and more recent releases, I'd like to see a version that is further filtered by locallly varying amounts depending on the amount (and type) of noise locally present. I think it would be more
eyecatching and informative. Add to that the completeness of Rene Pascal's image (filling in more of the holes in a single mosaic, with flat fielding, etc.) and we'd have a version that to me would look very intriguing. So, perhaps the best is yet to come, both on the amateur and professional sides

[dilo]

Here is an example of the results achievable with super-resolution, in case you have not seen this kind of comparison before.  It's part of a multispectral sequence of Dione from Voyager 1, taken as Dione transits across Saturn (the grey background).  The color version is well known.  There is very little color variation across Dione, making the sequence ideal for super-res (Io would be useless!)

The right image is one original frame, enlarged 400% to show pixel size.  The left is a composite of the entire multispectral sequence, weighted to give more prominence to the better images.  The black dot (a reseau mark on the original - thankfully NASA quit using them after Voyager as they served very little purpose... a leftover of the days when images were archived on film) has been removed with a patch from another image in the sequence.  I should add that other special processing to remove artifacts has also been done.

I put this here because it follows from the Huygens thread but obviously it could be in icy moons as well, or instead.




Phil

Very interesting, Phil!
I knew this technique, and I tried to do something like this on the post-landing Huygens pictures, without success.
In fact, there isn't offset at all between images (even at sub-pixel level, I think); moreover, compression artifacts are really very lossy and hard to remove...!
Anyway, I'm courious about method you are talking; in particular, how to register various images with sub-pixel offset?
I think isn't exactly trivial without aid of some "autoregression" software...
Moreover, you mentioned the "noise" removal of each image, followed by enlargement, contrast enhancement and then sharpening. In my mind, noise reduction and sharpening goes in almost exactly opposite directions...so I'm little bit disappointed!
Finally, how much contrast enhancement is needed? Is something like "automatic" PS function?
Sorry for the my pletora of questions, but I'm really interested to this method!!!
Thanks.

[Phil Stooke]

Dilo - only the worst noise is removed, plus other artifacts such as a faint pattern of vertical one column wide stripes, or occasionally a missing line of pixels. So this is not enough to negate the later sharpening. So - first the image is cleaned up as much as necessary in that way. The images are then enlarged by some integer factor, 2 or 3 or 4, using some higher order interpolation. The registration is done by limb fitting (on satellite images) or other major feature fitting on other types of image - matching the centers of the blurred limbs in the enlarged images.
They are then stacked and merged.

As for contrast enhancement... I was generating images which showed the best combination of large and small features (sharp craters and broad albedo markings) for mapping purposes. True photometric fidelity is lost in these images but features of all scales are shown at once. Each separate image was optimally enhanced before making the composite. I devised a method in which a single image was duplicated 4 or 5 times, and each version manipulated differently - different contrast settings and high pass filtering at different scales to optimize features of different sizes - and had its artifacts corrected or removed (e,g. limb artifacts in the high pass filter versions). These different versions were then merged. The Dione image is a typical result.

Phil

[dilo]

Dilo - only the worst noise is removed, plus other artifacts such as a faint pattern of vertical one column wide stripes, or occasionally a missing line of pixels.  So this is not enough to negate the later sharpening.  So - first the image is cleaned up as much as necessary in that way.  The images are then enlarged by some integer factor, 2 or 3 or 4, using some higher order interpolation.  The registration is done by limb fitting (on satellite images) or other major feature fitting on other types of image - matching the centers of the blurred limbs in the enlarged images. 
They are then stacked and merged.

As for contrast enhancement... I was generating images which showed the best combination of large and small features (sharp craters and broad albedo markings) for mapping purposes.  True photometric fidelity is lost in these images but features of all scales are shown at once.  Each separate image was optimally enhanced before making the composite.  I devised a method in which a single image was duplicated 4 or 5 times, and each version manipulated differently - different contrast settings and high pass filtering at different scales to optimize  features of different sizes - and had its artifacts corrected or removed (e,g. limb artifacts in the high pass filter versions).  These different versions were then merged.  The Dione image is a typical result.

Phil

Thanks Phil, I will try asap to use these suggestions...
PS: when you add different versions of the same image with different contrast/filtering, do you utilize different weight in the final sum?
Bye.

[alan]

Three new DISR mosiacs posted
http://www.lpl.arizona.edu/%7Ekholso/

[Decepticon]

Why are the channel areas sharper than the rest of the images?

Will they eventually clean up this Pixel soup?




Another Question will cassinni make Radar Observations of the Huygens landing site?

What type of resolution can we expect from the radar?

[dvandorn]

Three new DISR mosiacs posted
http://www.lpl.arizona.edu/%7Ekholso/

You know, it looks a little like Huygens might have landed on one of the "pebbly"-looking sandbar-like stretches that extend out from the brighter, obviously higher "islands" in the channel areas.

In other words, maybe Huygens didn't land in the channel itself, it landed in a sandbar. So, it's possible that lower, deeper areas of the channels are in fact filled with liquid. We just landed a meter (or less) above the "waterline" on a sandbar...

-the other Doug

[maycm]

This is an alternative location for the new mosaics, but with a nicer layout and explanations and the final mosaic of the landing area.

[David]

I still have not seen any attempted explanation of the peculiar shapes of the "sand bars" -- particularly, the arrow-shaped formation pointing to about 2:00, and the trailing shapes in its wake. The smooth curves suggest shaping by fluids, but I find it hard to imagine a specific pattern of flows that would produce these shapes. Are there any similar formations on earth?

[TheChemist]

From the DISR public website :

"Check out the Low Altitude Stereographic (.jpg)
and Low Altitude Gnomic (.png) projections. "

http://www.lpl.arizona.edu/%7Ekholso/

[gpurcell]

From the DISR public website :

"Check out the Low Altitude Stereographic (.jpg)
and Low Altitude Gnomic (.png) projections. "

http://www.lpl.arizona.edu/%7Ekholso/

We really got lucky that the best resolution pictures DISR took are the ones that show that shoreline.....

[volcanopele]

We really got lucky that the best resolution pictures DISR took are the ones that show that shoreline.....

Here are some posts on my blog related to these images:

http://volcanopele.blogspot.com/2005/05/ne...sr-mosaics.html
http://volcanopele.blogspot.com/2005/05/mo...sr-mosaics.html
http://volcanopele.blogspot.com/2005/05/fi...ns-surface.html

The three mosaics discussed in the second post have since been taken down but I have reposted them.

[GregM]

.

[Bob Shaw]

Hmmm... ...'Titan' - that was Chris Huygens' new band, after he split with the Encke Division, wasn't it?

Apropos of Titanian music, apart from 'Titan' by Wendy Carlos (Digital Moonscapes), can anyone think of any other music with an appropriate name (don't say Gustav Holst!) to listen to while surfing the latest from Cassini?

Oh, and W Carlos (if you don't know why she's better known as W Carlos, then don't ask!) also wrote pieces for Iapetus and Rhea. And she's an amateur astronomer (with an interest in harpsichord music, just to make the outer planetary connection even more real!) - one of her eclipse photos made it onto APOD.

[Phil Stooke]

Bob wrote:
Apropos of Titanian music, apart from 'Titan' by Wendy Carlos (Digital Moonscapes), can anyone think of any other music with an appropriate name (don't say Gustav Holst!) to listen to while surfing the latest from Cassini?

Al Stewart's "Sirens of Titan" on his Time Passages album from 1978. The song refers to Kurt Vonnegut's book, of course.

Phil

[Gsnorgathon]

...
Apropos of Titanian music, apart from 'Titan' by Wendy Carlos (Digital Moonscapes), can anyone think of any other music with an appropriate name (don't say Gustav Holst!) to listen to while surfing the latest from Cassini?
...

Not necessarily apropos, but this line from the Human League's Black Hit of Space comes to mind:

I reached for the tonearm, which was less than one micron long but weighed more than Saturn, and time stood still...

[alan]

In the box with the image release it says:
"More images will be released following our Nature publication this month."
When is hte Nature publication coming out?

[volcanopele]

In the box with the image release it says:
"More images will be released following our Nature publication this month."
When is hte Nature publication coming out?

Not this month...

EDIT: okay, I know not helpful.