[X] My Assistant

[tfisher]

I've been playing with the released raw ISS images (thanks to Bjorn for the img2png program!) trying to produce a better image near Huygen's descent location.

Here's what I've come up with. I started with images from two different imaging sequences, applied a Fourier bandpass filter, and stacked the images from the same pass. Because of different lighting conditions, or atmospheric conditions, or something, the images are significantly different, so I combined them as red and blue layers instead of just stacking further. Green is produced by the product of red and blue chanels, and a further contrast adjustment is also made.


I believe Huygens descent location is in the upper left corner. The correspondence with one of Rene's beautiful mosiacs is given here:


Unfortunately the geometry is different enough I can't really overlay the mosaic with the ISS image...

[tfisher]

Here's the correspondence to radar, as an animation.

[RNeuhaus]

Here's the correspondence to radar, as an animation.

I think the Hughens might be at the lower right corner, inside of circulars. These circulars might be the waves caused by the landing impact.

Rodolfo

[JRehling]

I think the Hughens might be at the lower right corner, inside of circulars. These circulars might be the waves caused by the landing impact.

Rodolfo

These images have a resolution of about 100 meters, so the circles would be at least 1 km in radius. The probe landed at 7 m/s and weighed about 50 kg on Titan. Now how the heck could something that weighs about as much a person, moving slower than a sprinter can run, land in mud and make visible circles 1 km in diameter?

[volcanopele]

I think the Hughens might be at the lower right corner, inside of circulars. These circulars might be the waves caused by the landing impact.

Rodolfo

?? First, not exactly sure what would have waves from the landing impact since it didn't land in liquid. Second, the images tfisher is showing (except the RADAR data) are from before the landing. Finally, those "circulars" are flatfield artifacts (how I hate those).

[RNeuhaus]

JRheling: Sorry! I have no tools (previous info and pictures) with me. What you told me is perfectally agreeable. I don't think a 50 kg slowing very slow as 7 m/sec can make waves of surface (the land is with ice and it is impossible to make waves). I didn't know about the resolution of the picture. After knowing it, I agree that it is impossible to make 1 km circle waves caused by a land surface impact. It might be see if the surface is water but it might be see with just few centimeters of height....and the water should be quiet before the splashing.

Volcanopele: I was not aware of the picture date. I tought it was taken after the landing. Sorry!

The circles pattern has attracted me the attention. Then I tought, as the probe was relatively hotter than the surface and when it landed on it and its heat might have propagated around it making the land softer until to some distance.

Rodolfo

[volcanopele]

That's okay. I just noticed that tfisher didn't mention which flyby he got that data from, Ta or Tb. Either way it was before the landing. Tfisher, what is the image numbers on the images you used.

[tfisher]

That's okay.  I just noticed that tfisher didn't mention which flyby he got that data from, Ta or Tb.  Either way it was before the landing.  Tfisher, what is the image numbers on the images you used.

The images for one sequence are n1477488603_2, n1477488635_2, and n1477488667_2 (from the Ta pass); for the other are n148162483_4 and n1481624423_4 (from the Tb pass).

[tfisher]

Here's an animation phasing between the Ta data and the Tb data a bit southeast of what I think is the Huygens landing site. I've marked one prominent fault line in yellow; you can see other fault lines coming up from the lower left corner and going across sort of parallel to the one I've marked. Notice how the brightness moves from one side of the marked fault line to the other.

I wonder if this is maybe a sunrise effect -- is the fault elevated, with one side shadowed from a certain sun angle? Does anyone know the direction and elevation of the sun at the Huygens site during the Ta and Tb passes?

[jmknapp]

  Does anyone know the direction and elevation of the sun at the Huygens site during the Ta and Tb passes?

Assuming Huygens is at longitude 169.0 E, latitude 10.6 S and not accounting for refraction:

Ta: 26OCT2004 15:28:20 UTC

sun elevation 57.6 degrees
sun azimuth 117.0 degrees E

Tb: 13DEC2004 11:36:40 UTC

sun elevation 56.1 degrees
sun azimuth 115.4 degrees E

Since the flybys were 47.84 days apart, and three Titan revs are 47.82 days, it's nearly the same geometry in each case.

[tfisher]

Since the flybys were 47.84 days apart, and three Titan revs are 47.82 days, it's nearly the same geometry in each case.

Hmm. That kind of blows my sun illumination theory. Too bad. What does that leave: low level haze?

[RPascal]

tfisher: Very interesting, and nice work! I have not considered this location up to now, but it looks plausible - including the cat scratches!

The images for one sequence are n1477488603_2, n1477488635_2, and n1477488667_2 (from the Ta pass); for the other are n148162483_4 and n1481624423_4 (from the Tb pass).

Where did you find these Ta and Tb raw images? I could not find them at the Cassini- or the ISS-website.

Sorry for me answering so rarely, I have very little time at the moment, but I am following this dicussion with great interest!

--René

[volcanopele]

Ta and Tb data are now available from the PDS.

For those interested in Titan processing, check out my LPSC abstract from March:

http://www.lpi.usra.edu/meetings/lpsc2005/pdf/2312.pdf

We have added a few steps since then (like actual photometry), but hopefully this will help.

[Phil Stooke]

tfisher - I hate to say it but I can't see any linear features at all in your image. And the fade suggested to me a less than perfect registration. So I must ask, how much confidence do you have that your line (1) actually exists, and (2) is correctly positioned? One problem to be aware of is that our eye/brain system is predisposed to see lines and patterns in complex textures - look at stucco, or even defocussed text on a page, and you will see lines. The more you look the more you see. I've been guilty of it myself.

Phil

[tfisher]

tfisher - I hate to say it but I can't see any linear features at all in your image.

Yeah, I worried quite a lot about the alignment, until I convinced myself there were enough things that lined up. Here is the location I see a line, with only images from one pass stacked, as a gif so no compression artifacts. Maybe "subtle" is a better word than "prominent", but I'm pretty sure I'm not completely imagining it.



It shows as dark against the lighter background above the central island or peninsula, and again as dark against the lighter island at the right. In between, it seems to still be there, though the graininess of the surroundings makes it harder to see.

[tfisher]

Where did you find these Ta and Tb raw images? I could not find them at the Cassini-  or the ISS-website.

The PDS data release can be searched through a form here, or if you know image numbers or just want to browse, you the data directories are here -- the at saturn imaging data is in the coiss_2007 and coiss_2008 directories.

[tfisher]

Ta and Tb data are now available from the PDS.

For those interested in Titan processing, check out my LPSC abstract from March:

http://www.lpi.usra.edu/meetings/lpsc2005/pdf/2312.pdf

Yes, that is interesting. You mention subtracting a "low-pass filtered" image for sharpening. By that do you mean a Fourier transform filter with only low frequencies allowed through? Can you say what pixel size you've been using for the wavelength cutoff?

[tfisher]

And the fade suggested to me a less than perfect registration. 

Hmm... Well, I was going to work out a markup to prove my registration, and ...
you're right. Its off. I'll be back with an improved version shortly.

(Feeling really dumb right now)

[tfisher]

Grrr. Here's a riddle for you. I still think I see a line, but I don't think its really there.

I took volcanopele's advice from the calibration thread, and tried to build a residual flatfield image as an average of bland images, in order to correct for artifacts not removed by the current calibration routines. Here's what I get as an approximation of that residual flatfield (only 3 images averaged, with some obvious noise removed, contrast stretched so you can see):



And with more contrast stretching around the area in question:



Playing with raw Cassini Titan images is a bumpy road!

[Bjorn Jonsson]

Yeah, I worried quite a lot about the alignment, until I convinced myself there were enough things that lined up.  Here is the location I see a line, with only images from one pass stacked, as a gif so no compression artifacts.  Maybe "subtle" is a better word than "prominent", but I'm pretty sure I'm not completely imagining it.

It shows as dark against the lighter background above the central island or peninsula, and again as dark against the lighter island at the right.  In between, it seems to still be there, though the graininess of the surroundings makes it harder to see.

I *think* I can see it but it could easily be due to inadequate flatfielding and not a real feature - as Jason mentions, additional flatfielding is required. There are linear features (*) in some of the images before flatfielding. If they are only 'partially' removed during flatfielding they will reappear during subsequent processing of the image.

(*) These linear features are most easily seen in BL1 images of Saturn.

[tfisher]

I've been working on improving my Titan processing methods. I've now got reasonable residual flatfielding and map projection based on the PDS data index information, and thanks to ImageJ's nice macro abilities the process of flattening, sharpening, and map projecting an image is pretty well automated. (which maybe means I should have bitten the bullet and used ISIS, but ah well...)

The "lines" turn out to be fully the result of residual flatfield artifacts, which unfortunately are quite persistant. The "sunrise effect" was a combination of residual flatfield and poor alignment.

I'll share with you the in-progress version of an improved super-resolution stack at the Huygen's landing site (at the moment using the same 5 images from Ta and Tb passes previously mentioned):



A question for volcanopele, if you're reading this: do you know how large the errors in the PDS latitude/longitude info can be? I've been trying to improve slightly on the registration by a small additional translation, but have been hoping error in size or rotation will be negligable.

[elakdawalla]

I've been working on improving my Titan processing methods.  I've now got reasonable residual flatfielding and map projection based on the PDS data index information, and thanks to ImageJ's nice macro abilities the process of flattening, sharpening, and map projecting an image is pretty well automated.  (which maybe means I should have bitten the bullet and used ISIS, but ah well...) 

You are actually using ImageJ for map projection? How??

--Emily

[tfisher]

You are actually using ImageJ for map projection?  How??

There is a plugin called TurboReg which will attempt automated registration of images or allows manual or scripted transformation. I'm currently using it in a script which uses TurboReg to perform a bilinear transformation moving the four corners of the image to lattitude and longitude coordinates based on the PDS index.tab information (really I've been grabbing it from the PDS search result page for each image individually -- the data at the bottom labelled "Image Parameters" is equivalent to a row from that table...) Really this isn't a true map projection: once the image is large enough curvature effects won't be handled properly. For the area I'm dealing with, though, of a few square degrees near the equator, the distortion is negligable.

On the odd chance you're interested in the script itself, here's the current form. (somewhat ugly, use at your own risk, etc...) It wants to start with a .png image open (produced by Bjorn's img2png and opened with the Image IO plugin) and to have the "Image Parameters" data saved as the same name as the image with a ".txt" extension.

CODE

// hardwire parameters
flatname = "flat-try4.tif";
filtername = "fft-filter-try1.tif";
dirname = "C:/stuff/cassini-work/"

map_width = 2000.0;
map_height= 2000.0;
map_lon1 = 195.0;
map_lon2 = 186.0;
map_lat1 = -6.0;
map_lat2 = -15.0;

// ------------------------- perform flattening and sharpening
imagename=getTitle();
extension=indexOf(imagename,".png");
if (extension>=0) basename=substring(imagename,0,extension);
else basename=imagename;
imageCalculator("Divide create 32-bit", imagename, flatname);

// new method: use a custom filter
run("Custom Filter...", "filter="+filtername);
rename(basename+"-flat");

// older method: subtract 85% of a low-pass filtered image
//rename("flat-temp");
//run("Duplicate...", "title=fft-temp");
//run("Bandpass Filter...", "filter_large=512 filter_small=85 suppress=None tolerance=5");
//run("Multiply...", "value=.85");
//imageCalculator("Difference create 32-bit", "flat-temp","fft-temp");
//rename(basename+"-flat");
//selectWindow("flat-temp");
//close();
//selectWindow("fft-temp");
//close();

// ------------------------ perform map transformation

metaname = basename + ".txt";
run("Edit...", "open="+dirname+metaname);
metainfo = getInfo();

y0 = getNumericTag("LOWER_LEFT_LATITUDE",metainfo);
x0 = getNumericTag("LOWER_LEFT_LONGITUDE",metainfo);
y1 = getNumericTag("LOWER_RIGHT_LATITUDE",metainfo);
x1 = getNumericTag("LOWER_RIGHT_LONGITUDE",metainfo);
y2 = getNumericTag("UPPER_LEFT_LATITUDE",metainfo);
x2 = getNumericTag("UPPER_LEFT_LONGITUDE",metainfo);
y3 = getNumericTag("UPPER_RIGHT_LATITUDE",metainfo);
x3 = getNumericTag("UPPER_RIGHT_LONGITUDE",metainfo);

run("TurboReg ", "-transform -window "+basename+"-flat "
    + toString(map_width) + " " + toString(map_height) + " "
    + "-bilinear " + "0 511 " + toString(map_lon(x0)) + " " + toString(map_lat(y0)) + " "
    +  "511 511 " + toString(map_lon(x1)) + " " + toString(map_lat(y1)) + " "
    + "0 0 " + toString(map_lon(x2)) + " " + toString(map_lat(y2)) + " "
    + "511 0 " + toString(map_lon(x3)) + " " + toString(map_lat(y3)) + " "
    + "-showOutput" );
rename(basename+"-flat-map");
setSlice(2);
run("Delete Slice");
selectWindow(basename+"-flat");
close();

return;

// --------------------- functions --------------------------

// return device coordinates from longitude
function map_lon(x) {
 return (map_lon1-x)*(map_width/(map_lon1-map_lon2));
}

// return device coordinates from latitude
function map_lat(y) {
 return (map_lat1-y)*(map_height/(map_lat1-map_lat2));
}

// This function returns the numeric value of the
// specified tag given in infostring. Returns NaN
// (not-a-number) if the tag is not found or it
// does not have a numeric value.
function getNumericTag(tag,infostring) {
 value = getTag(tag,infostring);
 if (value=="") return NaN;
 index3 = indexOf(value, "\\");
 if (index3>0)
   value = substring(value, 0, index3);
 value = 0 + value; // convert to number
 return value;
}

// This function returns the value of the specified
// tag given in infostring as a string. Returns ""
// if the tag is not found.
function getTag(tag,infostring) {
   index1 = indexOf(infostring, tag);
   if (index1==-1) return "";
   index1 = indexOf(infostring, ":", index1);
   if (index1==-1) return "";
   index2 = indexOf(infostring, "\n", index1);
   value = substring(infostring, index1+1, index2);
   return value;
}

[elakdawalla]

On the odd chance you're interested in the script itself, here's the current form.  (somewhat ugly, use at your own risk, etc...)  It wants to start with a .png image open (produced by Bjorn's img2png and opened with the Image IO plugin) and to have the "Image Parameters" data saved as the same name as the image with a ".txt" extension.[/code]

I am interested, thanks for posting the code. I've used ImageJ for simple stuff like making anaglyphs, color images, animations, and pulling out point spectra but not for any more complex image processing tasks or transformations. On my very long list of Web pages I'd like to write is a whole section explaining how to use freeware tools like ImageJ to do interesting image processing tasks on space image data...that task will be much easier thanks to all of the experts on this forum!

--Emily

[tfisher]

Further progress on the super-resolution composite:



This incorporates a weighted average of twenty-some images from Ta and Tb passes. These are flat-fielded, sharpened, and map-projected according to the metadata given in the PDS release.

This is improving closer to as good as I can do, but I think I can improve on edge smoothing, bad pixel removal, and micro-managing the registration. In another week or two maybe I'll have a final version.

My money is still on the landing location I indicated at the start of this thread. I think the further detail visible shows that to be a better candidate than the other two prominences also visable here.

[edit: tweaked the parameters in my composite program a bit, and updated the image: now the seams aren't so bad.]

[SigurRosFan]

Finally ... Exactly located!!

Great work, Travis!

Travis vs. René

Click for larger version (606.5 KB, not to scale):