Lunar animations: Aligning images
in [DSP]
|
From: Rune Allnor on 26 Dec 2009 10:57 Hi all. The past few days the moon has been unusually visible (that is, visible for an unusally long consecutive time), so that the change in phase is clearly visible over the course of a few hours. The idea occured to me that it ought to be possible to take a number of images of the moon, with a few hours interval, and assemble some sort of animation of the lunar cycle from these images. For this to work, one would need to align the individual images to some reference image frame. My telescope is of the 'terrestial' type, which means the tripod is oriented with respect to the horizon, not the earth's axis of rotation, which causes all kinds of alignment problems. Since such a sequence of images by necessity must be taken over the course of considerable time (at least hours and days, preferably weeks, probably months and years), one can not rely on fixed references or constant system settings. So there will be a considerable amount of fiddling required to align the images. What I have in mind is something like: 1) Take the image 2) Adjust brightness 3) Apply a Canny edge detector 4) Use the edge image to find translation/scale/rotation parameters relative to an overall reference frame 5) Scale/rotate/translate original image according to parameters from step 4) 6) Crop/resample the original image to a reference resolution 7) Insert into to existing sequence of images 8) (Re)generate animation The difficult steps here are items 4) and 5), first finding scale and rotation parameters and then scaling and rotating each individual image reative to the reference. Do anyone know how to do these kinds of things? I do *not* have access to any image processing software packages. Since there are less than 40 clear-weather winter nights per year and the moon is below the horizon in summer, most of the fun of this project would be to implement the algorithms... Rune
From: ImageAnalyst on 26 Dec 2009 11:23 Rune: I've uploaded an animated gif of the moon's lunation (what you described) to http://drop.io/lunation# (It's an animated gif - you'll need to download it for it to "play.") It's very cool. You'll see the size change and the wobble in addition to the shadow moving. There is also a lower quality YouTube movie of lunation at http://www.youtube.com/watch?v=COfoBAwRPHk I'm sure they used an algorithm like you described. I suppose that if the moon goes from full to full in 28 days and full to new in 14 days then the shadowline moves across the face of the moon in 14 days and would move 1/(14*2) of the distance across the moon in a 12 hour period so perhaps you might notice the shadow move in a single night (or day). Might I suggest moving to Tucson, Arizona or the big island of Hawaii for finishing your experiment? They'll have warmer, clearer weather and they're both very nice places to visit or live. Tucson, especially, has a huge optics and astronomy community. Maybe a web search for "lunation math equations" might turn up some algorithms, or try sci.astro newsgroup, or try someone at the University of Arizona's astronomy department (http:// www.as.arizona.edu/).
From: dbd on 26 Dec 2009 16:39 On Dec 26, 7:57 am, Rune Allnor <all...(a)tele.ntnu.no> wrote: > Hi all. > > ... > The idea occured to me that it ought to be possible to take a number > of images of the moon, with a few hours interval, and assemble some > sort of animation of the lunar cycle from these images. > ... > > Do anyone know how to do these kinds of things? I do *not* have > access to any image processing software packages. Since there > are less than 40 clear-weather winter nights per year and the moon > is below the horizon in summer, most of the fun of this project would > be to implement the algorithms... > > Rune Modern panorama stitching software can be used to perform most of the required operations. Output is usually available in the form of multilayered unblended formats that could be reassembled as animations. The stitching software would generate control points, scale, rotate, lens correct, align and resample the images. Output is available in unblended multilayer formats I know how I would accomplish this with the tool I have used to stitch a few thousand panoramas: PTGui, but this is a commercial product. I believe the same capabilities can be found in freeware products. Examples include: Stitching Hugin http://hugin.sourceforge.net/ Panorama Tools http://www.all-in-one.ee/~dersch/ Editors http://www.irfanview.com/ http://www.gimp.org/ Of course, this wouldn't allow you to be as hands-on with the algorithms. On the other hand, the stitchers above are open source and could be used as a source of algorithms. Dale B. Dalrymple http://dbdimages.com
From: Vladimir Vassilevsky on 27 Dec 2009 02:42 Rune Allnor wrote: > The idea occured to me that it ought to be possible to take a number > of images of the moon, with a few hours interval, and assemble some > sort of animation of the lunar cycle from these images. > > For this to work, one would need to align the individual images > to some reference image frame. You realize, of course, that digital image fiddling will result in the loss of resolution and probably some artifacts as well. Wouldn't it be better to modify the scope? > Do anyone know how to do these kinds of things? I do *not* have > access to any image processing software packages. С++ compliler is all that required :-) VLV
From: Rune Allnor on 27 Dec 2009 11:07
On 27 Des, 08:42, Vladimir Vassilevsky <nos...(a)nowhere.com> wrote: > Rune Allnor wrote: > > The idea occured to me that it ought to be possible to take a number > > of images of the moon, with a few hours interval, and assemble some > > sort of animation of the lunar cycle from these images. > > > For this to work, one would need to align the individual images > > to some reference image frame. > > You realize, of course, that digital image fiddling will result in the > loss of resolution and probably some artifacts as well. Wouldn't it be > better to modify the scope? It would probably be better what the end result is concerned, but it is not practical. The scope & camera combo I am using is a terrestial scope + hand-held compact camera, so the images are not exactly top-notch. I am observing from urban'ish areas at high lattitudes, so there is a lot of convecting atmosphere, due to heat over houses, messing things up at the time of year when observing is possible. On the other hand, I am playing with a 8-10 MPix camera, so some degradation might be allowed without results turning totally disastrous. . Again, this is as much a DSP/Image processing project as an astronomy project. > > Do anyone know how to do these kinds of things? I do *not* have > > access to any image processing software packages. > > С++ compliler is all that required :-) Got that one. An the Boost.GIL image facilities framework. A free Adobe library, apparently written by the guys who wrote Photoshop, that contains most of the basics for handling images, but without the nifty processing algorithms. Rune |