Part 2: Novices Test Astro-Imaging by Webcam

My last chemical process astrophotos were probably done twelve to fifteen years ago. Many of them were not even printed, as the negatives were unpromising. I have none of them and therefore cannot show my readers how unremarkable they were.

But to me, digital imaging is an entirely exciting and intriguing development. In some ways it resembles the never-to-be-replicated aspect of live observing; in others, it is equivalent to long-exposure astrophotography. Unfortunately it would appear that to date only the most expensive, state-of-the-art amateur CCD equipment, and fussy RGB compositing, can equal the beauty and depth of the finest traditional chemical process photography of the recent past. In many respects, up to the 1990s, the best way to record the fine details observed by eye when examining the planets was by means of careful drawing (as in the author's attempt, done while viewing Mars with a 10" Newtonian telescope in the 1980s, shown at left as reproduced in our software program Eyepiece.)

But planetary imaging presents us with an entirely different situation. There are few impressive amateur photographs of the planets, predating digital imaging. The best of these tend to have been surpassed by professionals in the thirties to fifties of the last century. And no terrestrial photos of a planet can come close to the clarity of satellite imaging. Just how much effort is required, though, to do DIGITAL pickups of the planets with our modest amateur scopes?

The previous page of our "Novice tests" focused largely on the first trials that Charles Chew and I made with the Meade LPI and DSI devices, showing some distinct possibilities for future refinement (but obtaining, frankly, no remarkable images worthy of the "masters".) The Meade LPI seemed easier to work with and quicker to provide satisfactory results than the relatively fussy and sensitive -- and somewhat lower-resolution -- DSI. But I found that friends and colleagues had often achieved better planetary images than my first trials by using simple, cheap, even obsolete webcams. Surely just about the best amateur digital pictures of Mars, during the 2003 opposition, seemed to have been made by the now-discontinued Philips ToUCam (sold only by a few specialty dealers, with astro fittings, at a premium price.) But a quick web search yielded articles by enthusiasts who had used every type of webcam, from lowly old B&W models to the latest wireless gadgets. Usually these required at least one simple modification: the removal of their internal lens system.

In the first part of these articles I demonstrated my first trials with a discontinued, "remaindered" Logitech Quickcam, purchased with software in the original commercial packaging for a mere $32 at "Surplus Computers" in Sunnyvale. I did not have time to modify the unit but merely tried it with afocal projection, held up to an eyepiece. The result -- shown at right -- was dismal: a grainy, soft image containing lots of noise (and some specks and threads that were on the surface of the eye lens of the ocular.)

Modification was essential. But I was so pleased with the way the little Logitech worked as a webcam that I did not want to modify the one I had purchased. So I bought another one!

The most appropriate web pages with instructions for suggested modifications for this model may be these: An Adventure in QuickCam Astronomy by Geoff Chester; a helpful step-by-step instructions with pictures and links at Stars Website by Chuck Shaw; Hanno Mueller's How To Disassemble a Quickcam (for an older B&W model, but quite appropriate to the later color Logitechs); Easy Raw Mods for Philips and Logitech Cams; Logitech Long-Exposure Modification, and The Firmament (with some excellent Mars and Saturn pictures.) Many of these and other QuickCam astronomy sites are on personal home pages, so they tend to come and go: be sure to do a new web search to confirm these addresses.

As usual, being an impatient person, I wanted to try the quick-n-dirty approach, to see if I could get some immediate results. So I unscrewed the case, carefully removed the little plastic filter-and-lens assembly, put a piece of black electrical tape over the surface-mount green "pilot light" LED, and merely glued on a plastic 35mm film case with the end cut off: an appropriate fit for my 1.25" telescope focusers.

I let the cement dry for a couple of hours and my gadget was ready for the first experiments, during daylight. Yes: the modification was a great success, producing an image with my inexpensive Orion ST80 refractor of a car almost a block away from my house. (I have blurred the license number intentionally.) The day was very overcast and dreary so the picture lacks color and vividness, not entirely the fault of the cheap webcam.

I found that focusing would be a great challenge. My garage setup has the computers facing entirely away from the driveway where the scopes are set up, making it inconvenient to run back and forth between scope and PC. Furthermore, the monitor was positioned wrong. So for the first tests of digital imaging I installed a second, old, monitor on a cabinet near the driveway. In my first tests I used a pair of binoculars to watch the screen while focusing my C-11. But this only works at night, when the dim screen of the monitor may be seen clearly.

For daytime focusing I happened to remember a never-used "accessibility" applet supplied for Windows computers: the magnifier. I installed it on my machine and set the magnification for 3 to 6 times, zeroing in on the Logitech control panel image display window. Now I could get a close, fairly bright detail magnified enough to see it clearly while adjusting the telescope some feet away (though on my old 500-MHz machine the processing overhead of the magnifier causes a certain amount of delay which one has to accomodate while focusing.) Unfortunately, the magnifier panel completely scrambles the arrangement of one's preferred Windows desktop placement of icons; so I use a handy $6 utility called Wintidy to restore the order that I've saved.

By means of the Windows magnifier, I was able to get a far sharper focus, as seen in this desktop capture picture. The round metal object used as the test for fine focusing was on the back door of an SUV located about a half-block away from my driveway. Focusing on this seemed almost more critical than for visual astronomy by eye (as, I presume, one's own eyeball-muscles may try to 'help'.)

In a few hours I was able to use the combination of the Orion ST80 refractor and the modified Logitech webcam to acquire a few test images of the Moon. The 80mm refractor seems to have a slight amount of chromatic aberration -- typical for its simple design and price range -- as well as some light scatter likely in the objective as well as the tube assembly: by eye, the limb of the Moon was surrounded by a slight halo. The first raw test images also had some faint horizontal scanning lines as well as a considerable amount of chroma noise. Some processing steps reduced the artefacts until the images seemed reasonably acceptable, though not by any means best possible quality, even considering the modest equipment under test. Below, the left image reproduces the effect at prime focus; the right picture is a narrower field since a 2x Barlow was added ahead of the webcam.

The raw image was crude and ugly, with low contrast and noise; and though the scope was tracking well, the orb of Mars seemed to be dancing around incessantly, moving about 1-2 arcseconds without cease. So each 1/200th second exposure was short enough to arrest the most of the motion, but detail would necessarily be blurred a bit.

This WMV file shows the "rolling seeing" that I was experiencing, in real time.

Heavy processing, including an anti-moiré pattern reduction, Gaussian smoothing, contrast stretching, and unsharp masking, yielded something useful that is a bit more gaudy than the eyepiece image, and not less detailed. One is always tempted to go a bit too far with processing: have I yielded to that here?

One assumes that careful stacking of a succession of the image frames in the video -- after throwing away the worst and fuzziest ones -- might provide a composite with lower noise and crisper detail. This is easy to do with Meade's Auto Star software provided with their CCD devices, but -- frankly -- bewildering with some inexpensive or free alternatives. I have spent a couple of days trying to fathom the complexities, jargon, andf interface of "RegiStax", almost to no avail.

My first attempt to register and stack about 300 frames from the full-scale AVI file I made while imaging Mars with the webcam yielded the results shown below. The pictures are smooth but almost completely featureless aside from a broad, smeared dark region that looks far less defined than it did in the eyepiece. After some further processing, including a correction of the camera's rotation in the ocular holder, very little latent detail could be uncovered: a waste of time.

I decided that, at least in this case, I'd be better off chosing ONE good frame, and processing that to try to correct the hue, rotation, contrast, and lack of detail. Here (below) is the point at which I have left off further modifications. (Did I once again succumb to the temptation to go too far?)

The setups for imaging the Moon are shown below:

I had been used to the fairly reliable tracking of the C-11, which made it possible to move from scope to "control point" and continue to have the proper image centered on the CCD. Of course the little alt-az manual scope has no clock drive, so the sidereal drift rate means that one has to move FAST! And I found that the Logitech camera CCD was just a little bit too far away from the focal point, requiring the use of a 2x Barlow. This narrowed the field and increased the apparent image drift, meaning that I had to move the 15 feet between scope and keyboard about a dozen times before I had managed to "take" a usable picture showing any lunar detail. Because of the drift I had to use a very short exposure, robbing the picture of the high contrast achieved with the C-11. The raw image shown above is typical of the disappointing results I achieved while trying to do a "still" of the Moon with this rudimentary telescope; and it also reveals the scan lines in all images that I have to try to remove with "anti-moiré" algorithms in one of my digital image editing programs.

I found a more interesting way to show the performance of my wife's little scope: by recording a "movie" that I named Moondrift (should it be, instead, 'Earthdrift'?) Of course, by eye -- even using one of the simple and inexpensive Ploessl oculars supplied with the telescope -- the view is dramatically better: brighter, sharper, more richly-contrasted. In this instance the webcam definitely fails to show as much as a live observer can discern.

UPDATE: Mars, Saturn, 16 November.   On the next occasion of decent seeing I set up the C-11 and once again imaged Mars. In my opinion the resulting processed version is slightly more detailed and "interesting" than the one shown above, acquired two days earlier when the seeing was rougher. (I have rotated the image to agree with the CalSky simulation for this approximate time.)

This final version has been processed from one best frame of many taken, with 1/100th second exposure time, using a 2x Barlow.

In this particular example, the eyepiece view -- which seemed satisfactory: about average for decent seeing conditions during this opposition -- was distinctly less detailed than the acquired webcam image as enhanced for improved contrast and chromatic richness.

I returned to the eyepiece to study Mars visually, and tested several colored filters. The sharpest and most detailed view, showing a trace of the bright polar region, seemed to be provided by a Wratten 56 green filter, preferable to the more "realistic" views with an orange or red filter. The dark markings seemed steadier and firmer. So I decided to test the webcam in black-and-white mode, to see if I could improve the detail while reducing the background noise. Unfortunately the resulting raw image certainly looked no smoother than the color ones; in fact, due to the increased gain I had to employ to make up for the light loss of the filter, the image was considerably nastier, as shown by this blowup of part of the raw frame.

But the raw material yielded nicely to processing, and in a few minutes I had an acceptable result; but, I cannot claim that it is superior to the color one above, nor even equal to what I could see in the eyepiece during steadiest glimpses. It does seem evident, however, that a slight trace of the polar hood is visible, while it failed to register on the color image.

I can infer that with patience, and extraordinarily careful telescope and imager alignment, one could get much closer to the superbly-detailed Mars pictures of the experts by employing a composited RGB technique: surely beyond my current crude setup (for one thing: the plastic 35mm film can that acts as the eyepiece holder projection for the modified webcam has considerable play and loose tolerance. It would have to be replaced by a properly sized sheath for repeatability of positioning.)

By 1 am on the morning of 16 November, Saturn was up high enough to consider studying, out of the blurry air near the horizon at my location in the valley floor near downtown San Jose. In the eyepiece the view was exquisite, so with great anticipation I began working on a webcam image. My first discovery was that with the exposure time, gain, and brightness settings that worked well for Mars, no trace of an image of Saturn showed up on the monitor! The planet's surface brightness was so much dimmer that the control panel settings for the Logitech webcam had to be radically altered: instead of 1/100th second exposure time, I had to use 1/30th or even 1/15th second. This made seeing turbulence far more significant. And the gain control had to be increased nearly to maximum, causing an ugly blotchy noise pattern to appear in the live image. After much effort to adjust best focus, I began taking frames, finding little variation in sharpness.

When the uncompressed raw 640x480 images were examined, in much larger scale than appeared in the Logitech monitor panel, I was disgusted to perceive an ugly blocky pattern across the entire image: obviously the background noise had acquired a "folded in" artifact of the x-y axis of the entire frame, repeated across the face of the image. This cropped blow-up at left reveals the limitations of the webcam used under these exposure and gain settings: a dreadful result.

Could something be done with this? I spent an hour, in my frigid garage, running the picture through every conceivable processing algorithm on three graphic editing programs. I removed the moiré pattern; did various types of blurring; used unsharp masking at every conceivable threshold level and pixel-range setting; and adjusted the color balance of the background, foreground, and middleground. Finally I rotated the image to agree with the CalSky simulation for this time of night, vertex 'up'.

The result shown at right is...less awful. The best that could be said is, I might admit, that now my $3500 telescope seems to work like a broken T---o 50mm refractor! Based on the quality of a raw frame of Mars, taken with the Meade LPI on Halloween night, I would assume that the Meade imager, with its lower intrinsic noise and higher sensitivity, would have "run rings" -- pun intended -- around the Logitech webcam Saturn desecration. I suspect that what one perceives as a slight trace of the Cassini division is really "manufactured detail" that occurred from the various stages of image processing, as the raw image has almost nothing of the major dark zone in the rings (observed with such exquisite crispness in the eyepiece view during the best moments of seeing.)

In these examples of the two planets, one cannot help but conclude that the webcam employed was only able to render the amorphous and vaguely-contrasting blotches of brilliantly-shining Mars with any degree of realism that resembled an ocular view, while Saturn's subtleties -- below the Logitech's resolution and straining its dynamic range -- almost completely defeated the device. The webcam may be pressed to work, straining every function to the maximum of capacity, for getting slightly soft but not unpleasant pictures of the brightest celestial objects, like the Moon or Mars (at opposition); but even fainter Saturn poses many challenges that must be overcome. I impatiently await the convenient placement of Jupiter in the night sky to see if it might fall somewhere in the usable range between these extremes!

UPDATE: Mars, Saturn, and Regitax: 18 November.   I was consistently dumbfounded by attempts to use Registax, and conflicted between my poor results and the obvious excellence of other users' best shots, thus processed. So I took my files to a local digital astro imaging expert, Ken Sablinsky (store manager of Orion Telescope Center in Cupertino), and carefully watched his handiwork.

My first discovery was that HE was using Version 3, while I had been diddling around uselessly with Version 2, which does not seem to be as effective. On his computer, stacking four frames of my Mars webcam images, Registax slowly began to demonstrate real surface detail and contrast buried in the data, more than I had been able to achieve by processing any one picture. So I came back home and immediately downloaded and installed the latest version of Registax : voila! Now I could make some sense out of the processing steps, as at last the "wavelet" processing function had a real impact on the sharpness, contrast, and residual detail. I repeated Ken's experiments with four frames each of my Mars and Saturn images, shown above in my previous best attempts. Here are the results, with further processing to clean up the remaining noise artefacts:

Compare the Saturn digital image, above, with a computer-colorized hand-drawing I made nearly twenty years ago, shown below, done while viewing with my old 10" Newtonian telescope in near-perfect seeing. This image is used in the gallery of objects in our telescope simulator program, Eyepiece.

While I am not entirely happy with either of the digital pictures (yet) I can see some distinct improvements in certain aspects of the images, with room for further experimentation. (Thanks, Ken!) Stay tuned...

UPDATE: Mars, 19 November.   Now that we have managed to get some results from Registax, the question came naturally: how SIMPLE can your equipment be, for a recognizable image of Mars? So we borrowed an inexpensive 5" (127 mm) aperture Maksutov telescope, an Orion StarMax model, and tried it with and without a 2x Barlow using no clock drive whatsoever! Yet we were able to get a series of digital webcam snapshots of Mars that had at least some trace of surface detail: perhaps about as much as could be seen by eye, with a 6mm ocular and a Wratten 21 orange filter.

I did about a half dozen images, 1/50th second and 1/100th second exposure times, with and without a Barlow, respectively. The picture at left an example of the prime focal image, no Barlow, in a cropped view of a raw, unprocessed frame. The short exposure was sufficient to freeze the steady drift of Mars through the field of view. You may see this as I watched it on the Logitech webcam control panel by clicking here for a 7-second WMV file, 33kB in size.

The raw stills, and the video, were rather pale and fuzzy. Registax Version 3 was then employed to stack and optimize 6 images and to do sharpening and contrast enhancement, followed by more processing to fine-tune the concluding results for a JPEG file. The result, while scarcely competing with the images we've obtained using our Celestron C-11, is still a credible picture with a distinct registration of light and dark contrast ranges but no sharply defined boundary regions. Frankly, that was about the way Mars looked by eye using the same telescope, though the seeing on this particular evening was very steady. Surface detail just did not stand out with stark clarity.

But the Moon was another matter. It looked fabulous to the eye; so I was certain that a picture would be gratifying. In just moments I obtained this 1/100th second digital snap with the webcam, despite the drifting image, not being held by means of a clock drive. The long focal length of the scope -- 1540mm -- and high focal ratio of f/12.1 meant that the field of view was narrow, but the light gathering power and sharpness of the optics combined to assure satisfying results. This small, convenient telescope would surely make a very enjoyable planetary/lunar instrument, and even offer some potential for digital imaging, assisted by appropriate equatorial mount and drives.

In the flash-assisted night picture of our setup, you can't quite discern that axis control drives had been affixed to this borrowed scope's mount, but they were found not to be working correctly during these tests and were left unpowered, their clutches disconnected. Re-centering of the images was accomplished by hand, my patient wife Regina sitting at the telescope while I operated the Logitech camera controls and took the videos and stills. As such, this experiment certainly indicates how easy it would be for a beginning amateur astronomer to start enjoying the fun of digital imaging, even with very simple equipment!

UPDATE: Jupiter! 7 June 2006.

We haven't had our C-11 telescope set up for months, but early on the afternoon of 7 June we set up the scope to do some power supply tests for our article on portable batteries. Why not leave the scope in the driveway, and try to image Jupiter?! Bad, sadly, it turned out that the seeing was rotten (perhaps about a 3 out of 10, at 9pm with the planet at about its highest elevation, compared to the nights of 7 or 8 out of 10 a few weeks ago, when we were trying out our new Orion 120mm refractor.)

Nevertheless, we took a lot of still images of 1/50th second exposure with the modified Logitech webcam, as well as two 20 and 40 second video clips in AVI format. A still frame (left) shows how poor the image was, unprocessed. The moon Io is barely visible (upper left) although it was bright and clear to the eye (we used it for focusing the image.) The seeing was turning the Galilean moons of Jupiter into a rolling cascade of odd shapes, never settling into crisp disks.

The magic algorithms of RegiStax 3, however, worked their usual wonders. Within a half hour, we had removed the really bad frames and stacked about 300 of the rest, the result being clarified into distinct details in the north and south equatorial belts, with a hint of loops and festoons as well as traces of color that weren't visible to the eye, with an ocular, or in each individual raw frame. It's definitely NOT a world-class image, but isn't appallingly bad, either. The perspective is North up, rotated after checking with the planetarium program we use, "TheSky Version 6".

We can only guess how much better it could have been (a) with the really good seeing we occasionally experience right in suburban San Jose at about 100 meters above sea level; (b) if we had borrowed the Meade LSI imager, with its intrinsically higher resolution; and (c) if the Great Red Spot -- and the new 'little red spot' -- were in view! Ah, well...

Meanwhile, we still have the raw images to work with. The composite at left, produced on July 20, 2006, illustrates that a slight improvement in detail could be made by various steps of processing: unsharp masking, contrast stretching, selected regions of Gaussian blurring, edge enhancement, and noise shaping.

The lower right version (with largest enhancement) is a bit artificial looking, but perhaps is closer to those moments of crisp seeing that could be detected in the eyepiece. The loops and festoons are almost as clear as they were by eye, and the belt contrast is even greater than in the live ocular view.

Comparison of Meade LPI and Modified Logitech webcam

A search for Jupiter images on the Net, to compare my first trial effort (above) with the Logitech webcam, yielded a number of superior, shaper images (such as the ones found here: the best I've seen with a Logitech as the imager), as well as some that did not seem any better than mine (such as the one on this page.) But, none of these had the "snap" of the best planetary pictures taken with a dedicated, specialized astronomical imaging device.

My comparative test of the Meade LPI against the Logitech webcam, in a daytime view of a light on the porch of my neighbor's house, across the street, through my Orion ShortTube 80 refractor, shows the difference between two such CCD imagers. The cropped picure below is unprocessed except for a slight rescaling merely to match the sizes: the LPI image (below, left) is sharp and crisp, with rich color. The Logitech picture (right) is washed-out and slighly blurry even though exquisite care was taken to adjust focus and to set the gain and exposure parameters.