Wednesday, September 26, 2012

Tri-band solar imaging - part 6

In previous blog postings I have presented the tri-band Solar imaging technique and discussed related acquisition and processing details while showing a number of example images. So, what can go wrong when doing tri-band imaging? In this blog posting I'll be presenting some of the main problems I have experienced so far.
Tri-band image with no flat field correction

1. Flat fielding is always non-trivial, especially when working at high f-numbers where the shadows of dust particles become smaller and denser. With narrowband solar filters flat fielding by means of the usual sky-flat or lamp sources does not work. When just colorizing single-band H-alpha images there is so much activity across the image that a few dust shadows don't draw very much notice. With tri-band images, these dust artifacts easily become color artifacts which show up more readily on our mental radar. So, take the time to experiment with establising a flat method that works well on your setup. I have used defocussed images of spot-free regions near the central solar disc with reasonable success. Especially with maximum detuning is the need for a good flat field evident since the solar landscape here is more bland than closer to the H-alpha wavelength.

Flat frame with vignetting, dust shadows and Newton rings

2. Newton rings are a particularly nasty problem associated with an extremely narrow filter bandpass. They are caused by interference between reflections inside the optical system and are usually of quite low contrast (typically less than 1% of the total signal). Hence, Newton rings are most evident when the solar filter is detuned to reveal more of the photosphere. The problem is reduced when working with webcam-like imaging devices where drifting solar features are tracked across many frames, hence the rings are often partially averaged out. On my images Newton rings appear as curved bands which easily move across the field of view over time. This mobility is their main curse, since it compromises the ability of a normal flat field to deal with the problem. For the better part of two years I have tried all sorts of acquisition and mathematical tricks to control them without much success. In the end I started doing it manually in photoshop by drawing dark lines in a seperate layer on top of the Newton rings, then blurring and using them as a mask to selectively brighten the regions affected.

Full disc tri-band view, showing effects of filter
non-uniformity (AR1520, July 12 - 2012)
3. Filter non-uniformity exists to some degree for all narrowband filters across their field of view and is in fact one of the main factors determining the filter price tag. The effects can be mildly irritating when working with a single-band image. With tri-band images I experience it much more severely since it causes unwanted color gradients. In fact, with my Coronado SM60 H-alpha filter (a high quality unit made by David Lunt) this problem effectively prevents me from creating decent, full disc tri-band images. See an illustration of the problem at right. At higher magnifications this effect becomes less of a problem since you then more easily can apply the 'background flattening' technique I mentioned earlier.

4. Time: remember to hurry when acquiring the tri-band images! Color artifacts will arise if significant movement has time to occur between frames. Staying within 20-100 seconds should be OK for normal situations, but with very high magnification and/or very fast moving features the restrictions can be much more severe. If you get an X-class flare exploding into your face (like this one!) don't even think about try to get it in tri-band! Micro-flares are also very fast and usually not possible to capture properly - I just accept them as colorful little dots here and there.
Rapid development of micro-flares results in funny colors on tri-band images.
In the next blog posting about tri-band solar imaging I will present a bundle of shots of what I have achieved so far.

Sunday, September 23, 2012

Tri-band solar imaging - part 5

This time I'll discuss the kind of image processing involved in the tri-band Solar images previously shown here on this blog.

Frame selection, alignment and stacking (AviStack)
After data aqcuisition you'll have three data sets; one for zero detuning, medium detuning and maximum detuning of the filter. Before these are combined into a tri-band composite the image processing is the same as you would normally do. If you made a movie you’ll select, align and stack the sharpest images with software such as AviStack or Registax. You might also perform flat field correction. Finally, the images may be sharpened and perhaps also corrected for solar limb darkening. You can choose whatever you want to do – the only important issue is that you do the same for all images prior to making the tri-band composite, or else strange color artifacts may appear in the final image. I won't go into further detail here about these steps because they are not particular to tri-band imaging.
Wavelet sharpening (Registax)
Background flattening (MaxIm)

On close-up images, like the one in the example above, I often 'flatten the background' - i.e. adjust the brightness of non-active regions to a uniform brightness level. This deals with any uncorrected vignetting and removes solar limb darkening effects. Background flattening ensures that the maximum dynamic range is used for displaying details of the chromosphere and active regions therein.
I use Photoshop for the remaining process steps where the tri-band image is created. First, all three images are opened and then merged as red, green, blue channels in a color image. Be sure to maintain the bit depths of the original images - I typically work with 16 bits. You can assign the images in whatever way you prefer; I have mostly chosen to use the most detuned image for the red channel, the medium detuned as green and the H-alpha channel as blue. This makes 'energetic' features blue/green or white and more quiescent regions red/yellow. There is nothing 'correct' in this, just do whatever you prefer!
Alignment of RGB levels.

Next, use the move tool to align the RGB channels to each other, while using some distinct feature such as a sunspot for reference. You don't have to do this perfectly, it's always possible to tweak the alignment later on.

Levels/curves details
(click for large version)
The following steps consist of several levels/curve adjustments with the purpose of increasing brightness/contrast and achieving a suitable color balance. All of these steps are really subjective - click the image at right to see details of the method a have chosen to follow for many of my images.

To illustrate that the results can be very different according to taste, look at the image below and see what happens if the hue is adjusted differently!

That's really all there is to it. If you are skilled doing regular narrowband Solar imaging the extra steps involved in tri-band imaging should be no big deal. However, there are limitations and challenges - in the next part of this blog I'll discuss some of these!

The colors of a tri-band image are as much subject to personal taste as regular narrowband images. Above, I  made a major hue adjustment to achieve a completely different look! (click for large version)

Friday, September 21, 2012

Tri-band solar imaging - part 4

Solar features can evolve rapidly and hence the tri-band acquisition process must be completed before these changes become significant. If you don't hurry, 'false' colors will arise that reflect movements of gas rather than changing physical conditions on the Sun.

For high resolution imaging with a webcam or similar device the quest for getting many images to catch moments of good seeing must be weighed against this requirement. Also, the more you zoom in on the Sun and the more active a feature you are targeting, the shorter the available time will be.

It is not possible to pin down a firm limit, but with my targets, preferences and setup (Coronado
SM60, f=1280mm, Skynyx 2-2M camera) I find the total available time to range from 1-5 minutes. Hence, if images are to be acquired for three wavelengths, the available time pr. wavelength is reduced to roughly 20-100 seconds. With my camera the acquisition rate is around 10-25 frames pr. second, hence each series consists of roughly 250-2500 images. Since Solar features change more rapidly in the chromosphere than in the photosphere, the time restrictions are least severe when the filter is detuned the most from the H-α wavelength. Consequently, images with the largest detuning from H-α ‘last longer’ for tri-band work.

Many filters are tuned mechanically in seconds, by tilting or by using a pressure regulator. A fast tuning mechanism is obviously best suited for tri-band imaging. Other filters are thermally tuned which is a slow process that requires several minutes to reach equilibrium between each wavelength change. With such a system, one could opt for ‘dual-band’ imaging where only two images are acquired for the red and blue channels, respectively– one at the H-α wavelength and another at maximum detuning. A green channel can then be synthesized as an average of the two other channels. An example is shown below. For this image a Daystar Quantum SE 0.5Å filter was used with 6562.8Å for the blue channel and 6563.8Å for the red channel. The result is not as rich in color variations, but I still consider it worth the effort.

Dual-band image of AR1195 from April 25, 2011. Daystar Quantum SE 0.5Å H-alpha filter.
Below is another shot - this time a real tri-band image - of AR 1087 on July 10, 2010. This was acquired using a Coronado filter that was detuned by tilting. The colors are obviously better.

Tri-band image of AR 1087 from July 10, 2010. Coronado SM60 filter.
Tomorrow I will discuss the processing flow I use for making these images. A lot of it is standard stuff for Solar work, and the tri-band part is actually quite simple.

Thursday, September 20, 2012

Tri-band solar imaging - part 3

In the previous blog posting I posted a question: how will spicules and prominences appear at the solar limb? I am using a color scheme that arranges the most detuned image as red, medium detuned as green and not-detuned as blue. When seen in silhouette against the photosphere spicules and filaments are revealed by their absorption of H-α light. They look dark in H-α images and since this is blue in my color scheme they will appear reddish. When seen against the blackness of space these same features are revealed by their emission of H-α light - and appear bright in the blue (H-α) channel. Check it out for yourself below!
Active region 1193 on April 22 -2011, with many internal color variations. The chromospheric limb appears bright blue since here we see H-alpha emission which I choose to map into the blue color channel. When seen against the photosphere the same features are revealed by their H-alpha absorption and thus appear reddish. Click image for full size view.
Now, take a shot and guess how a flare will appear in a tri-band image! I will show a picture of one later.

I had a hard day at work and blogged too late into the night yesterday, so you'll have to wait until tomorrow for the next posting where I will discuss acquisition aspects of tri-band imaging.

Wednesday, September 19, 2012

Tri-band solar imaging - part 2

Detuning by tilting - Coronado's T-max tuner

The Sun provides so much radiation that a tiny wavelength interval can be selected to enable viewing of certain layers of the solar atmosphere or certain elements – and still have plenty of light for fast exposures that freeze the seeing. Most notable is the hydrogen alpha (H-α) transition at a wavelength of 6562.8Å. Emission at this wavelength is a sign that hydrogen is being ionized and it comes predominately from the solar chromosphere. Narrowband solar filters can be detuned slightly to maximize contrast or to allow observation of Doppler shifted light from features that are moving rapidly along our line of sight. The tuning range is typically very limited – around one angstrom – but even such small changes lead to dramatic differences in the resulting image. I use a Coronado SM60  H-α  filter that can be tilted slightly using  the so-called T-max tuner; see the photo above.

Rates of change in brightness of various features with filter detuning and how they relate to colors in a tri-band image
(click for bigger version)
The figure above shows a typical scene, captured with this filter at three different wavelengths. The targeted object is active region 11087, which consisted of a small sunspot 20” across and a complex magnetic field which produced several low-level flares during July 2010. At the H-α wavelength of 6562.8Å (corresponding to zero tilt), the field is covered with spicules and contains two dark filaments and a bright region of plage. In all this action the small sunspot almost gets lost, with only the umbra standing out clearly as a dark, irregular spot. At 6562.3Å (half tilt) the contrast of the spicules, filaments and plage decreases. At 6561.9Å (maximum tilt) these features are nearly gone so that the sunspot with penumbra becomes clearly visible. A smaller spot, unnoticed before, becomes apparent towards the upper left.

As the wavelength is tuned away from the H-α transition we are viewing more and more light from the underlying photosphere. What is especially interesting in this regard is that the rate of change in brightness with wavelength is different for various features, depending on what they are. It is not the same everywhere. Hence, if detuned images are used as separate color channels in a composite image various colors will appear that represent different rates of change. This is evident in the tri-band image above where the R, G, B channels are chosen as most detuned, medium detuned and not-detuned, respectively, from the H-α wavelength. In such a color scheme the underlying photosphere appears yellowish, spicules and filaments are reddish, plague is white, etc.

Tri-band image of AR1087 in a quite moment. This active region had a beta-gamma magnetic field and harbored multiple M-level flares during July 2010. Click to enlarge.

Many more colors can be found in this way on the Sun and I'll be showing more later on. For example: how do you think the chromosphere and prominences at the solar limb will appear in a tri-band image as the one above? See for yourself in tomorrows blog posting!

Tuesday, September 18, 2012

Tri-band solar imaging - part 1

In this series of blog postings I will present a method for creating colorful solar images using a narrowband filter. In short, the idea is to tune the filter and acquire three images at different wavelengths. A colorful scene can be created by assigning each image to a different color channel of an RGB image. The resulting color image reveals real spectral changes which
are equivalent to changing physical conditions on the Sun. I call this technique ‘tri-band solar

Tri-band versus single-band H-alpha image (click for bigger view)
An illustration of the resulting view and how it compares to 'normal' colorization of a narrowband H-alpha image is given to the right. The tri-band version displays much more vivid and varying color details and represents sunspot penumbras better than the single-band, on resonance H-alpha image. All you need to make such images is a standard, tunable H-alpha filter and a bit of extra processing work.

Stay tuned to this blog since I'll be posting daily about tri-band solar imaging with the following headlines:

  • Background
  • A basic example
  • Acquisition techniques
  • Image processing
  • Tri-band pitfalls
  • Some colorful examples

I hope that others will find these presentations inspiring enough to try some of their own tri-band work. Please share any thoughts, comments and pictures!

Tuesday, September 4, 2012

Full disc Venus transit

The Venus transit of June 6th continues to unfold - on my harddisc. I still have lots of unprocessed data to play with and yesterday I did just that. The resulting image is my first of the event featuring a full Solar disc view (full res version HERE).

Venus transit in H-alpha light (Coronado SM60 filter, 70mm ED refractor)
- click for bigger view
The view is a composite of two images made out of 600 frame sequences; one for the solar disc with  2 msec exposure time and another with 27 msec exposure time for the prominences. Acquisition was done at 3:17 UTC when the altitude of the Sun was just 3.8 degrees. When processing the image it was quite evident that the solar disc was flattened significantly by atmospheric refraction. Still the view is not too fuzzy - it helped that I was WAY down at the beach front overlooking the straight between Denmark and Sweden.

Initially, I had set up a bit farther from the water front but as more and more bystanders gathered and flocked around - and in front of - the scopes you could forget about seeing anything. Failing to get anyone's attention to move out of the way I desperately grabbed the entire setup with clock drive running and carried it down to the waterfront. If someone wanted to get in front of my scope here they'd have to get their feet wet. On the picture below you can see me sitting at a table working the computer but the scope I used is completely blocked from view by people! But, hey, lots of bystanders - and not lots of clouds - is something to be happy about!
Imaging at the waterfront - surrounded by an enthusiastic crowd.
Image copyright: Palle Sonne