Someone might find this useful - "Star freeze" comet images tutorial

[Tonk]

I've just restarted astro image processing after a substantial layoff. My last imaging session was back in August and I just had a go at processing them - so I wrote up my steps.  The link is here:

The write up is reproduced below - with reference to the images in the link above,

+++++++++++++++

Comet C/2014 E2 (Jaques) images during the night 19th - 20th August 2014

This is presented as a panel of key processing steps to record how I do star freeze images of comets. I've been successfully applying star freeze procedure since 2004. A couple of methods (MinSubMax and Min2SubMax) are my own invention but are not generally applicable unless you have the right software (Images Plus to be specific). Other methods I've regularly used have been published by others or I have adapted and improved them. This time I'm using part of the Berhard Hubl method with by own "comet sieve" procedure added on top.

All star freeze processing aims to do the following 3 steps: a) isolate the comet with statistical rejection methods isolate the stars - either stats rejection OR reshoot the star field when the comet is gone c) combine the two images of stars only and comet only to create the final image. The final alignment should use the reference frame from your original raw images.

After acquiring the images ( I do this continuously - don't follow the Hubl suggestion that you leave a time gap between images - the time can be used wisely to gain a maximum number of frames, you just change the processing to use my sieving procedure) and the calibration frames I star align and calibrate using the DSS "Register Checked Images" function. This is just a preliminary step to grade the images (don't stack them yet). After they are registered check the scores and remove the frames with low scores (non round stars, clouds intervened etc).

Now we prepare a preliminary stacked image to work out how we are going to sieve the stack for the comet isolation step. Go to the first image (in the time sequence) and carefully mark the comet position (we are still using DSS here). Repeat this with the last image. Now chose the reference frame - usually this should be the frame with the highest score but equally you could choose it based on the framing you what (i.e. position the comet well with respect to the frame edges - possibly aiming for rule of thirds) - and mark the comet in this frame as well. To recap you mark the comet position in just 3 frames of the stack - not each frame. DSS is smart and uses the frame timestamps to accurately position all the intermediate frames. Now make stacked and *average* combined image. The result is shown on the left hand panel.

Now we look for the fattest star in the image and note if it overlaps with the time adjacent images of the *same* star in the trail of stars in the stacked image. Note in this example each star does not form a continuous trail but are instead a series of dots with gaps between. This is because the comet moved by a significant amount *during* each frame. The distance between each star image in its star trail is the amount that the comet moved. Take a close look at the comet - its elongated. This is not a worry as there is a simple correction that I'll apply later. [i don't guide on the comet and I wanted to reach long enough to record the very faint tail this comet sported].

If you look at the middle bottom of the left hand panel image the images of the largest star does touch its neighbour in the trail. This touching will interfere with the sigma-kappa procedure we are going to used to isolate the comet so we can now make a decision on how to divvy (divide) up the stack for the sieve. Count how many frames are needed for the largest star to separate from its prior images. In this case we get non touching after 2 frames - so we are going to create TWO separate stacks.

[TIP - if you are going to crop the image - pick the largest star that will remain in the image]

Now here is the sieve procedure. Deal each frame in strict time order into each of the N piles (2 in this case). If you have discarded frames, then for the purposes of the sieve you *must* pretend you still have them for the purposes of allocating each frame to the correct pile.

Take each pile in turn and in DSS choose comet stacking, choose to align on the comet and pick sigma-kappa stacking. Set sigma to 1.15 (but experiment) and sigma (iterations) to 10 to 15 (again experiment) and stack. The result stacked images are collected and aligned and combined (use average). I do this step in Photoshop using layers and an opacity sequence of 100%, 50%, 33%, 25%, 20% etc (you rarely ever need more than 3 sieved stacks)

The result is shown in the middle top panel. The elongated comet was corrected by using "darken" blending in Photoshop. Duplicate the comet only image, set the duplicate to "darken" blending mode. Then use the "Offset" filter (under "filters -> Other") to shift the top image pixel by pixel to align the start of the elongated trail to the end of the trail with maximum roundness.

The "stars-only" image was also created using the Sigma-Kappa combining method in DSS - in the comet stacking tab (on the stacking dialog) do that same as before but choose "align on stars". You should now do this on the entire collection of frames as the sieve is not practical (the coma diameter is too wide to find enough piles with enough frames in each to make Sigma-Kappa work here). The result is shown in the bottom middle panel.

Most of the comet does get removed - the tail certainly did in this case (its thin enough). You will see strange artefacts left by the overlapping portions of the comet coma (green and purple in this case - this is usual). These artefacts are removed using Hubl's background extraction method (its a standard method - see http://www.astrophoton.com/tips/comet_images.pdf - Photoshop in this case). Some changes I made are: a) the Hubl method results in a pair of images and one is subtracted from the other. In the Photoshop dialog I add 25 to the "offset" field before completing this step as otherwise the background is set at or very close to RGB (0,0,0) which is far too black and look clipped. I deliberately *add* noise (the Add Noise Filter) using a value of 0.1 to 0.2 - the subtracted background is unnaturally smooth and you need to add noise to match the levels found in the comet only image - otherwise the combined images look - well - "combined".

Now you have the two images - "comet-only" - and "stars-only". If you have poor results isolating the stars you can always fall back to re-imaging the same star field later (I'm improving some of my early efforts this way).

Now the trick is to combine the images. First have the single reference frame (the stacking reference frame) at hand to guide the alignment. I use Photoshop for this. There are two methods - one based on "screen" blending and the standard "lighten" blending. The former correctly blends star and coma colours but is difficult to use (so I'll ignore for now), the latter is easy but the method just picks the brightest pixels from the two blended images - these can create strange "punched through" stars that otherwise lie behind the coma - the colour boundaries are not blended well - too sharp!.

[TIP for the next step the background brightness of the "comet-only" image should be slightly brighter than the "stars-only" image - if its not then *reverse* the "comet-only" and "stars-only" images in the following description]

For lighten blending you create 5 layers in Photoshop (one is temporary). Bottom layer is the "stars-only" image. Then add a "levels" layer, then the "comet-only" image set to "lighten blending" and then another "levels" layer. Finally add the reference frame as the 5th layer with opacity set to 50% - you will use this to get the first and 3 layers aligned (then turn off the reference image). Now use the 2nd layer (levels) to set the stars only layer to match the background of the comet-only image - use the black point value in the "Output Levels" control (under the levels histogram) to do this. With careful adjustment of the output level you can get the two images perfectly blended (you should not need to cut out the comet from the background - that can *always* bet detected on at least some ones monitor!!). Don't forget to turn off the reference image visibility when you do this. Finally the 4th layer (the top levels) is used to reset the black point on the histogram to darken the image to the required level - I choose a dark neutral grey background (25,25,25) to (30, 30, 30) when using a 0-255 brightness scale)

The result of the above processing is shown in the right hand panel. I haven't said anything about noise reduction, colour boosting and such like - that's up to you to do wither on the pre-combined images or on the final result.

SUMMARY

So in summary the key elements were: DSS for time stamp interpolation (Pixinsight can do this too) - for perfect comet alignment. DSS for Sigma-Kappa - allows control of the 2 key parameters to control what is rejected. Do an average comet aligned stack to chose the number of piles for the sieve using the "fattest star" test. Deal the images into the sieved piles on strict time sequence (accounting for discarded images). Comet align sigma-kappa stack each pile, average combine the aligned results. Star align and sigma-kappa stack *all* the raw images - then use the Hubl method to do background extraction and artefact removal. Do your own processing on the comet-only and stars-only images to clean up (try and do the same actions for both) and then combine the result (cleverly!!!). Try not to cut out parts of images - it always looks bad and can always be spotted - use smart blending instead.

[Tonk]

PS - those stupid emocons are supposed to be "b )"  :shrug:

[symesie04]

Only skimmed through but looks like a good guide. Maybe a mod might make this a sticky somewhere.

[Tonk]

The one I'm keeping up to date on astrobin  is definitive ... Its here http://www.astrobin.com/151422/0/

I updated it extensively this morning with section titles, additional tips and caveats