Telescope Reviews: AutoStakkert! 2

As some of you might already know, a week ago I released AutoStakkert!2. I thought it might be a good idea to have this thread where people could post feedback and ask question about the software.

Let me start by saying that a new update will be released that will be capable of batch processing 'surface' recordings. Here is a short list of the fixes/changes of the upcoming release (if there are more changes, I'll let you know).

2.0.0.41 - Februari 7 2012 - 1:01AM
- Should be more stable now, especially for surface recordings containing bad frames.
- Automatically turn off bad frames with horizontal or vertical shifting artefacts (they should have a quality of 0 and be at the end of the quality list. Set TurnOffMisformedFrames=0 in the ini file if you want to turn this feature off).
- Speed increase for surface alignment (approximately 30% faster).
- Introduced two Surface options: 'Expand' will try to make the very biggest image stack possible, the edges will contain less frames (this was the default option). '100%' will crop the image such that each pixel will contain the same amount of data (the edges should be fine).
- Batch processing for surface recordings. When more than one surface recording was opened, when processing, for each frame all APs are replaced by a set of automatically placed APs (in a grid, just like when you press that button). This feature is very experimental and hasn't been tested!!!!! (so please test it

)

---

At the moment there is no manual available, and I don't have time to write one. But perhaps this thread could give you some pointers on how to use the software. I'll start off by copying an email I sent to the beta testers:

-----

- File support for large (> 1 GB files, up to 4 GB should not be a problem). You can still use Castrator before running the images through AutoStakkert!, but the image quality will not be any different one way or the other.
- MAP: Multiple alignments (and along with that, multi-core support. So if you have a computer with 4 cores, the program should run about 4 times faster).
Potential drawback: it's not allowed to make more than one stack in MAP mode (you used to be able to make up to 8 different stacks using different frame percentages or frame numbers, but it was rather difficult to implement that in MAP mode. Perhaps it will be possible again in a future version)
- It can stack surface recordings as well as planetary recordings. There are two methods of global aligning, one based on the fact that a planet is always brighter than the background and the center should thus be easy to determine, the other will perform a course alignment on surface recordings (for images on the sun, or the moon, or potentially even some deepsky-recordings if you know what you are doing). I haven't really tested the program for moon recordings, but I have found it easily outperforms Registax 6 or AviStack2 on white light solar images.
- supports RGB recordings (avi recordings: many kinds of YUV formats, and RGB24. And TIF files). Color processing is generally much slower than monochrome processing (a lot of buffering takes place in monochrome processing)
- raw bayer image support (avi recordings: Y800, many kinds of YUV formats - for those who used the wrong codec to store the data - SER frames). Processing is much faster than RGB processing, and no interpolation techniques are used (R, G and B pixels are treated independently from each other, the natural drift in the images will fill in the gaps in the bayer pattern ).The correct bayer scheme is auto detected (but you can also override this if you prefer to). I won't go into too much detail here, send me a mail if you'd like to know more.
- Support for SER recordings (thanks to Heiko Wilkens for letting me use his file format!). This includes 8 or 16-bit monochrome recordings, as well as raw bayer recordings.
- Support for TIF, FIT and BMP files. Drag and drop one or more folders containing these image files onto the main form of AutoStakkert!2 to open them.
- Detail validation. A fancy word for creating two independent stacks with the same quality that can be used to separate features from noise. If features differ a lot between these two stacks (usually seen after stacking), than you can be quite sure it was simply noise you were looking at. This only works in Single AP mode (the main reason being that those stacks are slightly more reliable than MAP stacks)
- You can view all the frames of a recording.
- A new 'Analyze' function that prepares the recording for actually stacking (by calculating the quality of each frame using the chosen quality estimator). The frames will be ordered afterwards, so you can see how well the quality estimator does it job (when using the program in MAP mode, the quality of each AP will be calculated, each AP will have its own subset of images to stack). When we are dealing with a surface alignment, the analyze function will automatically set the maximum image size, and determine the area where the APs can be placed (pretty much inside the red rectangle, this is the area for which there is information available in ALL the frames)
- When the program is used in planetary mode, you can drag the planet around by holding the shift button and dragging the image. Click on the offset number, to reset the offset (it will center the planet again).
- Automatic AP placement. This is likely a preliminary version, because it is difficult to do this task automatically. It can place a bunch of APs of a particular size with the brighter parts of an image (there is a simple intensity threshold, which gives no feedback at the moment...). I prefer to manually place APs for images of planets. I will include a couple of examples on how I place the APs in my images.
- HQ Refine. Makes refining slightly more accurate (especially for really fine images). Processing will be a bit slower though, but I would leave this option ON for each serious processing efforts.
- Drizzling. 1.5x drizzling is nothing but the 3x drizzling reduced to 50%. Drizzling tries to improve the image quality of under sampled (!!!) data. It often does not produce more details, but it might be fun to play around with. Typical drizzling artifacts might occur (blocky pattern), reducing the 3x drizzling back to 50% removes pretty much all of these artifacts (but you could also work with the 3x drizzled recording if you want to. Drizzling makes the stacking phase go rather slow (about 9 times slower.. Aligning does not take more time).
- And finally: Unlike Registax or AviStack2, each alignment point does not have a predetermined area that it will represent in the final stack. After aligning and stacking all the alignment points, AutoStakkert!2 uses only the best APs to rebuild the final image stack. A single misplaced AP does not mess up the entire image, data from an adjacent AP that overlapped the same area will be used instead.
However, this does not mean that you can simply cover your image will an infinite amount of differently sized APs and an optimal image will always come out. This will not only result in a VERY slow program, it will also increase the chance that one of the poor APs by chance managed to make it to the final image after all. So place your APs wisely, and add a bit of overlap. APs produce the best results if they are placed in such a way that BOTH the X and Y direction can be determined correctly. Place a small AP right on the edge of a planet will ensure one direction is tracked properly (perpendicular to the edge), but the other direction will likely be blurred out.
One size of APs is usually the best way to go for a specific recording, unless for example there is a small moon visible inside the image, which might be better off with an extra smaller AP. Also, in the case of solar recordings, you might want to replace some of the small APs right in the center of sunspots by larger APs that cover a slightly wider area.

Proper stacking takes time. A single alignment point placed around the entire planet will always produce a decent looking image, but it will hardly ever be the best possible stack. If the seeing was absolutely excellent, it might come very close to a multiple alignment stack, but even then there is often a bit to gain by using multiple alignments.

Some tips:
You can add alignment points by clicking with your left mouse button inside a frame. In 'Single' mode, you can only draw the APs (left click once, move the mouse to another position, and click once more to place a rectangular AP), but in MAP mode you can also add square APs using a single mouse click. Use the right mouse button to remove the closest AP, and the mouse wheel to change the size of the APs you want to set.

Details in the image are aligned on a reference image which is simply a small stack of the best frames. You want this reference image to be representative for the entire recording, and contain little noise. A single frame is always a poor choice for a reference image, as it is usually slightly distorted and thus the final stack will also be slightly distorted. Also, a single frame is often too noisy to align on.
"Last stack is reference" tries to use the last stack as a reference frame (IF that last stack is available, and has the same size and same offset as the current recording, otherwise it will simply create a new reference frame).

Jupiter recordings at F/30:
- Open 1 or more recordings
- We are dealing with a planet, so make sure 'Planet COG' is selected
- Use the gradient quality method, Noise Robust 4 would probably be fine (use smaller value if the image contains very fine details, or a larger value if the recording is extremely noisy)
- Click analyze, en look at the ordered list of frames to see how well the currently quality estimator does its job.
- Determine how many frames you want to stack.
- use Multiple alignment points (MAP) and place around 30 APs with a size of around 80-120 (depending on the image scale and image quality). If there is a small moon or shadow visible, you can add an additional smaller AP around that.
- Determine how many frames you want to stack. If after the best 500 frames the image quality drops, it will generally NOT be a good idea to stack the best 1000 frames.
- Click Stack, to make stacks of all open recordings using the APs that you placed. A cancel button also becomes clickable just in case you made an error.

Mars (similar to Jupiter, with a couple of differences):
- Edge detector often works well, I would turn it off for the shadow side of mars, because it is more difficult to determine the quality there.
- If the images are really good, you might also try the gradient quality estimator. You'll have to experiment. If you use edge, every AP will use the same subset of frames to stack, in gradient mode all APs use their own subset of frames.
- Single or MAP, if Mars is big enough (> 8"), an MAP alignment should give you more detail. Add enough overlap between the APs, and use an AP size of about 8" x 8" to get you started. A smaller AP placed around the polar cap, and/or two placed near the edges of the polar cap, might be a good idea too.

Venus (similar to Mars)
- The recordings often contain VERY little noise and are made with short exposure times. The gradient quality estimator can incorrectly see the seeing bubbles as extra details, so you might be better of using the edge quality detector.
- If Venus has a crescent shape, I would definitely use MAP. Place one AP on each pointy end, and add a couple more APs around the center (large enough so you see a bit of black space on either side of Venus, as it is pretty impossible to stack on features on Venus.

The Sun
- Always use gradient quality. A small value of 2 works best for my white light recordings that contain no noise and were shot at F/15. If you use a larger image scale, I would use a value of 3 or 4.
- MAP
- The size of the APs greatly depends on the image scale and the amount of noise visible. I can use the smallest APs on my low noise data recorded at F/15 (with a 5.6x5.6um camera), but I would start off with larger APs (do a run with 100, and decrease the AP size in the next runs until you don't see an increase in the image quality).
- Use the button 'Place APs in Grid' to add APs.
- Have a look at the Analyzed frame before placing the alignment points!!! In my recordings of the sun in white light, the quality often drops very fast. Don't stack a lot of frames if only a few of them are good enough, but if you are lucky to have steady seeing, stack more frames.

Moon
- Similar to the sun

----

So, if you have any questions on how to process your recordings, just ask them here. If you run into serious problems, please send me an email as well.

Edited by MvZ (02/07/12 06:44 AM)