Registax Video Analysis Procedures
By Tony George
(revised June 27, 2007)
Registax is a shareware program designed to stack still frames and AVI format NTSC videos. IOTA members have used Registax to process video files to:
1. Create confirmation star charts,
2. Enhance faint stars for occultation detection,
3. Compress video files to create ‘time lapse’ videos of long events such as NEO flybys, and
4. Detect faint transient events such as lunar impact flashes.
This document provides step-by-step procedures on how to process video files to enhance faint stars for occultation detection in typical situations that IOTA members might encounter.
General Installation and Comments:
Registax may be downloaded on the main download page located here. On this page, you will find download instructions for the latest release. As of the writing of these procedures the most current version is 188.8.131.52. There is also an instruction manual available in PDF format. Registax is supported in English format as well as a variety of different languages. Registax will analyze video files up to approximately 2Gb in size. Anything larger will not run on Registax. For good insurance, don’t try to run any files larger than 1.95 Gb.
Registax is sometime sensitive to the type of Codec used to create the AVI file. It is always possible to translate an AVI file into a version that Registax can read using the freeware program VirtualDub.
Enhancing Faint Stars in Occultation Videos
When occultations of faint stars or when processing videos to detect low magnitude occultations or very short occultations, we often use the video photometry program called LiMovie. This program allows us to extract data from events that would be nearly impossible to observe visually. One problem with LiMovie is that of discriminating the target star from the background noise. High gain video cameras such as the WAT-902H2 Ultimate or the Supercircuits PC164C both have the problem that they can have high noise levels at full gain. Registax gives us a method to preprocess videos so the noise is suppressed and the signal from the target star and other field stars is enhanced. This is the Integration/Running Average method. To use this method, we do the following steps:
Registax allows you to ‘pre-register’ the AVI video to a bright star anywhere in the field of view. This allows the frames in the video to be aligned on that star or other relatively bright feature. This technique could be extremely useful for ‘drift through’ occultations where the telescope is preset to a specific altitude and azimuth and the target star drifts through the field of view at the time of the occultation. It is also useful for occultations were there is a lot of movement of the image due to wind shaking the telescope. To ‘pre-register’ the video, follow the steps below:
1. Open Registax
2. Select the file to analyze (you can use this practice video or one of your own videos)
3. Registax should open the file and show the first frame of the video.
4. If you have a relatively bright field star in the video, or if the target star is bright
enough, you can do a pre-registration that will further improve the ability to detect faint
stars and reduce noise. To pre-register the file, first select the 32-pixel
Alignmentboxsize and the Default (single) alignment Method. Then Left Click the cursor
on the brightest star in the field of view. Be sure to uncheck Automatic Processing.
Now click on Align. This will align all the frames (and fields) on the star selected.
Next move the Stack Slider on the bottom of the screen all the way to the right. When
integrating frames, we will leave the slider at the far right.
Now click on Limit. This selects all the frames for stacking that are within the limit
set by the slider.
Do NOT click on the Optimize and Stack button.
Instead, press the Stack tab in the menu.
5. The next step is to save this pre-registered file. Click on the Save Register/Integrated AVI tab. On this tab, select the
Maximum Area option under the Create registered AVI file option.
Finally click Save. You will be prompted for a file name. Select a name that will indicate this file is ‘pre-registered’ and then
the program will save all the pre-registered images to a new AVI file.
Registered/Integrated AVI Moving Average:
Whether or not the video file is ‘pre-registered’ the signal-to-noise ratio of the video can be enhanced by integrating each video frame into a moving average. To integrate each frame into a moving average, follow the steps below:
1. Open the file (or pre-registered file you just created) in Registax. Check the None method
of alignment. The Stack Slider on the bottom of the screen will move to the right. Do not,
however, limit the number of frames that will be stacked by moving this slider. When
integrating frames, we will leave it at the far right.
Click on Limit. This selects all the frames
for stacking that are within the limit set by the slider.
2. Click on the Save Registered/Integrated AVI tab on the menu bar (see image in #5 above). Click the Moving Average
Method of integration. Select the number of frames you want to integrate in each frame running average. Only an odd number
of frames are available, as this allows Registax to maintain exact time continuity with the original video file when the video is
time inserted. If 5-frames are selected, Registax will start with video frame 3, stack and integrate frames 1, 2, 3, 4, and 5 into
frame 3 and then write that frame to a new video file. It will then move forward one frame and repeat the integration on frames
2, 3, 4, 5 and 6. It repeats this process for the entire video all the way to the third to last frame.
3. Click Save. (Moving Average) Registax will prompt for a file name. Give the file a name that will
indicate it is an integrated moving average file. Once you name the file, Registax will begin
integrating the file. Depending on the number of frame integrations selected, this may take a long
time, so be patient. When Registax is done processing, AVI written will appear in the progress bar
at the bottom of the window. You may now process yet another file, or leave Registax and analyze
your new brightness-enhance and noise-reduced video with LiMovie.
Here are some examples of improvements that can be made with Registax.
Figure 1: Screen print of unprocessed occultation video taken with WAT 902-H2 Ultimate on manual gain.
Notice there are three stars visible on the screen.
Figure 2: Screen print of video 5-frame ‘integrated’ AVI processed by Registax. This video was not pre-registered, as the noise was slightly too great to get good registration.
Notice that five stars are now visible. These stars could be easily processed by LiMovie for occultation photometry.
Figure 3: Screenprint of a different occultation video raw data – visible stars are indicated.
Figure 4: Pre-registered 3-frame integration running average. Notice one additional star becomes visible above the background noise.
Figure 5: Pre-registered 7-frame integration running average. Notice two additional stars become visible above the background noise
Figure 6: Pre-registered 31-frame integration running average. Notice one final star becomes visible above the background noise.
Also notice how well all the other stars stand out against the background noise. LiMovie would easily detect an occultation of any of these stars,
however, due to the integration, the duration would have to be longer than one-second to be detectable.
Timing Issues Using Registax Register/Integrate AVI
While using Registax’s features to enhance the visibility of faint stars and reduce the background noise allows us to observe events that would otherwise not be observable, the procedure does have some disadvantages:
1. Because adjacent video frames are registered on top of each other and added to create an equivalent new video image, the time
stamp on GPS time inserted videos is ‘blurred’ in the millisecond, hundredths, tenths, or even seconds unit, depending on the
integration number selected.
2. The event times of the occultation are also ‘blurred’.
We can deal with these issues as follows:
On the following page is a sample of an artificially generated random video data stream with an embedded occultation in the data, including a simulated Registax processed output video three-frame and 5-frame integration running average.
Notice for this example that the data has some random noise superimposed on it, making it look like an actual occultation data stream. The average for the ‘bright’ portion of the data is 16 +/- 1.5. The average for the ‘occulted’ portion of the data is 10 +/- 1.5. Our signal to noise ratio is 6/3 or 2:1 or 200%. For a normal occultation of a bright star, we would just read the occultation times from the raw data and use the inserted times from each frame. However, when we have a faint star, we cannot reliably ‘see’ the occultation in the raw data and we only have the integrated data to work with. In this case all three data series – raw, 3-frame, and 5-frame – have been plotted on the same chart so the equivalence can be seen. Notice that the midpoint of the slope from the ‘bright’ average down to the ‘occulted’ average coincides with the actual occultation point that would be selected from the raw data. This relationship holds for even very noisy data if the occultation event is ‘binary’ e.g. the star disappears and reappears in a single frame. For binary events the midpoint of the sloped line works to select the event time even where it is nearly impossible to detect the occultation point in the raw data. Try this technique with your own past occultation data and you will see that it can work. It may not work for every occultation however.
Also, there is an issue with what timing error to report on your occultation report when using the integrated moving average capabilities of Registax, we cannot simply apply a +/- 0.008 seconds error limits per field or +/- 0.017 seconds error limits per frame as we can for observation made using raw data. Even though the midpoint of the sloped line should give us a similar accuracy for a binary event, there is some as yet ‘unquantified’ additional error involved. You should report your observed error as +/- the field or frame error, whichever is used times ½ the frame integration factor. If you use a 5 frame integration, use +/- 2.5X0.017 seconds or +/- 0.042 seconds until we determine a better way to report the error limits.
Sample Occultation with Noise
Sample Occultation without Noise