Registax is a powerful program, and watching it work is amazing.  Registax can analyze, align, and stack thousands of images in just minutes.  And the whole procedure is displayed for you --  watching your computer think has never been so much fun!  Registax is a free download.  Click here to visit the Registax website.

Registax is typically used with a video file (.AVI) captured using a webcam.  See the Capturing Images with a Webcam page for details on capturing these images.

For this section, a 200-second video clip of Mars will be used.  The video sequence contains 2000 frames.  The image was taken through Rick Auerbach's TMB 254mm f/9 triplet refractor, using a 5x Powermate lens and Lumenera webcam.

Above:  On the left, a single raw frame from the video clip of Mars.  On the right, the final processed image using the best 750 frames out of 2000.

Selecting Input

The first step is to select the file you wish to process.  For webcams, you will use an .AVI file.  Click on the Select button and navigate to the folder where you saved your video file and open that file.

Note:  Registax sometimes gives error messages with AVI files containing more than about 2200 frames.  Keep the input file to this size or less to avoid getting an error.

Above: The Mars video clip selected in Registax.

Initial Frame Selection

There is a slider along the bottom of the window that allows you to select which frame from the video clip to use for the starting frame.  Move the slider slowly along and watch for a fairly sharp frame or one where the planet looks the most round (sometimes poor seeing squishes the planet).

Initial Settings

Normally you will check the Color Processing box for planetary images.  For lunar images, you may leave this unchecked since the moon shows little color.  For color images, the LRGB option can be checked as well.  The differences the LRGB processing makes can be subtle, but there are benefits in noise reduction and in making sure bright highlights do not appear overexposed.  There is no real disadvantage to checking the LRGB box since the processing time is the same and you don't have to use the LRGB processing if you like the default settings, so it is recommended that you go ahead and check the LRGB box.

The Processing Area can be changed if desired.  If your image is very large, you can select a subframe to be processed.  This speeds things up by just processing a small section of your larger image.  You can see results faster and decide on the processing that you like and then process the final larger image after choosing the desired setting.  For webcam images, which are generally very small, the default 512 pixel setting is usually fine.  In 640x480 resolution mode, the edges of the image will not be processed.  For planetary images, this is fine, since the planet only occupies the center of the field.  However, for lunar or solar images, where the object fills the image, you will want to increase the Processing Area to 1024.

The alignment box can be changed depending on the subject.  If a planet appears quite large in your image, 128 or 256 should be used.  For small targets, or for aligning on a small crater on the moon, use one of the smaller boxes to keep other craters or features from interfering with the alignment.  For the example used here, the 256 pixel setting was used.

Move the cursor over the image and a box is drawn.  Make sure the box is larger (but not too much larger) than your chosen alignment feature.  For the image of Mars, the whole planet can be used.  The 256 pixel box surrounds the entire planet.  Click on the center of the disk of the planet and the Aligning window will appear.

Above:  Selecting the alignment feature, in this case the whole planet.

Aligning

After selecting the planet the FFT Spectrum window appears showing the alignment feature being used.  In the case of a planet, you should see a distinct small red spot in the center.  In the case of a lunar image, where there could be multiple alignment features to choose from, make sure only a single distinct dot is visible.  If more than one red spot is seen, Registax may not know which feature to use for alignment.

Increasing the FFT filter number will decrease the spot size and potentially increase the accuracy of the alignment.  Usually a good rule is to increase the FFT Filter until any colored areas around the edges of the window disappear, leaving just a single dot in the center, as shown below.

Above:  FFT Spectrum

Align Window Options

The only setting to change here is the Quality Estimate.  The Gradient method works well for planets.  Setting the Lowest Quality to 80 is a good starting point.  This means all images with a quality level greater than 80% will be aligned and included for further processing (or later rejection).  If your entire set of frames has a quality greater than 80%, all the images will be included in the alignment.  Normally, with good atmospheric conditions during imaging, this will be the case.  However, for less than ideal images, the full range of quality may go below 80%.  It might be best to lower this value to 60% as a starting point.  You can always raise the quality settings in the next step, but it is best to try to include the entire range of images at this point since you don't know what the quality level will be.

Click Align to begin the automated alignment procedure.  The bar at the lower left corner of the screen will indicate the progress of the Alignment function.

After running the initial alignment, the Align window displays the Registration Properties graph.  This graph shows a red line displaying image quality.  The graph is not numbered but each horizontal line represents 10%.  In this example (see figure below) the image quality line starts on the upper left at 100%, then drops off to about 75% on the right.  The green line indicates alignment error, which will be discussed more below.

Note:  You can get a good feel for the seeing conditions and the likely final quality of your image from the quality graph.  Average conditions tend to produce images with a lower quality range of 75% or so.  Poor conditions might lower the quality to 60%.  In spectacular conditions, the worst images might be 95%, indicating excellent seeing (and excellent optics).

Above:  Registration Properties graph showing the 40% cutoff described in the text

You can now select which frames will be used for the final image.  The best frames are on the left end of the graph and the worst images on the right.  By moving the green line (using the slider at the bottom of the Registax window), you determine the quality cutoff.  Everything to the left of the green line will be used for the next step (stacking).  A good rule of thumb is to use the best 40% of the images.  (See the box below for more details on choosing the quality cutoff.)  To select 40% for the cutoff, move the green line left until it reaches the fourth vertical line on the graph, the 40% mark.  As you move the slider, note the number of frames being indicated along the bottom of the screen (listed as stacksize).

Find the point where the image quality line crosses the 40% frame line.  In this example this is at about the 85% quality level.  This means that the best 40% of the frames are 85% quality or better.  In this example, this selects the best 800 images for stacking.

Optimization

Usually the default settings in the Optimization window are sufficient.  Using the Resampling or Drizzling settings will enlarge the image.  Drizzling will enlarge the image after the next Stacking step.  Resampling, on the other hand, enlarges each individual frame before optimizing.  This allows a better optimization but requires much more processing time.  Many imagers will Resample to get a more accurate alignment, then resize the image back down to the original size later.  Lanzcos and Mitchell methods seem to produce the least noisy results when Resampling.

Artifacts can creep into the image if it is resized, so my recommendation would be to leave the image size alone and resize it later, if necessary, in Photoshop.

Above:  Optimization window settings

The bar at the lower left will indicate the progress while the images are optimized.

Above:  Registration Properties window after optimizing.

The next step is to Stack the frames.  Click on the Stack tab to select the next window.

Stacking

At the top of the screen you will see a checkbox to open the Stackgraph.  The Stackgraph basically looks like the Registration Properties graph, but there are two sliders that will be used to make the final selection of frames to be included in the final image.

Above:  Stackgraph with cutoff sliders.

The Stackgraph contains two cutoff sliders.  One for Difference Cutoff on the vertical axis, and one for Quality Cutoff on the horizontal axis.  As you adjust the sliders, notice that the number of frames to be included in the final image changes.

Stackgraph Settings

If you want to reject some of the images that are not registered as closely, you can bring the Difference Cutoff slider down.  This will reject images with a greater alignment difference.  Bringing this down to 90% rejects some of the more poorly aligned images.  You can also limit the frames by sliding the Quality Cutoff slider to the left to remove some of the lower quality frames, but this was already done when we Limited the frames after aligning them.  The settings used here reduced the number of frames for the final image to 750.

Click on Stack and Registax will stack the images and display the final stacked image.  Now it needs to be processed.  Click on Wavelet Processing to begin enhancing the final image.

Above:  On the left a single raw frame from the video clip, on the right 750 frames stacked.  Note the significant reduction in noise.

Wavelet Processing

Notice that the final stacked image is pretty blurry.  However, it shows significantly less noise than the original.  The image now has 6 layers that can be adjusted using the Wavelet sliders at the right of the screen.  Every image is different, and some experimentation is definitely necessary.  Normally, most of the adjustment is done to layers 2 and 3, with just a little on each other layer.  Try running the sliders to the right to see what happens.  If you start to see noise creeping into the image, back off on the slider.

Tip:  One of the most common errors in processing planetary images is to oversharpen.  This leads to a noticeable halo effect where there is a dark line around bright portions of the image.  This is most noticeable on Mars around the edge of the planet.  On Saturn it might show up where the rings meet the edge of the planet, and it might appear as artifacts around high-contrast craters on the moon.

Above: Wavelet Processing settings for the Mars example

A handy feature of Registax is the ability to save wavelet schemes that you like.  To do this, set the wavelet sliders how you like then click on the Save Scheme button below the sliders.  I find that settings tend to apply to a given planet, so I have schemes named "Mars" and "Jupiter" and so on.  Once you find a good setting you can easily use it on a new image by clicking the Load Scheme button.

Other Processing

There is still a little work to be done, as can be seen in the image above.  Four other processing features in the Wavelet Processing window were used on this image:  Contrast, Gamma, Histogram, and the L=R+G+B mixer.  Techniques for using these features are outlined below.

Contrast

The stacked image tends to be a bit bright, especially after Wavelet processing.  Usually lowering the brightness and contrast is helpful.  For this image the Contrast had to be lowered quite a bit.  This brings detail back into the overexposed bands near the middle of Mars.

Above:  Contrast/Brightness settings

RGB Balance

This feature is ideal for controlling color balance.  The red, green, and blue Colorweight sliders adjust the color balance of the image.  This image of Mars is a little off in color.  Adjusting the color balance is a matter of moving the sliders until the planet appears as you want.

Above: Histogram settings for the Mars example.

L=R+G+B Mixer

This adjusts the mix of each color channel in the luminance channel of the image.  It does not affect the actual color of the image.  It just changes how much weight each color contributes to the overall intensity of the image.  Changing these sliders can help enhance certain details.  For Mars, increasing the blue slider tends to increase the brightness of the clouds but some contrast is lost on surface features.  For this example, red and green were increased to bring out more detail.

Above: LRGB Mixer settings for the Mars example.

When finished with the Wavelet Processing window, click on the Final tab to finish processing the image.

Final

Final color adjustments can be made in this window using the Hue and Saturation sliders.  Since we corrected the color using the RGB Balance window in the last step, changing the Hue is probably unnecessary for this image.  Saturation was increased to 10 to help intensify the color.  The rotation dial can be used to align the planet differently, if desired.  For this image, the Flip Y option was checked to put the planet's north pole on top.  The Flip X option would mirror the image, if desired.

The last step is to save the file.  Normally you will want to save your file as a TIFF file.  This will prevent any compression from being applied (as would be the case for a JPEG file).  This is preferable if you plan on opening the file for more processing in Photoshop.  Go to File > Save and select TIFF from the pulldown menu after naming the file.

Above:  The Final window

Post-Processing in Photoshop

Normally, some slight post-processing is done in Adobe Photoshop.  A little more sharpening, noise reduction, and color balance adjustments can be made easily in Photoshop.  Various techniques are possible and the methods used and results achieved will depend greatly on the particular image being processed.  See the Photoshop section of this website for more details on the different processing methods that can be used.

For this image, the Despeckle filter was run twice to remove any remaining noise, a slight Unsharp Mask was applied, and a few adjustments were made using Curves and Levels.  The Saturation was boosted slightly.

Above:  The final results of processing using Registax and Photoshop.  Note the pink-colored dust storm on the right side of the planet.  Image by Rick Auerbach and Scott Tucker.

Return to Software Instructions Page