headerphoto
 
Setting Alignpoints

    In RegiStax V4 the user allready had several ways to set Alignpoints. In RegiStax V5 we have extended the options for the user. There are 3 groups of setting alignpoints, by hand, let RegiStax estimate them or a mix of both. In the text below you can see more about this.
  1. Setting Alignpoints by hand
  2. Let RegiStax estimate Alignpoints
  3. A mix of both

    alignmethodsPreparations
    The image to the right shows the Align options area at startup. Before you set alignpoints you need to decide to either go with the default (1 alignpoint) or use multi (many alignpoints) or no alignpoints at all.

    Before you start setting alignpoints, select a good frame that looks better than average. If you have a large amount of drifting dont select a frame that is too far into the set of frames as this might give problems during the start of aligment.
    In RegiStax V4 the referenceframe was allways the frame the user had located in the sequence of images to position the alignpoints on. In RegiStax V5 we allow the user to load a referenceframe which serves two purposes, first off all it allows the user to find the best alignpoints better (for instance using the estimate-alignpoints function) and secondly it can be a far better reference for the topology of the image. 

    When using the default method, the user can only set one alignpoint. The user also needs to set the alignboxsize (the area used to estimate quality and estimate the alignment). The last used alignboxsize will be default at startup. If the "User set Alignbox" is checkmarked the size of the alignbox set by the user will be used. The user will see -when positioning an alignpoint - a double box in this case. The outerbox is the nearest default alignboxsize, thats the area that will be used for image-quality estimation and the inner alignbox is the area that will be used for alignment. This allows the user to use small alignboxes that align properly but also will estimate image-quality based on a good amount of pixels. Most other features are merely usefull in multi-point alignment. The examples below are therefore aimed at explaining the ways to set multipoints.
     

    1. Setting Alignpoints by hand
    This is the method we allready had at hand in previous versions. The user selects Multi in the alignmentmethod box and a multi-alignpanel pops up.

    This panel will show the alignpoints and allows the user
    alignmulti1 to load alignpoints and save them also for later usage. A new feature is the maximum-drift setting.  This is very usefull in multipoint-alignment. If any of the alignpoints drifts away more then the user-set amount of pixels (relative to alignpoint# 1) it will be restricted to the shift of alignpoint #1.  We refer to Alignpoint #1 as the Master-Alignpoint. Below the save/load point buttons a new control is visible. When openend (use the arrowbutton on the rightside) it will show the available saved sets of alignmentpoints that have the same name as the file that is being processing. So if you work on an AVI called MOON.avi and save your alignpoints with names like MOON100points or MOON4x64 Registax will show those alignmentpoint-files since they have MOON in their name. Everything else is much as before.

    Now the user can set points by hand, simply set the alignbo
    alignmulti2xsize and position a point at a location that looks rich in contrast. After setting a point the panel shows the X,Y position and the size of the alignbox (FFT). The user can set many boxes of the same size but also can mix them, just select another alignboxsize and position the larger/smaller boxes. Its often wise to make sure that the master-alignpoint (#1)  is located at the richest area in terms of contrast. Also the alignbox of the master-alignpoint should not be small, 64x64 is the smallest usefull for this point.  The user can also use the User-set alignbox sizes mixed. In the image to the right you can see we set alignpoint 1 using a 64x64 box, alignpoint 2 has a 128x128 alignbox and alignpoint 3 has something special. The alignbox for alignpoint 3 is 64x64 in size but it also has a ZP (zeropadding) value of 11. This means that on all sides of this alignbox 11 pixels will not be used during alignment. The alignbox is therefor 64x64 for quality estimation and 42x42 for alignment. Smaller alignmentboxes are considered better to align on finer details. When you use a lot of user-set alignpoints its wise to record them, when saving them (save points) be aware that RegiStax also saves the number of reference-frame you have been using to position. So if you load the points again they will be at the same location and on the same imageframe.


    2. Let RegiStax estimate Alignpoints
    In RegiStax V4 we had a way to estimate alignpoints, the method uses was however not very powerfull. In RegiStax V5 you will see things are a lot better.
    To estimate alignpoints you need to select the ESTIMATE tab in the
    alignmulti3 multi-alignmentpanel (next to alignpoints). When pressed you see the panel to the right. There are many controls you can use and most of them are recorded, so next time you dont need to set them again ! Starting from the top we have the button "SET MASTER ALIGNPOINT" .When pressed the user can position a master-alignpoint based on the size of the alignbox or user-set alignbox. This point will alignpoint #1 and when you estimate the other points they will not overwrite this point and also keep their distance. Next control is the Method, we have method1, method2 and matrix. The matrix will setup a set of alignboxes arranged as a matrix, the user can set both the number (X,Y) and the size of the alignboxes. The default X,Y of 2 will setup a scheme of 2 x 2 = 4 alignboxes all 64x64 pixels large. This can sometimes be usefull. But most of the time we will be using method 1 or 2, they work slightly different and the user has to find out what works best for the user's images. The method uses the threshold, detectionlimit, radius to calculate possible alignpoints. Increasing the threshold (range 2-15) or detectionlimit will lower the number of alignpoints. The detectionlimit will block setting alignpoints on pixels with an intensity lower than the detectionlimit. The radius calculates the alignpoint strength based on an area around the pixel. A larger area will create less alignpoints. The minimum distance and free space to edge control how close alignpoints are allowed to be near to each other. When calculating the best alignpoints RegiStax orders the points on strenght and then works from the strongest to the weakes found alignpoint. If the next alignpoint-candidate is too close to either the edge of another (stronger) alignpoint it will be neglected. Finally we also have "variable sized alignbox". In this mode RegiStax will mix alignbox sizes to create a scheme with minimal overlap between alignboxes. WHen you are not happy with the estimates, simply change the settings and press estiimate again.

    This example shows how the schem looks when not using variable sized alignboxes. Notice that alignbox1 is highlighted.
    alignmulti4


    This example shows how the schem looks when using variable sized alignboxes. The overlap between boxes is minimal this way. In large smooth areas few alignboxes appear and near to rich areas many boxes appear.
              
alignmulti5






















            3. A mix of both
        
A user can simply first set a group of alignboxes/points using the estimate method and then still add several points by hand or remove points. Adding them is like we are used to and they are based on the alignboxsize set or the user-set alignboxsize. To remove an alignbox you have to double-click using the RIGHT-MOUSE button.