Proper configuration of the ALIGNMENT TOOL for AUTO ALIGNMENT in PREMIER 9.3 . . .

2017-08-09-171122

All --

I would like to use the PREMIER 9.3 ALIGNMENT TOOL within an APPLICATION but I cannot seem to get the CONFIGURATION right for it to run AUTOMATICALLY.

In the SCREEN CAPTURE below, please see that images C3.TIF and D3.TIF are open within PREMIER 9.3 and that a MANUAL ALIGNMENT has been performed to TRANSLATE and ROTATE C3 to align with D3.

The results are C3 + TRANSLATION (+42,-63) + ROTATION (136) = D3.

I have attempted to start with a fresh copy of C3 and D3 and set the METHOD to FFT CROSS-CORRELATION and then adjust the settings for ROTATION, SCALING, TRANSLATION, and CORRELATION to get the same results AUTOMATICALLY.

Unfortunately my results are either:

1)  A ROTATION of approximately 20 deg and no TRANSLATION resulting in an bad alignment
2)  PREMIER HANGS at 29% or 43% on the STATUS BAR

If a PREMIER GURU would direct me to the appropriate way to configure the ALIGNMENT TOOL, it would be much appreciated.

I have attached C3.TIF and D3.TIF so you will be working with exactly the same IMAGE DATA that I have been working with.

If I need to get C3 within 90 DEG of D3 before attempting the AUTO ALIGNMENT, that is doable.

Thanks in advance.

-- Matt

*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-





Best Answers

  • edited August 2017 Accepted Answer
    Matt,

    Alignment by rotation and translation is not as sensitive as by translation, so if objects look different or close to a circular shape, it may not work, though for other shapes it works well. I've attached 2 test images that are aligned properly using Rotation and Translation with FFT Cross-correlation (Phase correlation). The result should look like this:



    Yuri
  • Accepted Answer
    Matt,

    Here is the screenshot of my options:

    Alignment tool could work better if you would have original image with texture (not just a binary mask).

    Yuri

Answers

  • edited August 2017
    2017-08-10-105249

    Yuri --

    Thank you for your response.

    2 questions . . .

    1) Will you please post a SCREEN CAPTURE of the SETTINGS for ROTATION, SCALING, TRANSLATION, and CORRELATION that you used within FFT CROSS-CORRELATION to get your results?

    2) Do you think that the ALIGNMENT TOOL would work better if the BLOB was replaced by a BOUNDING BOX or an EQUIVALENT ELLIPSE oriented along the BLOB MAJOR AXIS using measurements from the two BLOBS like those shown below?

    Thanks.

    -- Matt


  • 2017-08-10-125643

    Yuri --

    Thank you for your response.

    I'll give your settings a try.

    Due to the OPTICS used to capture the C and D images, the INTENSITY VALUES within the FEATURE BOUNDARIES are very different so I think it would probably confuse the ALIGNMENT TOOL.

    Thanks again.

    -- Matt
  • 2017-08-10-170637

    Yuri --

    Your settings did not produce good AUTO ALIGNMENT RESULTS for the BINARY BLOBS.

    I even tried to get the BLOBS within 90DEG of ROTATION of each other and attempted a TRANSLATION without ROTATION and then a ROTATION.

    Unless there is another option that you can suggest, the USER will need to TRANSLATE and ROTATE manually.

    If there is another option, please let me know.

    Thank you for your assistance.

    -- Matt
  • Matt,

    The settings on the screenshot (default) worked well for my images. There are no other settings in auto-alignment. 
    You may try to use original image (with texture) and see if it will work better. 
    If your objects are elongated, you may try to use Sort function in Count/Size to arrange the object Vertically and then align sorted images just by Translation.
    If nothing works, then manual alignment can be used.

    Yuri 
  • 2017-08-10-190403

    Yuri --

    Super suggestions.

    Since the BLOBS are very similar, I may use the CENTROID to get the TRANSLATION and the MAJOR AXIS to get the ROTATION and then then have the user FINE TUNE if necessary.

    Thanks again.

    -- Matt
Sign In or Register to comment.