Notice: This page will be redirected to the Epic Developer Community. Spend a few minutes to get familiar with the new community-led support today.

Retry specific images in alignment?

Answered

Comments

11 comments

  • Avatar
    Wishgranter
    Hi Neil

    try recommend to change only this values:

    Max features per image - amount of all detected points in image
    Preselector features - how much from the max features will be used in alignment
    Detector sensitivity - sensitivity of point detection

    try use 40k/20k then 80k/40k and eventually 160k/80k

    If want not align all the imgs can use ControlPoints to put it together
    https://www.youtube.com/watch?v=d8naLEtLqDY

    But even from my experience if get say 5% not aligned imgs its OK. best if try to understand what was the cause ( low points detected= featureless subject or not good angles between the images or etc )
  • Avatar
    Neil Nafus
    Wonderful! Thank you for the additional information and video. The video helps quite a bit!
  • Avatar
    akiro
    So "160k/80k " means
    Max features per image = 160k
    Preselector = 80k
    :?:
    And does "Detector sensitivity" has to be higher or lower to get more images aligned?
  • Avatar
    Wishgranter
    Hi Akiro

    Yes you understand it right
    So "160k/80k " means
    Max features per image = 160k
    Preselector = 80k

    the detector sensitivity is useful when say set 160k points and you get say 100k points per image detected ( "every" img get just around this points detected ) so you can set higher sensitivity and it will try detect more points.

    But more detected points with higher sensitivity mean it will detect even "weaker" points" so its cannot be directly said it help better align.. because weaker points mean weaker alignment results..
  • Avatar
    Götz Echtenacher
    Dang, and I played with the mpx value instead of the preselector.
    That explains why it didn't improve much. :-)
    What does that do then - is that points per megapixel?
    So is the limit whichever value is reached first, per image or per mpx?
  • Avatar
    Wishgranter
    Hi Götz


    What does that do then - is that points per megapixel?

    Max features per image = 160k its the MAX points per image :D
    Preselector = 80k = it take from the 160k points 50% = 80k points so its take the "better" 80k points

    max features per Mpix = default value is 10k, its say as its name how much points can be detected per Mpix.
    if set say 20k it will try ( and highly probably NOT success except the image have a very features rich subject surface texture there ) detect 20k points in one Mpix so it sooner fill the limit on max features per img = 160k/20 = 8Mpix of whole image so it can be just small part of image with detected points...

    if set higher it detect a LOT of ( not useful ) points in small areas that will cause alignment stability issues = harder to align properly. so its advisable to not to play too much with this settings at all.

    if get alignment issues PM me with screenshot of the scene so can give advice what can help out...
  • Avatar
    Götz Echtenacher
    Hey Wishgranter,

    your offer is much appreciated!
    Since I understand the variables better now I am confident that I will be able to get better results on my own.
    If not, I might still get in touch... :)
  • Avatar
    Wishgranter
    OK.. :D

    feel free to ask if stuck somewhere...
  • Avatar
    Götz Echtenacher
    Holy Moly, that was it!
    Set to 40k/20k and got 399 of 404 in one component!
    Yay!
    Now I will do the same with the main set...
  • Avatar
    akiro
    Götz Echtenacher wrote:
    Dang, and I played with the mpx value instead of the preselector.
    That explains why it didn't improve much. :-)

    Same here... :mrgreen:

    Thanks again Wishgranter! It works great now! 8-)
  • Avatar
    Tom Foster

    Brilliant insight and the only explanation I've found of Max features per mpx, which I've found has no effect - here's why.

    Is it fair to say, as guideline, that Max features/mpx should be based on:

    Max features/img / Image size mpx = Max features/mpx?

    e.g. 160k / 20.2mpx = not more than 7.9k Max features per mpx?

    Then the Align default settings of 10k Max features/mpx and 40k Max features/img, would be right for images of not less than 4mpx?

    And that with Align default setting of 10k Max features/mpx, with 12mpx images Max features/img should be not less than 120k?

Please sign in to leave a comment.