Forum MicMac

This forum is dedicated the the community of MicMac users


All times are UTC + 1 hour



Post new topic Reply to topic  [ 5 posts ] 
Author Message
Offline

Joined: Sep 2017
Posts: 14
Gender: None specified
PostPost subject: Camera calibration
Posted: 12 Sep 2020, 09:52 

Dear MicMac users,

trying to fine tune my camera calibration (Mavic Pro 2 Hasselblad camera), I am expecting to find coherent results between two approaches :
1. gave a try a the Camera calibration toolbox for MatLab ;
2. performed several Micmac runs on a set of images.
The main issue I cope with is, I guess, a units concern : while focal length, PP and Cdist are similar, I am having trouble with distortion parameters, i.e. they differ from each other by several (well, many) orders of magnitude, e.g.
AutoCal_Foc-10790_Cam-Hasselblad.xml :
Code:
...
<CoeffDist>-8.68814596800474489e-09</CoeffDist>
<CoeffDist>1.03782603772602363e-16</CoeffDist>
...

Calib_Results.mat :
Code:
...
%-- Distortion coefficients:
kc = [ -0.131544042569461 ; -0.044695677971104 ; 0.001131062101727 ; -0.000547469947769 ; 0.000000000000000 ];
...


Any advice ?

Vincent.


Top
  Profile 
 
Offline

Joined: Jul 2011
Posts: 1050
Gender: Male
Age: 31
Posted: 16 Sep 2020, 11:13 

Hei,

Look up this : https://www.mathworks.com/help/vision/r ... nsics.html

And the MicMac doc section 15.2.2

Seems to be the same formulas (as expected). But it seems that the Matlab option gives you the inverse distortion ( Distort=f(ideal) ), while MicMac gives you the direct ( ideal=f(Distort) )

Hope this helps!

_________________
Join the MicMac community on Reddit : /r/MicMac/
Don't forget to check the wiki : http://micmac.ensg.eu


Top
  Profile 
 
Offline

Joined: Sep 2017
Posts: 14
Gender: None specified
Posted: 16 Sep 2020, 12:42 

Thank you Luc for your reply, I'll look into this.

I recently stumbled on a notation correspondence between the Matlab calibration tool I use and some previous work by Heikkilä et Silvén :
http://www.vision.caltech.edu/bouguetj/ ... eters.html

looks like I should divide kc(1) and kc(2) respectively by f^3 and f^5 to get back k1 and k2 coefficients: in my set of data, values then still differ but orders of magnitude fit quite well. Did not go further for now.

On another hand (sorry for my ignoramus questions), do you consider it's fondamentally a good practice to calibrate a camera, or do you consider it's ok to run micmac on raw images? The point is Mavic Pro 2 perfoms its own distortion correction while translating raw (.dng) images to jpg format, which I guess is certainly confusing for micmac calibration routines, isn't it?
Is it worth trying to (i) calbrate my camera (ii) un-distort raw images out of micmac (iii) run Tapas process on the latter, relying on a set of parameters corresponding to an ideal camera?

Vincent.


Top
  Profile 
 
Offline

Joined: Jul 2011
Posts: 1050
Gender: Male
Age: 31
Posted: 24 Sep 2020, 14:24 

Hei,

Tapas should be fed the rawest image possible, as any previous correction might actually create some weird edge effects that are harder to fit than the raw distortion pattern. Pre-correction should only be considered if the distortion pattern is known to be very non standard (like if the camera is not using a standard lens but something more exotic). Tapas includes a very large variety of options to fit even some rather weird distortion patterns (fish eye, non centric...).

_________________
Join the MicMac community on Reddit : /r/MicMac/
Don't forget to check the wiki : http://micmac.ensg.eu


Top
  Profile 
 
Offline

Joined: Sep 2017
Posts: 14
Gender: None specified
Posted: 25 Sep 2020, 15:05 

Thank you Luc for these clarifications,

V.


Top
  Profile 
 

Who is online

Users browsing this forum: No registered users and 1 guest

Permissions of this forum:

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
Post new topic Reply to topic  [ 5 posts ] 


Créer un forum | © phpBB | Entraide & support | Forum gratuit