| View previous topic :: View next topic |
| Author |
Message |
redcore
Joined: 01 Jan 2010 Posts: 18 Location: prague
|
Posted: Fri Jan 01, 2010 2:56 pm Post subject: big lens distortion problem.. |
|
|
hi there...i am trying to solve camera for this shot (part of imax "space station" movie)..i just would like to do it for myself to gain new skills.
..the track by itself - i dont think its too difficult to match. the only problem is a big lens distortion..and tbh i dont have any info about that shot like lens value, back plate value etc... so i would appreciate any help u can give me, or "show" me the direction i should go.
please download attachment from rapidshare..about 35 mb..400 frames in jpg files
http://rapidshare.com/files/328892711/astro.rar.html
thanks in advance..
rob |
|
| Back to top |
|
 |
Gorf
Joined: 10 Mar 2007 Posts: 111
|
Posted: Sat Jan 02, 2010 9:13 am Post subject: |
|
|
| Tried for ten minutes to download them, but Rapidshare is claiming to be overloaded. |
|
| Back to top |
|
 |
redcore
Joined: 01 Jan 2010 Posts: 18 Location: prague
|
|
| Back to top |
|
 |
Gorf
Joined: 10 Mar 2007 Posts: 111
|
Posted: Sun Jan 03, 2010 10:44 am Post subject: |
|
|
I'm not an expert - not even a regular user, but if I mess up and use bad terminology or am just plain wrong, someone will dash in and correct me I'm sure...
It's definitely fisheye, but it's not a spherical lens - there is further distortion at the edge of the lens, so I've zoomed in by 25%. As far as I'm aware, SynthEyes can't get rid of this second level of distortion.
You can download my initial solve here - Like I said, I'm still a newcomer to this. I bought SynthEyes a couple of years ago and am still building up the courage to make use of it in a film of my own. |
|
| Back to top |
|
 |
redcore
Joined: 01 Jan 2010 Posts: 18 Location: prague
|
Posted: Sun Jan 03, 2010 11:37 am Post subject: |
|
|
thanks gorf for your solution....
looks very good to me..so basically u tried to rid of that lens distortion first and then tried to solve camera...and yes, i have seen that further distortion as well..but lets say i have a footage to solve..so i have to rid of that distortion first, then solve camera, then make some effects (whatever) in 3d soft, after i have to bring it back to syntheyes and apply distortion again, so i can get the right picture...hmm...i have watched tutor about it but still..have to have a look whether i can save 32 and 16 bit tiffs from syntheyes + alfa channels...we have syntheyes at work so when i have a time i just try to do some things...still have to learn a lot.
thanks again gorf...
bye
sorry for my english |
|
| Back to top |
|
 |
Gorf
Joined: 10 Mar 2007 Posts: 111
|
Posted: Sun Jan 03, 2010 12:28 pm Post subject: |
|
|
To be honest, it was my second solve. I always just throw the footage in SynthEyes, let the cache fill, then do an auto solve. I then make a note of the pixel error and the field of view, just for a benchmark.
While still in this first session, I used the lens distortion panel to get a figure for the lens, just by using the slider until the three bars in the picture were straight. A quick scan of the footage showed that there was another distortion at the edges of the frame, so I guessed that 25% zoom would bypass that.
I started a new session, and put those figures into the image preprocessor and did another auto solve. I got just over one pixel of error, and a quick scan of the new solve showed a tracker following the neckline of one of the astronauts, so I deleted the tracker and refined the solve, getting the pixel error under 1.
It's possible it could benefit from more work, or the use of supervised trackers.
You might also find that the way I've arrived at a solve is no good when it comes to integrating with CGI imagery. It depends if you want to have distorted CGI or undistorted (but reduced) source footage.
Good luck - don't worry about your English. I am in awe of anyone who's made the effort to learn a second language, especially English. |
|
| Back to top |
|
 |
|