Hi,
I'm trying to find out if OpenCV is the right solution for this. I'm attempting to stitch VR video on a Java server.
Currently I extract all frames with FFMEG and stitch using PtGui, before combining the stitched frames and audio to make the 360 video.
I'm hoping I can do something similar on the server using OpenCV.
The stitch would involve 2 fisheye images that are always in the same position, so OpenCV wouldn't need to find control points etc. I'm hoping the data from a PTGui template could be used to position and warp the images for the stitch.
Are there any examples out there of something like this, or at least stitching 2 fisheye images?
Here are some example images and the data from PTGui that stitches them
> #-dummyimage
> # The following line contains a 'dummy image' containing some global
> parameters for the project o w1 h1 y0
> r0 p0 v220.1167582353273 a0
> b-0.1083809038961362 c0 f2 d0 e0 g0 t0
> #-imgfile 2304 1296 "00001_1.jpg"
> #-metadata -1 -1 -1 0000-00-00T00:00:00 3*8 0 -1 -1 -1 * *
> * curve -1 * T * 13105066406946011
> #-viewpoint 0 0 0 0 0
> #-exposureparams 0 0 0 0.355634835327623
> #-sourcemask o f2 y-120.3422990063066 r-0.9297019831167859
> p4.326622464952266 v=0 a=0 b=0 c=0 d=0
> e=0 g=0 t=0 C362,1972,-145,1465
> #-imgfile 2304 1296 "00001_2.jpg"
> #-metadata -1 -1 -1 0000-00-00T00:00:00 3*8 0 -1 -1 -1 * *
> * curve -1 * T * 13105066443910340
> #-viewpoint 0 0 0 0 0
> #-exposureparams 0 0 0 -0.3556348353276229 o f2 y61.44044174667408 r-88.7859508891198
> p-1.528835919598066 v=0 a=0 b=0 c=0
> d=0 e=0 g=0 t=0 C362,1972,-145,1465
