Hi Oskar, On September 20, 2010 09:38:48 am Oskar Sander wrote: > Nice walk though!
thanks,
> * Is there something inherent in CP detection that is thrown of by linear
> mosaics? My experience is too that it doesn't work well, but i tho
> thought it was my low-contrast low-light and monochrome application (very
> dark underwater scenes) that was throwing the search off.
depth (pun intended). The CP detectors will detect features that look the
same, but they still can't tell if all of those features are on one plane.
> - One experience that I have form manual CP detection is that IF your
> mosaic subject is not a flat plane, you need to analyze the images and make
> sure that the CP are selected in the image plane you want
exactly that. My experience with building fronts is that even small relief
makes a big difference.
> When you are laying the images out manually, are you
> just picking a number?
I help myself visually with the feedback of the fast preview.
> The X,Y,Z parameters are a scaling factor of the
> radius to the unit panosphere I understand.
the unit is relative. The way I understand it is that with X you shift the
center of the panosphere left and right. With Y you shift it up and down.
with Z you shift it forth and back ("zoom"). Each image is on its own
panosphere that gets projected on the mosaic surface. Knowing this, playing
with the numerical values in the Images tab becomes a "fly by wire" exercise
where you zero in on an approximate position, before optimizing it with the
optimizer.
> Maybe a simple calculator
> button next to the X, Y X fields to place the image a whole number of image
> lengths away from the anchor could be useful.
The image length is constant on the mosaic only if you shoot from the same
distance and angle - a special case. I don't see it as being helpful to me.
I rather have some sort of 3D joystick/trackball. one joystick would do X/Y
the other would do Z/roll and the trackball would do pitch/yaw.
> * I like your in-line lens calibration in the work-flow, I have to try
> that to se if I get better results!
It's a fortunate coincidence that I have the necessary features in the
picture. It's really much better to work with calibrated lens profiles in the
first place, but it has been so long since I've been serious about pano
stitching that I have not bothered to calibrate this lens I bought about a
year ago (Sigma 15mm fisheye).
> Another question that I think I asked before but can't find the answer to
> is: When you have generated a mosaic project, is it possible to generate
> an output where all image seams are highlighted?
If my memory does not betray me, there was a debugging version of enblend that
would do just that, but I can't find any references to it right now.
Given the extreme sensitivity to depth/perspective, it is anyway better to
place the seams manually, using masks. That's the most time consuming part of
the work. Fiddle, fiddle, until it looks acceptable.
Yuv (still fiddling)
signature.asc
Description: This is a digitally signed message part.
