middle and pan are taken to check the movement of the reference point when turning to the right. Might be he should have moved a little bit further to the overlap area. The nails don't seem move in these example.
For LRS he has focused down to the logo. Would have been better if we could see the focus point from the camera.
RC, you get better results about a "moving reference point" when you check it to something which is far away, like a house etc.
Last edited by hindenhaag; 08-08-2011 at 01:06 AM.
Sorry about that guys! Look like I someone how saved pan shot over the middle. I attached a new link with some quick pictures I took this late evening. I apologize for the composition it was the best I could do this late at night. But you should be able to see what's going on...
The new pair of images are fine. What exactly is your main concern? There's a small amount of parallax shift visible, but probably not enough to prevent a good stitch from being obtained for typical panoramas. You should really be checking for parallax in the same conditions as for shooting a panorama - i.e at yaw positions -/+ half the normal yaw increment, e.g. -15° and +15°.
In what context does "focus point of the camera" appear?
I mentioned that I can see my "focus ring" while focussing to the logo for example.
He has problems with his setting.
Checking the two new pics, you can see that the top of the nail in front moves with the movement of the camera: in the first, the left side of the top ends in the middle of the wood on the right side. On the second, it has moved to the line in between the two boards. RC, this means your reference point moves with the movement of the camera and as consequence you have to move forward on the upper rail.
BTW: I have learned from John, that we always shoot the pano by moving to the right. This is the normal way to do it. This will keep your screws fixed, turning to the left may loosen your screws and the head will move.
Last edited by hindenhaag; 08-08-2011 at 03:38 AM.
Issue 1: My main concern is that even though it appears my pan head is calibrated correctly I'm still seeing parallax in my panoramas. It's primarily in the set of images that are taken at a 45º tilt. From my understanding if pan is calibrated correctly so should tilt. If that's the case how would I adjust for tilt?
Issue 2: I'm currently shooting 12 images at 30º apart at varying angles of 90, 45, 0, -45 and -90. The issue I'm having is aligning the 90º straight up shot. PTGUI has issues plotting control points between the image 0 and 1. Mainly because there is not enough information to plot points between images. " this is not related to the parallax issue but I just thought I would bring it up"
I wonder how you "see parallax in your panoramas"? It would be a good idea to supply the 12 images at pitch 45 so that the real problem can be investigated. They need not be 2MB each if you choose a quality setting equivalent to 9 in Photoshop, at which they should be around 300KB. Irfanview will resave in batch mode quite simply.
If there's no detail for control points at the zenith, then you should be able to position the zenith image manually. You should know the nominal yaw, pitch and roll angles of all the images. If you enter these to initialize the project, the zenith will remain in position reasonably aligned without control points.
I had to key in the lens parameters, assign control points to tie in the zenith and do just a touch of masking between a couple of images (in order to move the seam into the shadows and such). Otherwise PTGui 9.1 beta 1 stitched this one just fine. I see no alignment or "parallax" issues.
Ryan, I stitched it ok. No masking applied. I used temporary line control points (t3) to align the ceiling. I haven't done an intensive search for stitching errors, but it looks reasonable. I did make an adjustment to the white balance when I resized the output down. I put the project file and 5000x2500 stitch at: http://www.johnhpanos.com/project2jh.zip
t3,t4,t5 ... tn points can be assigned along line features that cross from one image to another such as the beam across the ceiling in your images. The PanoTools optimizer (not PTGui's normal optimizer) will then try to get all those points in straight lines across the output area and thus align those features. However, in the equirectangular projection, those features should be curved, not straight, so it's not apprpriate to use that technique without some trickery.
The trick is to shift the zenith down to the middle of the output area and then switch to rectilinear projection. The horizontal and vertical fov needs to be reduced to 120, say. Then, the line features in the ceiling will be straight and the optimizer has a sporting chance of getting the rows of tn points aligned.
So, after the alignment of the zenith image has been successfully performed by PTOptimizer, the panorama projection can be switched back to equirectangular and the zenith shifted back to its usual position. The tn points are no longer needed and can be deleted. The panorama can be levelled with t1 points using the normal optimizer or the levelling option on the Panorama Editor window without disturbing the alignment of the images.
I did not see any t3, t4 or t5 control points used in the project file. I did however, see that you used a couple of t1 control points to adjust the vertical alignment in the panorama. Which proved to be very useful when aligning the ceiling. I used the project file you provide me as a learning guide to see where I went wrong. My biggest problem was that I wasn't using any of PTgui plugins. Having the right optimizer plugin with the correct settings, fixed a lot of issues.
Ryan, there were no tn points in my project file as they were temporary. Once the equirectangular image was resored to its normal orientation, they were not needed and I then added the t1 points to level the panorama. The fairly new version of PTOptimizer available from the PTGui web site is useful occasionally, but for the most part, PTGui's own optimizer is all that's needed.