Is there any chance "Free Nodal Point" will be made available in a future release of Autopano Pro, or is this going to remain an Autopano Giga option?
The reason I'm asking is that like [bo], I do KAP and have the occasional need to stitch sets of orthogonal images into a single planar image. I have a license for Autopano Pro, and would like to continue using it. But if upgrading to Autopano Giga is the answer, so be it.
As I am curious to the result of the “free nodal point” possibility in the test version of APG I did some tests.
I used 9 horrible pano's, consisting of 3-14 pictures., all handmade, most with a compact camera, some even from a moving and swinging skilift.
All pano's were processed in the test version with and without free nodal point setting.
All pano's could be stitched without free nodal point setting, RMS 2.2 up to 4.0
All pano's could be stitched with free nodal point setting with a RMS that was 0.2 -1.8 better. Also the output in pixels was different (mostly fractional smaller) and the “shape” was somewhat different.
I further found out that if I processed the pano's in APP 2.0.9 the result is not the same, f.i. 14 pictures in APG (Without Free nodal) RMS 2.95, in APP RMS 8.50.
My conclusion up until now (based on this limited test) is that the use of the free nodal point gives a better RMS value of 0.2 – 1.8 points.
Also without free nodal, APG can handle all the terrible pano's without stitching problems.
APG is better than APP 2.0.9 (stitching no problem, but worse RMS).
I know that this is not a scientific test but I am wondering are there more people who have tested APG and have a more or less equal experience.
I still wonder what the influence is of the free nodal setting in terms of what it is doing, but until now my conclusion is that it works.
Leave the question if it is implemented in the actual version will it also be available in APP, then buying APG only for this item is a little bit to expensive.
PS I also used all the horror pano's in an other stitcher and most of the time I got the message that it was not a pano. (It are really horrer pano's)
In the 4th Autopano User Meeting, I've made intensive use of this 'Multiple Viewpoint' feature. It works well on 'aerien photo' ( helicopter, balloon, skite, etc ), linear panorama ( walls, painting, etc ), as well as on handheld panorama ( even 360 fisheye panorama ).
The alpha 1 doesn't at all reflect the versatility of this new feature. I'm afraid, you'll have to wait for beta 1 which has really the feature working well.
As I am very interested in the NP correction I did further tests with a specific panorama.
Input: 7 Pictures of a fence, with lots of background and depth, by moving the camera linear along the fence, section by section, total 7 pictures. (of coarse by hand)
APP 2.09 Fence stitched with 2 big mistakes and lots of, what I call doubling s, same subject on two positions. (RMS 4.07)
APG, no NP correction, fence with 1 small stitch fault, background 1 doubling, but acceptable. (RMS 3.8)
APG with NP correction, fence ok, one doubling in rear but acceptable. (RMS 2.6)
I was pleasant surprised by the result as. f.i. same set in PSE 7 stacked all pictures on top of each other as the poles of the fence were identical.
In the pictures the focusing was on the fence and I noticed that APG gave a good result on the fence with only a doubling in the (far) distance. Question has the NP control something to do with focusing and being in or out focus?
I find this option more and more interesting and I am still wonder what is the idea behind it, can it be influenced and what improvements on this already good result can we expect
The alpha version was really an early version of this system. We just finished one the background algorithm needed for that here and now, we can see really the possibilities of that feature. I did intensive testing on panorama with that feature and it is really nice. Balloons panoramas, cartography stitching ( with 200 images ), etc. Everything comes out really nice.
BTW : RVK, if you have some test cases that you want this feature to work on, please share them. I'm dedicated to test that.