Seams in QuickTime and Flash
Disclaimer: I'm a hobbyist with only the most rudimentary graphic editing skills.
Over much of the past year, I made QTVR panoramas with PTMac. My 6 horiztonal shots stitched wonderfully, but adding the zenith often caused errors with seams, in the triangle formed by the zenith shot and the first and last horizontal shots. Usually the error would be a small section off by 1-3 pixels. I'm working on acquiring the graphic editing skills to take care of such errors.
Not too long ago I bought PTGui and remade a few of my earlier PTMac panoramas. In some cases seam errors have disappeared or they have become harder to find because a bit of blurring at a seam makes a 1-pixel offset harder to see. After going through my optimization stages, I usually end up with a message informing me that my final optimization is very good. In all honesty, does PTGui offer a better evaluation, perhaps "excellent" or "panoramic Enlightenment achieved"?
I also have a copy of Pano2VR, although I haven't been using it much, because I need to improve other skills. However, I have converted a few of my remade PTGui panoramas to Flash 9 with the default settings and I have found seams that I can't see in the QTVR version but do appear in the Flash version. Why is this? Am I simply not looking close enough at the QTVR output and thus missing the flaws that show up in the Flash versions?
Re: Seams in QuickTime and Flash
Patrick, thank you!
It's good to know that there is a higher level of stitching. My problem is, quite frequently, the zenith shot where auto-generated points are rarely
Re: Seams in QuickTime and Flash
Hi Christopher
The optimizing report errors can not tell you if the controlpoints are correct,
It can only tell you if the controlpoints match but it may very well be an error of 20 pixels in real image matching.
Perfect match especially at zenith and nadir demands manual controlpoints and manual optimizing.
Hans
Re: Seams in QuickTime and Flash
Thank you too, Hans!
I had gotten frustrated with zenith shots in PTMac (always a small area where seam trouble appeared), which is largely why I purchased PTGui. It has done a better job, for me, with the zenith shot. Yet I am often surprised at how far apart are the control points it generates. I find I must make a lot of manual points and corrections, and sometimes the points I provide seem to make the others worse. It probably all comes down to practice and patience.
I go through optimization in the steps recommended at Kekus: FoV (toss almost CP >5), b + FoV (toss >3 and add if needed), a + c + FoV (add/adjust CPs), d + e + FoV (usually I stop here). I should not repeat the same level of optimization once I accept the results, correct?
It's just under 11MB, but this 3-node tour of Ryozenji Temple is about as good as I got with PTMac, no post-stitching editing. The third and final node is inside the main building (the first two nodes have sky), so it posed more trouble with the zenith shot; you can see the zenith seam trouble on the face of the large lamp to the right and above the praying pilgrim. That is what I want to be able to fix; a later stitching in PTGui did fix it (enough for my amateur eyes, that is), but I've not yet done the other two nodes. Oh, this panorama has sound, probably somewhat too loud.
Chris


Re: Seams in QuickTime and Flash
Hey Christopher:
Every so often I receive a "too good to be true!" message from the PTGui optimizer... :)
I have found that simply hunkering down and manually adding control points to "problem child" images can take care of most misalignments.
Cheers,
Patrick Cheatham
--
CheathamLane
Berkeley, California