Forum: Q & A

Thread: PTgui & Stitching problems

back to threads list | this thread is closed
Search the forums:
Author  Message 
spwd01

Posts: 8
Location: Blooming Grove, NY, United States
Registered: 24 May 2005
PTgui & Stitching problems
Posted: 1 Jul 2005 at 17:23 GMT
updated: 1 Jul 2005 at 20:39 GMT
I am having a terrible time stitching multi-row images with PTgui. The seams never stitch well and straight lines don’t line up.

I am shooting with a 14mm, two rows of ten images at 35degrees. When I optimize I always get a “Very Good”. I try and keep the distance between control points below 4 however some times this is imposable no mater how many times I add and subtract control points. Can anyone look at my panos and tell me what I am doing wrong? Please ignore the color problems I am just trying to get these to stitch before I worry about color and value. I can post the source images once I get them small enough to put on the web if anyone wants to look at them and see if the problem is the images.

www.sarno.biz/qtvr/050527_p1.htm

www.sarno.biz/qtvr/050623_p1top.htm

Also,can anyone answer the following questions for me? I am tired of wasting time with experiments.

What is the max distance the control points should be in relation to the output image size?

What is the recommended output size (l x w) for full screen 360x180.

Those how are doing multi row panos with a DSL, do you use full resolution tiff input images. Or do you down size you images and convert them to 72 ppi before you bring them into PTgui?

Any help would be great. I am going nuts trying to get things to line up and to do things faster.

SP&WD
alert moderator
John Houghton

Posts: 3487
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
Re: PTgui & Stitching problems
Posted: 1 Jul 2005 at 21:02 GMT
Given that there are discontinuities in some of the vertical features like sides of windows, I would suggest assigning some line control points (t3,t4,...) along them, if you haven't already done so. You should also make sure that these features are truly vertical by adding some t1 points and optimize for y,p,r on all images, except you can keep yaw fixed on one image only. And have you optimized the lens parameters d & e too?

There is no particular maximum value for the control point distance. People have different standards. To some extent, it will depend on how accurately you have positioned the entrance pupil of the lens on the rotational axis. But then with some lenses, the position of the entrance pupil varies as the lens rays enter at larger angles to the lens axis. So any position of the lens will have to be be a compromise and absolutely perfect stitching cannot be assured.

For full screen on the web, 4000x2000 should be enough for good quality. 72dpi (or 72ppi) has no relevance to PTGui.

John
alert moderator
spwd01

Posts: 8
Location: Blooming Grove, NY, United States
Registered: 24 May 2005
Re: PTgui & Stitching problems
Posted: 2 Jul 2005 at 16:11 GMT
John,
Thanks for the reply.
How do you create t3, t4 control points along two or four different images? I will try to figure it out but if you have any pointers or tutorials it would help.

I don't believe I have optimized for d & e I will try that.
I have posted the PTGUI projects if those will help anyone to look at them.
www.sarno.biz/ptguiexamples.html

Thanks
SP&WD
alert moderator
John Houghton

Posts: 3487
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
Re: PTgui & Stitching problems
Posted: 2 Jul 2005 at 16:36 GMT
To assign a set of line control points, select the CP Type as New line (t3). Add the two points anywhere along the chosen straight line feature (separated - not on matching points). To do this, you can display either the same or different images in the two control point windows, as needed. Then assign further points along that same feature, selecting CP type line3, repeating as necessary. To assign points on a different straight line feature, select CP type New line (t4). Do a similar process for line4 (and line5, line6 etc if need be). Note: it is important to select only features that would be expected to be straight in the currently selected output projection.

John
alert moderator