PTA 6.x problems with Widefield Lens single Exposure Stacks

Discussion forum for Tawbaware's PTAssembler software, Helmut Dersch's Panorama Tools software and any other photography related software

Moderator: spamkiller

Post Reply
seigell
Posts: 90
Joined: Tue Mar 21, 2006 10:26 pm

PTA 6.x problems with Widefield Lens single Exposure Stacks

Post by seigell » Tue Oct 15, 2013 2:01 pm

After a fruitful Fall Colors trip, I have a large number of Exposure Bracketted 3x sets of Single Image (non-pano) sets - most of them at Wide Focal Lengths (24-38mm @35mm equiv) from a "standard" 15-85mm Canon EF-S on my 60D APS-C DSLR.

Last year's crop from the same Camera-Lens combo had some difficulties while being Exposure Stacked in PTA 5.x - mostly a Bad Horizon Placement about 45* CCW Rotation. (Recoverable - just difficult when Batch Stacking)

This year,using PTA 6.2 (x64) (and tested in older v6.02_x86 - I no longer have PTA v5.x), I'm encountering the nearly always fatal issue of Exceedingly Large Lens Distortion "B" Values after Auto-Optimization - largest being >15K or others as Negative Values <-1. After this, PTAStitcher_x64 fails to produce Intermediate Files during Create Final, and Preview fails similarly after drawing nearly round Image Outlines.

I've tried most everything I can imagine. On a few cases, I've been able to generate "OK" results by changing the Step#2 "Lens Type" from "Rectilinear(Normal)" to "Panoramic (Cylindrical)" - but even this is mostly unsuccessful.

When the same exact images are included in Panos (single-row 2 image or larger; or multi-row) then everything is OK, and the previously "fatal" images are properly Optimized (with small "B Distortion" values) and readily presented in Preview and rendered in Create Final.

maxlyons
Posts: 3649
Joined: Fri Jun 20, 2003 8:55 pm
Location: USA
Contact:

Re: PTA 6.x problems with Widefield Lens single Exposure Sta

Post by maxlyons » Sun Oct 20, 2013 5:20 pm

seigell wrote:I'm encountering the nearly always fatal issue of Exceedingly Large Lens Distortion "B" Values after Auto-Optimization
My guess is that the stacks have one or more bad control points.

A brief explanation may shed some light on a whole host of problems. PTAssembler attempts to select control points using the configured control point picker program (e.g. PTAPicker, Autopano, etc.). If any of the control points that these programs create are bad (i.e. incorrectly placed on two features that aren't the same), then it is almost certainly the case the optimizer will give a bad result. This is because neither the control point picker program nor the optimizer "see" the images the way that you and I do. To these programs, the images are just a series of numbers. So, to you and me it may be obvious that a control is placed in the "wrong" place. The control point picker program, however, it is looking for regions with similar features (described as a series of numbers), and sometimes two different features may end up with very similar series of numbers which can confuse the control point picker into believing that it is looking at the same feature. In fact, a good deal of the complexity in most control point picker programs is logic that tries to avoid this situation and remove "false positives" like this, but this logic is almost impossible to make 100% error-proof. Anyway, if the control point picker identifies one or more bad control points, the optimizer ends up forcing the images to fit by estimating what appear to it to be reasonable values, but the result is almost certainly going to be bad.

So...in the event that the optimizer gives unreasonable results, the first thing to do is check for bad control points. You can do this by visual inspection. Often, it can be helpful to look at the control point information table (press F9) and sort by optimized distance. It is usually the case (although not necessarily) that the bad control points will have significantly larger optimized distances than the good control points, so this can speed up the hunt for bad control points.

Luckily, PTAssembler allows you to fix this type of problem; you can manually identify and remove the bad control points, and then optimizing the project manually.

If this isn't the problem, please post (or send me) a project and images, and I'll take a look.

Max

seigell
Posts: 90
Joined: Tue Mar 21, 2006 10:26 pm

Re: PTA 6.x problems with Widefield Lens single Exposure Sta

Post by seigell » Mon Oct 21, 2013 12:33 am

I've checked the F9 CP Info, and on several of these Single Image Exposure Stacks the largest Optimized CP Distance was in the range 2-3.5 yet the Lens B parameter were posting at values >1.5 (once 34.xxx) - and the resulting Preview Images showed as seriously distorted or even causing PTStitcher to silently Bomb Out.

I've posted one such Exposure Bracket set to Dropbox, involving IMG_0027 thru ING_0029, and also the Project File and the Stitching Script set. With these Images, the largest Optimized CP Distance was 0.365 yet the Lens B value was 1.64 and PTAStitcher was unable to generate the Preview images (but the Preview outline showed something that would have looked like a Stop Sign).

seigell
Posts: 90
Joined: Tue Mar 21, 2006 10:26 pm

Re: PTA 6.x problems with Widefield Lens single Exposure Sta

Post by seigell » Mon Oct 21, 2013 12:35 am

I've also noticed that when one Unchecks the Lens B Optimize checkbox on Step#4, Auto-Optimize will still modify that parameter.

Post Reply

Who is online

Users browsing this forum: No registered users and 3 guests