[PTGui] HowTo: use lens calibration database values with subsequent panos?

[l_d_allan]

I'm unclear how to make use of the lens database values stored in the
advanced Lens Settings tab.

FAQ 6.4 describes the steps involved to make a single row 360 degree
pano. My impression is that an objective of this step is to get
Optimizer results that are reported as at least "This is very good"
and hopefully results reported as "This is too good to be true". This
would typically involve:
* a carefully done pano in near ideal conditions
* calibrated panohead offset for the focal length of the lens being
used
* carefully cleaned-up CP's (shouldn't be too many to remove or add if
pano taken carefully ... ymmv)

My understanding is that the end result of this step is to have
calculated values for horizontal FOV, a, b, c, d, and e that are more
or less as good as you are going to get for that specific lens (and
specific focal length for a zoom). These values would typically be at
least as good and usually better than the FOV, a, b, c, d, and e
values that the optimizer calculates for a non-360 degree pano,
especially a pano taken in less than ideal conditions, and especially
for multi-row panos..

Near the end of FAQ 6.4, there is the statement:
"Use the Lens Database button to save the values for later use."

I'm fuzzy on just how to use these values.

For purposes of discussion, suppose I have taken a 360 degree single
row pano with a 50mm prime in portrait mode. After CP cleanup, the avg/
max Optimizer values are down to under 0.4 / 1.0. I've saved the FOV,
a, b, c, d, and e values in the Lens database for this lens.

Then later I take a multi-row pano with the same 50mm prime that
covers about 150 degrees horizontally. It consists of about 3 rows of
20 images each, for a total of about 60 images. The conditions were
less than ideal, with moderate wind, moving clouds, and some object
movement. With several iterations of careful CP cleaning and re-
optimizing, I've been able to lower the avg/max results reported by
the Optimizer from about 3.0 / 15.0 to 1.5 / 3.0.

Now what do I do?

My impression is that I want to use the Lens Database values to
substitute for the less than ideal FOV, a, b, c, d, and e values that
the Optimizer has calculated for this multi-row, non-360 degree pano.
I would use the "Load From Database" option. This should result in a
better stitch, which is the point of this effort.

Some questions:

* Do I run or not run the optimizer again? My speculation is that I
wouldn't, as re-running the Optimizer would throw away the lens
database values. Or is the Optimizer run again, but with different
options?

* If leveling was involved with use of vertical line CP's, should this
be done before the "Load From Database"? What about centering?
Straightening the horizon?

* Are there checkboxes in certain advanced tabs that would be changed
from default/previous settings to have the Lens Database values used
appropriately?

* After using "Load From Database", am I now ready to go to the
"Create Panorama" tab for cropping and stitching? Are there one or
more remaining steps?

With non ultra-wide lenses, it isn't trivial to prepare a 360 degree
pano, and I'm fuzzy on how to take advantage of it once it's been
prepared.

--
You received this message because you are subscribed to the Google Groups "PTGui" group.
To post to this group, send email to pt...@googlegroups.com
To unsubscribe from this group, send email to ptgui+un...@googlegroups.com
Please do not add attachments to your posts; instead you may upload files at
http://groups.google.com/group/ptgui/files
For more options, visit this group at http://groups.google.com/group/ptgui

[John Houghton]

On May 13, 12:41 pm, l_d_allan <lynn.d.al...@gmail.com> wrote:
> I'm unclear how to make use of the lens database values stored in
> the advanced Lens Settings tab.

You need to understand that one purpose of running the optimizer is to
evaluate the lens parameters. If you have available calibrated values
(via the lens database, a template, or simply entering the values
manaually), then the optimizer can be relieved of this task. This is
achieved by unchecking these parameters on the Optimizer tab (Advanced
mode). However, the lens shift parameters can vary from one shoot to
another, owing to mechanical tolerances in the lens and lens mount, so
these parameters can be optimized for optimum results. You set up the
lens parameters after loading in the camera images.

The other purpose of the optimizer is to evaluate the position
parameters: yaw, pitch and roll of each image. Again, if you happen
to already know what these values are (if not too accurately), you may
give this information to PTGui in various ways (via a template, or
entering the information manually with help from the Fill Yaw option).

All this will help the optimizer to home in very quickly and easily on
a good alignment - given a reasonable assignment of control points.
Only if you are using a high precision panorama head would you be able
to dispense with control points and running the optimizer.

It may be that PTGui will produce a perfectly good stitch without
doing any of these things. OTOH, it might not. There may be orphan
images without control points, for example, that are hopelessly out of
position. Or the spread of control points may be inadequate to enable
accurate lens parameters to be evaluated, leading to stitching errors
despite a "good" optimization report from the optimizer. Generally
speaking, you need far fewer control points when you are using
calibrated lens parameters, and their spread is also much less
critical.

You therefore need to take all the above into consideration when
planning how you will stitch a given set of images.

John

[l_d_allan]

Hi John,

Thanks for the feedback. My understanding is that your answer is more
or less equivalent to "it depends on what you want to do."

How about in this situation?

* I've carefully taken a 360 degree pano with a 50mm prime, calibrated
panohead, in best case circumstances (outside, no wind, no clouds, no
close objects so miminal parallax issues, lots of detail from top to
bottom at 30+ meters away, plenty of sun so fast shutter speed and
plenty of DOF, etc.). With minimal CP cleanup, the avg/max CP
distances are under 0.5 / 1.0 My impression is that the values I would
put in the lens database from this pano for Horizontal FOV, a, b, and
c would be pretty much as good as I'm going to be able to get.

* I also believe you advised not to put d and e values (for sensor
shift) in the Lens Database because they could change.

* For purposes of discussion, suppose that I later take a pano with
the 50mm prime in much less than best case circumstances (hand held,
ad hoc so somewhat in a hurry, about 90 degrees instead of 360
degrees, some movement from wind, etc.). The avg/max CP distances
might be 5.0 / 30.0 (or worse).

* I unclear how to make use of the values from the Lens Database from
the earlier "best case" pano.

* This is my speculation on how to proceed. Correction appreciated.

- Load the images for this "much less than best case" pano and
align.

- Use the CP Assistant and CP Table to get an idea of the overall
pano quality. Re-optimize (F5) to see the avg/max CP distances. Maybe
save .pts intermediate file.

- See what is reported by PTGui's "Delete worst CP's". Accept or
decline as appropriate (probably accept)

- Add CP's as necessary to make CP Assistance happier.

- Re-run optimizer and save project .pts file (with perhaps
intermediate name). Avg/max CP distances may or may not improve.

- Maybe use APClean by opening .pts file just saved

- Re-load from .pts file just written by APClean.

- Re-run optimizer (F5). Avg/max CP distances should be better ...
hopefully significantly better.

- Use Control Points tab to enter vertical CP's for leveling, as
appropriate

- Use Panorama Editor to level, center, and otherwise adjust pano

- re-run optimizer (f5). Maybe save .pts intermediate file.

- This is the point as which I'm unclear how to proceed. Go to Lens
Settings tab, and click on "Lens database ..." button.

- Select entry from 50mm prime lens, and click on "Load from
database" button

- Go to Optimizer tab and click on "Advanced >>" button

- Uncheck boxes for "Field of View", a (lens distortion), b (lens
distortion), c (lens distortion)

- re-run optimizer (?). Avg/max CP distances should be same? better?
than previous?

- Examine pano in Panorama Editor to see if it looks reasonable.
Crop as appropriate.

- Go to "Create Panorama" tab. Use "Set optimum size ..." button.
Set other options for file format, stitch using, blend using, etc.
Save .pts file using perferred file naming convention.

- Click on "Save and send to Batch Stitcher".

- Look at rendered pano and check for stitching errors.


My speculation is that some or much of the above is flawed (wrong,
unnecessary, wrong sequence, etc.). Revisions appreciated.

[Matthew Rogers]

Omg, you're making things so much more difficult then need be.

1. If you have well calibrated values then ALWAYS USE THEM REGARDLESS OF THE SCENARIO.

2. DELETE APCLEAN, YOU DO NOT NEED IT !

3. Unless you're using a different camera body the D and E will NEVER change.

4. Just save a template with your calibrated values. Simply import the images, apply the template and optimise. As you're using the calibrated a,b,c,d,e and fov you must uncheck them all and simply calculate the control point distances.

That's it, it's no more difficult then that. You seem to take 10x longer to stitch images then I've ever seen. Maybe you want to make it difficult so it seems like more of any achievement when you final stitch an image ?

Matt



Sent from my iPad

[Joergen Geerds]

On May 15, 8:07 am, Matthew Rogers <matt...@360precision.com> wrote:
> Omg, you're making things so much more difficult then need be.

I agree, it is easier in reality than posted.

> 3. Unless you're using a different camera body the D and E will NEVER change.

this statement is unfortunately wrong.
D/E values constantly change, they can even change during a shoot, if
the PG isn't careful.
The main cause for shift is the lens mount... a good lens might have
only 0.1-0.25mm play in the mounting flange, but that means already a
possible shift of 5-40px, in addition to the off-center sensor/lens
shift.

> 4. Just save a template with your calibrated values. Simply import the images, apply the template and optimise. As you're using the calibrated a,b,c,d,e and fov you must uncheck them all and simply calculate the control point distances.

sorry to be a stickler, but it's import images, apply template,
generate CPs, optimize everything but ABC and fov (include fov if the
user is using a zoom, since it is impossible to hit the same zoom
position). otherwise you are right, it's that easy.

joergen

[Hans]

On May 15, 2:07 pm, Matthew Rogers <matt...@360precision.com> wrote:
> Omg, you're making things so much more difficult then need be.
>
> 1. If you have well calibrated values then ALWAYS USE THEM REGARDLESS OF THE SCENARIO.
>
> 2. DELETE APCLEAN, YOU DO NOT NEED IT !
>
> 3. Unless you're using a different camera body the D and E will NEVER change.

Unfortunately this is not correct.

Shift changes with lens, and this is actually often a larger change
than sensorshift.

I just done several teststitches with a Canon 15mm and with the Tokina
at 15mm.
Exactly same setup.
Canon 15mm gives me Horisontal 0 and -38-43 pixels vertical. The
Tokina -11 to 24 horisontal and -24 vertical.

You will also experience that the shift may change when you remount
the lens.
Even my Canon 15mm has a small play which changes the centerpoint.

Hans

>
> 4. Just save a template with your calibrated values. Simply import the images, apply the template and optimise. As you're using the calibrated a,b,c,d,e and fov you must uncheck them all and simply calculate the control point distances.
>
> That's it, it's no more difficult then that. You seem to take 10x longer to stitch images then I've ever seen. Maybe you want to make it difficult so it seems like more of any achievement when you final stitch an image ?
>
> Matt
>
> Sent from my iPad
>

> > For more options, visit this group athttp://groups.google.com/group/ptgui

>
> --
> You received this message because you are subscribed to the Google Groups "PTGui" group.
> To post to this group, send email to pt...@googlegroups.com
> To unsubscribe from this group, send email to ptgui+un...@googlegroups.com

> Please do not add attachments to your posts; instead you may upload files athttp://groups.google.com/group/ptgui/files
> For more options, visit this group athttp://groups.google.com/group/ptgui

[l_d_allan]

On May 15, 6:07 am, Matthew Rogers <matt...@360precision.com> wrote:
> Omg, you're making things so much more difficult then need be.

Oh, maybe that's why I posted the question. I didn't describe it as
the correct way to use, but rather "hey, this doesn't seem right.
Could someone with more experience look this over and 'correct the
error of my ways' ".

> 2. DELETE APCLEAN, YOU DO NOT NEED IT !

Maybe, or maybe "YOU", Matthew Rogers, don't need it.

My speculation is that our use, expectations, and experience with
PTGui are different. I'm curious what the average number of images in
your panos tends to be? 2? 4? 6? Or 50+.

Do you have a full frame, or crop camera? What is the focal length you
tend to use?

My widest lens is an 18mm from the EF-S 18-55 kit lens on a Canon 50d
"crop camera". That makes it equivalent to a ~28mm, which is hardly
wide angle. I'd like to have an EF-S 10-22, but my budget ain't what
it used to be. A full frame 5dm2 with EF 16-35L ain't gonna happen
anytime soon. In the meantime, I try to do the best I can.

Many of my panos here on the Front Range of Colorado Springs are 3+
rows, and 180 degrees and wider. 30 images would be a rather small
pano. 50 to 100+ images aren't unusual. I submit that PTGui is a very
different experience when your average pano size is 10 images or
under, compared to 50 and over, and 100 to 200+ image pairs.

> 3. Unless you're using a different camera body the D and E will NEVER change.
>
> 4. Just save a template with your calibrated values. Simply import the images, apply the template and optimise. As you're using the calibrated a,b,c,d,e and fov you must uncheck them all and simply calculate the control point distances.

So you don't Align images? THAT certainly simplifies things, and will
speed up 100+ image panos immensely. OMG, I've been Aligning images
unnecessarily?

As H. L. Mencken wrote, "For every complex problem, there is a
solution that is simple, neat, and wrong."

> That's it, it's no more difficult then that. You seem to take 10x longer to stitch images then I've ever seen. Maybe you want to make it difficult so it seems like more of any achievement when you final stitch an image ?

Maybe, or maybe I have doubts about my current technique, and was
hoping to get constructive feedback to get simpler, more efficient,
and improved results in less time. But thanks for the pyschological
speculation.

[l_d_allan]

On May 15, 8:01 am, Joergen Geerds <jgee...@gmail.com> wrote:
> On May 15, 8:07 am, Matthew Rogers <matt...@360precision.com> wrote:>
> Omg, you're making things so much more difficult then need be.
>
> I agree, it is easier in reality than posted.

Thanks for your patient, constructive reply to this confused semi-
newbie. IMHO, the documentation is adequate for creating the values
for the Lens Database from a 360 degree pano, but rather sparse for
actually using the values. From FAQ 6.4,
"Use the Lens Database button to save the values 'for later use'. "

I submit that "for later use" is rather vague, but maybe it's just me.
<g>

Hence the request for clarification, as applicable to a more specific
rather than abstract situation.

> > 3. Unless you're using a different camera body the D and E will NEVER change.
>
> this statement is unfortunately wrong.
> D/E values constantly change, they can even change during a shoot, if
> the PG isn't careful.

"PG" ??? I've known people who get PG who aren't careful, but I don't
think this is what you are saying. <g>

> The main cause for shift is the lens mount... a good lens might have
> only 0.1-0.25mm play in the mounting flange, but that means already a
> possible shift of 5-40px, in addition to the off-center sensor/lens
> shift.

Of my four lenses, 3 have plastic lens mounts. I suspect the
tolerances are not so great.

[l_d_allan]

On May 15, 8:55 am, Hans <hans...@gmail.com> wrote:
> On May 15, 2:07 pm, Matthew Rogers <matt...@360precision.com> wrote:

> > 3. Unless you're using a different camera body the D and E will NEVER change.
>
> Unfortunately this is not correct.
>
> Shift changes with lens, and this is actually often a larger change
> than sensorshift.

And also, thanks for the constructive, patient feedback to a confused,
ignorant, semi-newbie such as myself, and perhaps other participants
on this forum.

I didn't know what to expect from the D and E values, but they seemed
very small, practially miniscule if I understood the values correctly.
For my EF 50mm f1.8 prime (a.k.a. 'nifty fifty / plastic
fantastic' ... one of the worst built lenses ever sold, but seems to
be very sharp with excellent IQ) ... D = -0.00833 pixels and E =
-0.0007018.

I'm trying to comprehend what ~1/100th and ~1/1000 of a pixel amount
to. I suppose that is smaller than the "itsy bitsy, teenie weenie
yellow polka dot bikini" from the 1960's song, which I recall
listening to out at Lake Maurer. <g>

> I just done several teststitches with a Canon 15mm and with the Tokina
> at 15mm.
> Exactly same setup.
> Canon 15mm gives me Horisontal 0 and -38-43 pixels vertical. The
> Tokina -11 to 24 horisontal and -24 vertical.

Thanks for taking the time to do the test, and post the results. I'm
curious where those numbers come from. I don't see anything resembling
them on the Lens Settings tab, or the Image Parameters. Are these CP
distances?

I'm also curious if you are using a full frame, or crop camera. A 15mm
on a 1.6x crop camera compared to a 15mm on a full-frame would be a
MUCH different lens.

[Matthew Rogers]

On 16 May 2010, at 10:13, l_d_allan wrote:

> On May 15, 6:07 am, Matthew Rogers <matt...@360precision.com> wrote:
>> Omg, you're making things so much more difficult then need be.
>
> Oh, maybe that's why I posted the question. I didn't describe it as
> the correct way to use, but rather "hey, this doesn't seem right.
> Could someone with more experience look this over and 'correct the
> error of my ways' ".
>
>> 2. DELETE APCLEAN, YOU DO NOT NEED IT !
>
> Maybe, or maybe "YOU", Matthew Rogers, don't need it.
>
> My speculation is that our use, expectations, and experience with
> PTGui are different. I'm curious what the average number of images in
> your panos tends to be? 2? 4? 6? Or 50+.

I shoot panoramas ranging from 3 images with a D700/10.5 right up to 200+ images with the D700/300mm f4.

Let me repeat myself, you do not need APCLEAN, you're simply wasting your time using it and adding unnecessary steps into the production process.

>
> Do you have a full frame, or crop camera? What is the focal length you
> tend to use?

I don't need to get into my list of equipment but I have enough to know what I'm talking about.

>
> My widest lens is an 18mm from the EF-S 18-55 kit lens on a Canon 50d
> "crop camera". That makes it equivalent to a ~28mm, which is hardly
> wide angle. I'd like to have an EF-S 10-22, but my budget ain't what
> it used to be. A full frame 5dm2 with EF 16-35L ain't gonna happen
> anytime soon. In the meantime, I try to do the best I can.
>
> Many of my panos here on the Front Range of Colorado Springs are 3+
> rows, and 180 degrees and wider. 30 images would be a rather small
> pano. 50 to 100+ images aren't unusual. I submit that PTGui is a very
> different experience when your average pano size is 10 images or
> under, compared to 50 and over, and 100 to 200+ image pairs.

I don't see how it's any different whether using 3 images or 300, the basic principles are the same as are the optimisation and CP deleting processes.

>
>> 3. Unless you're using a different camera body the D and E will NEVER change.
>>
>> 4. Just save a template with your calibrated values. Simply import the images, apply the template and optimise. As you're using the calibrated a,b,c,d,e and fov you must uncheck them all and simply calculate the control point distances.
>
> So you don't Align images? THAT certainly simplifies things, and will
> speed up 100+ image panos immensely. OMG, I've been Aligning images
> unnecessarily?

Well of course you need to Align the images, some steps are a given, just like I didn’t say you need to use your mouse button to click the PTGui icon to open it.

>
> As H. L. Mencken wrote, "For every complex problem, there is a
> solution that is simple, neat, and wrong."
>
>> That's it, it's no more difficult then that. You seem to take 10x longer to stitch images then I've ever seen. Maybe you want to make it difficult so it seems like more of any achievement when you final stitch an image ?
>
> Maybe, or maybe I have doubts about my current technique, and was
> hoping to get constructive feedback to get simpler, more efficient,
> and improved results in less time. But thanks for the pyschological
> speculation.
>

And I gave you a method to simplify things but you clearly want to take the long road.

[Matthew Rogers]

Yes Hans, but what does the information below have to do with the original question ? Nothing !

Lisa was basically asking at which step in the process do you apply CALIBRATED parameters to a new project. What the values of other lenses are is completely irrelevant and your post will only server to confuse her more.

I'm not sure why you'd want to interchange calibrated values between lenses ?

Might help to actually read the post properly and reply in context without confusing matters more.

Matt

[Bjørn K Nilssen]

On 16 May 2010 at 2:48, l_d_allan wrote:

> > Shift changes with lens, and this is actually often a larger change
> > than sensorshift.
>
> And also, thanks for the constructive, patient feedback to a confused,
> ignorant, semi-newbie such as myself, and perhaps other participants
> on this forum.

I have a feeling that you're using PTgui in simple mode?
Don't do that!
Wherever you see an "Advanced" button - click it to exit Simple mode.
Ie first in the Project assistant tab.
But for some odd reason you'll also have to go to the Optimizer tab to leave "Simple"
mode completely. You want full control to understand what's happening

Until you get familiar with panos and PTgui, don't waste your time with hundreds of
shots. Make it simple and use something like 2 rows and 5 columns, or even only 1 row of
3 images. The principle is the same.

Steps to make a pano:
Go to the Project assistant
Open a window (Explorer) with your images, select them and drag/drop them into PTgui.
Make sure PTgui says Lens type = Rectilinear (unless you use a fisheye)
Do *not* click that Align button.
Instead, from the menu choose Control points/Generate control points.
After a while PTgui has made all the CPs it could find.
You could check the Control Point table if there are any images that didn't get CPs.
If you got no CPs then you need to add some manually in the CP tab.
Next is to go to the optimizer tab.
On the left side check 'Field of View' and 'b' and leave Yaw/pitch/roll checked for all
but the first image.
Click the 'Run optimizer' button at the bottom. (F5)
If you get good results check the 'a' on the left side and run again.
Next is to check the 'd' and 'e' and run again.
Usually you don't need to optimize/use 'c'.
Use Panorama Editor to view your pano.
Use 'Control Point table' to find any high values (sort on column header by clicking it),
and delete the CPs (as APclean does?)
After optimizing again you should be ready?
If it is not level you could use the Editor to drag it up/down with left or right MB.
Sometimes it works to click the "Straighten panorama" button, and sometimes not.
And the 'Center' and 'Fit' panorama buttons.

There are more tools and options, but this will get you started.
Forget about lens database etc for now. You don't need it

--
Bjørn K Nilssen - http://bknilssen.no - panoramas and 3D

[John Houghton]

On May 16, 10:13 am, l_d_allan <lynn.d.al...@gmail.com> wrote:
> So you don't Align images? THAT certainly simplifies things, and will
> speed up 100+ image panos immensely. OMG, I've been Aligning images
> unnecessarily?

Never say never. If you look at PTGui Pro's Project Settings tab, you
will see that you can customize the Align Images function, so that
exactly what it does is up to you. When you have more experience, you
can can take advantage of this, but for now - be aware of the settings
when using the batch stitcher.

John

[l_d_allan]

On May 16, 5:07 am, Matthew Rogers <matt...@360precision.com> wrote:
> On 16 May 2010, at 10:13, l_d_allan wrote:
> Let me repeat myself, you do not need APCLEAN, you're simply wasting your time using it and adding unnecessary steps into the production process.

I suppose it is a true statement that I don't "need" APClean. However,
based on my so far limited use of it for my circumstances, my
preliminary assessment is that I'm finding it useful, and am
interested to learn more about it. It may turn out I arrive at the
same opinion as yourself, but that isn't the case so far.

> > Do you have a full frame, or crop camera? What is the focal length you
> > tend to use?
>
> I don't need to get into my list of equipment but I have enough to know what I'm talking about.

It is perhaps a relevant question ... if you have a full-frame with
ultra-wide angle lens and your pictures are not that FOV wide, you are
almost certainly going to take a LOT fewer pictures in a pano. On the
other hand, with the equivalent of a ~28mm non-UWA lens, my panos tend
to have 30+ images in them. My less-than-fully-informed opinion of
APClean is that it can be really handy with lots of images in a pano.
With fewer panos, you don't "need" it.

Another factor is that not all of my panos are taking in best case
situations. Some are taken hand-held, or on a tripod and I don't have
my panohead with me. Or I'm rushed, and my technique is not so great.
Or there's wind and moving clouds. This morning I was taking 4+ image
panos of a praise band on a stage ... moving people, mediocre
lighting, shallow depth of field, etc. Those panos are likely to take
some work, and so far, APClean seems like a useful tool to have in my
quiver.

My impression is that such panos are likely to start off with mediocre
or worse avg/max CP distances. So far, APClean seems useful in these
scenarios. With a carefully taken pano in a best case situtation, I
can certainly believe APClean would be less useful.

> I don't see how it's any different whether using 3 images or 300, the basic principles are the same as are the optimisation and CP deleting processes.

Perhaps we are on "different wavelengths". Handling a pano of 3 images
is so much of a 'quantitative difference' in scale that it becomes a
'qualitative difference'.

With three images in a pano (2 image-pairs), imho, you can afford the
time to look at each CP in each image-pair, and carefully decide
which, if any to delete. You can take into account factors such as
favoring/keeping CP's that are distant from each other. You might
remove CP's with low distance numbers that are right next to another
excellent CP, because it is more or less redundant (but I don't
suppose you would, but you could consider it.)

With 300 images .... perhaps 1000 image-pairs ... you could
conceivably have 10,000 to 20,000 CP's. Maybe I'm the only one this
applies to, but I don't feel like looking at each CP with the same
care as when there are 20 to 30 CP's in two image- pairs. To me, that
is a scenario where this PTGui semi-newbie and APClean newbie see
potential value in a utility like APClean. YMMV.

> And I gave you a method to simplify things but you clearly want to take the long road.

Not really. I'm trying to learn the error(s) of my ways from members
of this forum with constructive feedback, who seem to know what they
are talking about.

As noted, the OP was HowTo? to ask how, rather than HowTo! to inform
people what I'm doing, and encouraging them to do likewise.

Reject filter, meet Matthew.
Matthew, meet reject filter.

[Bjørn K Nilssen]

On 16 May 2010 at 12:59, l_d_allan wrote:

> Perhaps we are on "different wavelengths". Handling a pano of 3 images
> is so much of a 'quantitative difference' in scale that it becomes a
> 'qualitative difference'.
>
> With three images in a pano (2 image-pairs), imho, you can afford the
> time to look at each CP in each image-pair, and carefully decide
> which, if any to delete. You can take into account factors such as
> favoring/keeping CP's that are distant from each other. You might
> remove CP's with low distance numbers that are right next to another
> excellent CP, because it is more or less redundant (but I don't
> suppose you would, but you could consider it.)
>
> With 300 images .... perhaps 1000 image-pairs ... you could
> conceivably have 10,000 to 20,000 CP's. Maybe I'm the only one this
> applies to, but I don't feel like looking at each CP with the same
> care as when there are 20 to 30 CP's in two image- pairs. To me, that
> is a scenario where this PTGui semi-newbie and APClean newbie see
> potential value in a utility like APClean. YMMV.

So could you please tell us what you think APclean can do for you that you can't get from
PTgui then?
PTgui lets you compare/list CPs on a global or pair-wise basis, and it even lets you
remove all the worst CP pairs automatically. I have never used APclean (never needed it),
but after reading the specs for it I can't see that it can do anythng that PTgui can't do
faster and easier.

> > And I gave you a method to simplify things but you clearly want to take the long
> road.
>
> Not really. I'm trying to learn the error(s) of my ways from members
> of this forum with constructive feedback, who seem to know what they
> are talking about.

Then why don't you listen to all the good advice you get?
When you're learning to sail a boat you don't start off with crossing the Pacific alone,
and when learning to run you don't start off with a marathon. Neither do you start
learning to drive by driving solo across the US.
So why do you think it is such a good idea to start learning making panos with 300 images
or more - and to not listen to advice from those that have done this for a bit longer
than you have?
By all means, keep on using APclean and make Gigapixel panos to learn stitching if you
want to, but don't expect to get much advice if you never listen to it!

--
Bjørn K Nilssen - http://bknilssen.no - panoramas and 3D

[Keith Martin]

Sometimes people ask questions regarding extreme cases as a way of
learning where 'extreme' actually is. I agree that it would make much
more sense to work your way up to this, but nobody needs to get
stressed about it.

k

[Bruce Hemming]

On May 16 12:59PM -0700 l_d_allan wrote

> I don't need to get into my list of equipment but I have enough to know what I'm talking about.

 
>It is perhaps a relevant question ... 

a quick google of Mathew Rogers - or 360 precision - would have told you that Matt is familiar with most equipment combinations.  He might be a little "direct" ( I believe that the correct Antipodean term is "toey" - is that right Matt?) in his answers but is always worth reading.

 
>Another factor is that not all of my panos are taking in best case
situations. Some are taken hand-held, or on a tripod and I don't have
my panohead with me. Or I'm rushed, and my technique is not so great.
Or there's wind and moving clouds. 

I've a feeling that right there is the answer to most of your problems.  If your shooting technique has flaws then you will always be struggling to correct the panos you make, regardless of how well versed you are in the minutiae of PTGUI.  My advice would be to get a fish eye lens and start working with that - there is a very good reason why just about every pano maker whose work is worth viewing uses FE lens in the majority of circumstances, it is a recognized and efficient workflow.  

Making a panorama is actually a very straightforward and simple process, there isn't any "secret" knowledge that guarantees instant success and unfortunately you do seem to take a perverse delight in making things difficult for yourself - partly by selectively listening to advice.  I'm not being hostile, everyone knows that you have to start somewhere and we all make mistakes ( God knows I did when I began) if you re read answers to your posts I think that you will find the answers that you are looking for.

Bruce Hemming

[Matthew Rogers]

Hi Bruce,

Thanks for revealing my secret identity :) I see people all the time in the same situation as Lisa and as soon as they realise that it's actually a very simple process they just get it. I mean, some people really do over analyse the entire calibration/control point issue almost to death.

I guess I'm direct because I answer the same questions all the time and often reply to 100+ emails every day, 7 days a week. Unfortunately I simply don't have the time to sugar coat things to appease the sensibilities of the few.

I'll try and get some workflow documents that I've been working on that should explain the issues in a more user friendly way.

Matt

[l_d_allan]

On May 16, 7:00 am, John Houghton <j.hough...@ntlworld.com> wrote:
> On May 16, 10:13 am, l_d_allan <lynn.d.al...@gmail.com> wrote:
>
> If you look at PTGui Pro's Project Settings tab, you
> will see that you can customize the Align Images function, so that
> exactly what it does is up to you.  When you have more experience, you
> can can take advantage of this, but for now - be aware of the settings
> when using the batch stitcher.

Thanks. I hadn't looked at this Tab in a while, since I first started
using PTGui a few months ago. Some of the options make a bit more
sense to me now. My normal practice is to use the Batch-Stitcher with
the "Create the panorama" button unchecked. I've ignored most of the
other settings, as I didn't really understand several of them.

My impression is that if you have a fully automated panohead robot
that can "communicate" with PTGui, you might not need to Align the
images, but otherwise I can't think of a scenario where you wouldn't
Align images. But my use of PTGui has been limited, so I conceed
ignorance.

[l_d_allan]

On May 16, 5:43 pm, Bjørn K Nilssen <b...@bknilssen.no> wrote:
> On 16 May 2010 at 12:59, l_d_allan wrote:

> So could you please tell us what you think APclean can do for you that you can't get from
> PTgui then?

> PTgui lets you compare/list CPs on a global or pair-wise basis, and it even lets you
> remove all the worst CP pairs automatically. I have never used APclean (never needed it),
> but after reading the specs for it I can't see that it can do anythng that PTgui can't do
> faster and easier.

I really regret mentioning APClean. In retrospect, it is not all that
relevant to the question of how to use the lens calibration database.
This thread might have had more to do with some clarifications about
the lens database if I'd just lumped APClean in with "clean up CP's".
The APClean "tempest in a teacup" has been an unfortunate distraction,
imho.

But a separate thread on APClean might be very valuable. Erik K.
originally highly recommended APClean in an unrelated thread earlier
this month.

> > > And I gave you a method to simplify things but you clearly want to take the long
> > road.
>
> > Not really. I'm trying to learn the error(s) of my ways from members
> > of this forum with constructive feedback, who seem to know what they
> > are talking about.
>
> Then why don't you listen to all the good advice you get?

About the only advice offered in that post was:

> 2. DELETE APCLEAN, YOU DO NOT NEED IT !

And some possibly wrong info about D and E. I didn't find it to be a
particularly constructive reply, and in retrospect should have ignored
it.

I'm not aware of any other advice I've rejected regarding use of the
lens calibration database, since the discussion since then has been
almost entirely related to APClean, unfortunately.

> When you're learning to sail a boat you don't start off with crossing the Pacific alone,
> and when learning to run you don't start off with a marathon. Neither do you start
> learning to drive by driving solo across the US.
> So why do you think it is such a good idea to start learning making panos with 300 images
> or more - and to not listen to advice from those that have done this for a bit longer
> than you have?  

The mention of 300 images came up due to:
> I (Matthew R) don't see how it's any different whether using 3 images or 300, the basic principles are the same as are the optimisation and CP deleting processes.

To this PTGui semi-newbie, the optimization may be the same, or at
least very similar regardless of the number of images involved (other
than how much time they take). However, it seems there can be huge
differences in the CP deleting process if you are dealing with 3
images, or 300 images. Again, in retrospect, I should have ignored
this reply because it had little or nothing to do with use of the lens
calibration database, or perhaps continued it in another thread
regarding APClean, or privately with Mr. Rogers.

I do recall previous threads involving whether a panohead was
essential, where I was admittedly slow to take advice, and perhaps
perceived as less than coachable. I've since gotten a spherical
panohead, and made several homemade do-it-yourself spherical panoheads.

[l_d_allan]

On May 17, 2:03 am, Keith Martin <ke...@vortex.co.uk> wrote:
> Sometimes people ask questions regarding extreme cases as a way of
> learning where 'extreme' actually is. I agree that it would make much
> more sense to work your way up to this, but nobody needs to get
> stressed about it.

The original question was how to use the lens calibration database. To
me, the documentation such as FAQ 6.4 is essentially non-existent
about "using" the lens database. The APClean discussion has been an
unfortunate distraction.

FWIW, this is why I asked about the lens database. I think it may be
useful for some or many of my panos to "mix and match resolutions".
For example, I'd have 4 to 10 images from my EF-S 18-55 at 18mm with a
single row, and perhaps 30 to 100+ images from my EF 50 1.8 prime
(nifty fifty / plastic fantastic) with multiple rows.

The 18mm would supply the pixels for the upper sky/clouds, and the
sharper prime lens would supply the skyline and city content. This
"mixed resolution" is possible with Photoshop-CS4, and easy with
another panorama software package that I have the trial for, but
considered an advanced use of PTGui.

Here's an example to illustrate:
http://berean.myphotoalbum.com/view_album.php?set_albumName=album414

My speculation is that values from the lens database for the 18mm and
50mm would be helpful to merge such mixed resolution panos together,
but I could very well be mistaken.

I would agree this is going to be a tough pano for someone such as
myself to get to work. I'm not all that experienced or proficient with
PTGui, or even the proper techniques to take the panos in the first.
It may be premature to attempt it based on my current lack of
experience. But by tackling it, I've already learned a lot on multiple
issues, and hope to learn all that more.

[l_d_allan]

On May 17, 4:34 am, Bruce Hemming <br...@bhphoto.biz> wrote:
> On May 16 12:59PM -0700 l_d_allan wrote
>
>  > I don't need to get into my list of equipment but I have enough to  
> know what I'm talking about.
>
>  >It is perhaps a relevant question ...
>
> a quick google of Mathew Rogers - or 360 precision - would have told  
> you that Matt is familiar with most equipment combinations.  He might  
> be a little "direct" ( I believe that the correct Antipodean term is  
> "toey" - is that right Matt?) in his answers but is always worth  
> reading.

Interesting. That clears up some things. I surmise he is one of, or
the, principal with 360 Precision?

I suspect Mr. Rogers and myself might have several issues to be out-of-
sync about. He is part of a vendor of top-of-the-line panoheads, with
prices to match. I'm in the "most bang for the buck" camp, very frugal
with a very limited budget, and enjoy "Pareto preincipal 80/20
solutions".

I'm more and more using a variant of the home-built spherical "Nodal
Samarui", based on measurements and experience gained from a
discounted Panosaurus spherical panohead. I have a moderate-weight
tripod for which I paid $40. My only sharp lens is a 50mm 1.8 prime,
which was about $100 (plus kit lenses for the Canon 50d).

>  >Another factor is that not all of my panos are taking in best case
> situations. Some are taken hand-held, or on a tripod and I don't have
> my panohead with me. Or I'm rushed, and my technique is not so great.
> Or there's wind and moving clouds.
>
> I've a feeling that right there is the answer to most of your  
> problems.  If your shooting technique has flaws then you will always  
> be struggling to correct the panos you make, regardless of how well  
> versed you are in the minutiae of PTGUI.  

Again, the original question was about *using" the lens calibration
database. My speculation is that it would be helpful to improve panos
taken in marginal conditions. I could be wrong. It seems to me that
the FOV, a, b, and c values would be suspect with marginal panos, and
by applying lens calibration database values that were more accurate
for that specific lens and specific focal lenth, those panos would
tend to improve.

Sometimes I have the time to carefully take the pano, and have my
tripod and spherical panohead with me. I know my techique is far from
perfect, but I think it is getting to be decent, and improving based
on feedback from this forum, and learning from my many mistakes.

But there are plenty of other times when this isn't the case.
Sometimes I just have a 3+ year old 7mpx point-n-shoot with me. Or I
have my DSLR and maybe my cheap monopod, but not my tripod and/or
spherical panohead? Or I only have a very limited amount of time?

Do I not take panos then?

I'm certainly not a professional photographer, but a comment from a
professional on a completely unrelated topic was "delivering the goods
in less than good situtations". I would like to be able to
increasingly get good results from not-so-good shooting conditions.

I don't want to be sloppy when I have the time, but it seems like
lessons learned from challenging panos will contribute to better
results from carefully taken panos, especially in saving time, and
perhaps avoiding expenses.

> My advice would be to get a  
> fish eye lens and start working with that - there is a very good  
> reason why just about every pano maker whose work is worth viewing  
> uses FE lens in the majority of circumstances, it is a recognized and  
> efficient workflow.

I realize this isn't what you ar saying, but sometime on this forum, I
get the impression that unless you can afford a $350+ spherical
panohead, FF camera, one or several ulta-wide angle lenses, and one or
several fish eye lenses, you shouldn't be trying anything beyond the
basics of taking panos.

> Making a panorama is actually a very straightforward and simple  
> process, there isn't any "secret" knowledge that guarantees instant  
> success and unfortunately you do seem to take a perverse delight in  
> making things difficult for yourself - partly by selectively listening  
> to advice.  I'm not being hostile, everyone knows that you have to  
> start somewhere and we all make mistakes ( God knows I did when I  
> began) if you re read answers to your posts I think that you will find  
> the answers that you are looking for.

Personally, I find it very satisfying to get what are to me
increasingly acceptable panos on a very limited budget, rather than
solving the many technical challenges involved with pricey budget
busters. Agreed, my panos are never going to match results from a FF
5dm2 with EF 16-35L, and/or EF 15mm fish eye, and $350 to $1000+
spherical panohead. So be it.

[PTGui Support]

No personal fights on this forum please, let's finish this thread here.

Joost

On 17-5-2010 16:24, l_d_allan wrote:
> Interesting. That clears up some things. I surmise he is one of, or
> the, principal with 360 Precision?
>
> I suspect Mr. Rogers and myself might have several issues to be out-of-
> sync about. He is part of a vendor of top-of-the-line panoheads, with