Stitch large panorama in PTGUI
Alfred
Hi, there,
First of all, happy holidays to every one! I wish all of you the best in the coming 2011.
I recently released a large panorama on Gigapan at http://www.gigapan.org/gigapans/66626/ . This panorama received many serious debate about the true pixel count. Since I used Autopano, my argument was, when there is no standard way to calculate the total pixel count, we should not manipulate the software output, rather, everyone use the same baseline. Then everyone should also provide theoretical estimation of the optical pixel count. In this case, 112G pixels.
The questions here are:
1. Is there a way to stitch super size panorama in PTGUI? Can we emit raw files so that it is not subjected to image format limit?
2. Is there a way to import Autopano config file and process in PTGUI?
3. Can PTGUI accurately report the optical pixel count for super size panorama?
Thanks!
Cheers,
Alfred
Sacha
awhile. I think perhaps it should be based on total images subtracting
overlap and using the mtf of the lens. Final pixel size is irrelevant
for obvious or numerous reasons.
Unsure about generating super size 360s buy I think a giga is wildly
used for a reason. a good area for joost to adress.
On Dec 23, 5:44 pm, "Alfred" <rongkai.z...@gmail.com> wrote:
> Hi, there,
>
> First of all, happy holidays to every one! I wish all of you the best in the
> coming 2011.
>
Tom Sharpless
angular size of the pano and the angular size of one pixel, without
any guesswork about overlaps or projections. Here is an example.
Your pano covers 175 x 65 degrees and was taken with an 800mm lens on
an EOS 7D. Canon's specs give the width of the sensor on the 7D as
22.3mm, and the horizontal pixel count as 5184, so the width of one
pixel is 22.3 / 5184 mm. With your 800mm lens, that covers an angle
of arcsin ( 800 / pixel width ) which comes out to 0.000311 degrees.
There are 562,232 pixels in 175 degrees, and 208829 pixels in 65
degrees; product 117,410,395,104 pixels - 117.4 Gpixels.
Regards, Tom
Sacha
to prime.
Though zoom to extreme focal length muddies things again. Lens edges
etc..
I think we're finally on the right track here.
Kudos Tom.
Alfred
the panorama coverage during the assignment. Therefore, the stitching
software should accurately sum up the valid pixels and report the result.
But for some reason, Autopano does not report accurately, that's why I am
falling back to PTGUI for help.
In case you are not familiar with the context we are discussing here, please
refer to http://www.gigapan.org/gigapans/66626/
In fact, my estimation was also along the line Tom described.
Cheers,
Alfred
Regards, Tom
--
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
Sacha
Sorry alfred.
On Dec 23, 9:37 pm, "Alfred" <rongkai.z...@gmail.com> wrote:
> The issue was, many photographers do not have accurate instrument to measure
> the panorama coverage during the assignment. Therefore, the stitching
> software should accurately sum up the valid pixels and report the result.
> But for some reason, Autopano does not report accurately, that's why I am
> falling back to PTGUI for help.
>
> In case you are not familiar with the context we are discussing here, please
Tom Sharpless
On Dec 23, 9:32 pm, Sacha <sachagriffin...@gmail.com> wrote:
> Excellent input, but how to compare l lenses to cheap plastic, or zoom
> to prime.
> Though zoom to extreme focal length muddies things again. Lens edges
> etc..
> I think we're finally on the right track here.
> Kudos Tom.
camera pixel or better. If lens quality is the issue, then one must
do the calculation with a larger 'resolution element width'; but the
principle remains the same.
Unfortunately there is no good way to measure the practical size of
the resolution element for camera lenses. The simple formula for the
'Airy disk' (diffraction pattern of the aperture) used in astronomy
badly overestimates the resolution of camera lenses in the larger
aperture ranges we mostly use. Scientific camera review labs often
publish "mtf" or "ctf" curves that show how the contrast of a pattern
of equal black and white bars falls off as the bar width gets
smaller. That provides an objective basis for comparing lens quality,
but to get a 'resolution width' one has to rather arbitrarily choose a
mimimum acceptable contrast level.
A related method, that can be done on any image having plenty of
detail contrast, is to systematically smooth the image in small steps,
and plot the mean squared difference at each step. This measures the
'contrast power' lost due to the increase of the effective resolution
element (smoothing disk) size. If one starts with an image that has a
lot of empty magnification (true resolution element much larger than a
pixel) then the first steps will remove little contrast. When the
smoothing disk becomes equal to the 'true resolution' the lost power
will start to increase rapidly.
I think it is a silly waste of time to do such measurements on a bad
lens. But the pixel size of the 7D is small enough to reveal the
limits of even some good lenses, so you might want to try to estimate
the 'true resolution element size' if you care deeply about 'true
pixel counts'. Especially if you use zoom lenses.
Sacha
from center to edge especially for wides, cheap zooms, and fisheyes.
For instance, the final stitch might be full of fringe fuzz that
fisheye stitchers know about.
In the end I feel gp measuring is extremely difficult. So many apples
to oranges. But maybe we can get closer with your help.
Alfred
with 800mm F5.6 for the next project.
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
Tom Sharpless
Sent: Thursday, December 23, 2010 9:21 PM
To: PTGui Support
Subject: [PTGui] Re: Stitch large panorama in PTGUI
Regards, Tom
--
Sacha
so we can proceed.
PTGui Support
> The questions here are:
>
> 1.Is there a way to stitch super size panorama in PTGUI? Can we emit raw
> files so that it is not subjected to image format limit?
PTGui can create .psb files; the 300,000 pixel limit for psb was removed
in PTGui 9 so you can now create panoramas of (nearly) unlimited size.
> 2.Is there a way to import Autopano config file and process in PTGUI?
No.
> 3.Can PTGUI accurately report the optical pixel count for super size
> panorama?
If you do Optimum Size -> Maximum size, PTGui will set the output size
such that the angular resolution of the output equals the angular
resolution of the source images.
(to be more specific: the angular resolution in the center of the images)
Joost
hd_pano
first of all congratiolation to your great picture.
As far as i know you can generate a PTGui project file dirctly ouf of
Autopano Giga.
Regards and best wishes for Christmas
hd_pano
On 23 Dez., 23:44, "Alfred" <rongkai.z...@gmail.com> wrote:
> Hi, there,
>
> First of all, happy holidays to every one! I wish all of you the best in the
> coming 2011.
>
hd_pano
another remark to your picture.
I found on Gigapan.org, that you are jusing an aperture of 36. I think
on this aperture you get not the maximum sharpness of your lens
because of the diffraction. I myself tested it at a 100 mm macro and
found at a aperture smaller then 11 (larger f-stop) a significant loss
of sharpness. Maybe it is at your lens a bit better, but i think 36 is
too small.
The other point is the tele extender. In my mind i think it doesn't
extend the resolution but only the largeness of a picture. In times
the resolution of the chip is such high as today it doesn't make
sense.
cheers hd_pano
>
> First of all, happy holidays to every one! I wish all of you the best in the
> coming 2011.
>
Erik Krause
> I found on Gigapan.org, that you are jusing an aperture of 36. I think
> on this aperture you get not the maximum sharpness of your lens
> because of the diffraction. I myself tested it at a 100 mm macro and
> found at a aperture smaller then 11 (larger f-stop) a significant loss
> of sharpness. Maybe it is at your lens a bit better, but i think 36 is
> too small.
I didn't find where it is stated that the images where taken at f/36,
but that's a ridiculously small aperture. You get a diffraction circle
of about 0.025mm diameter no matter what lens you use. This reduces the
effective resolution of a 7D to about 600x920 pixel (0.55 megapixel)!
Even the largest aperture (f/5.6) of the lens gives an effective
aperture of f/11 with the tele converter. This results in a diffraction
circle of 0.007mm diameter or 2150x3300 pixels (7 megapixels) on the
APC-C sensor. This are the theoretical maximum values. Real lenses will
have less resolution.
But I can't imagine someone goes out to break a gigapixel record without
such basic knowledge. Furthermore I'd expect that the 70% rule is
applied for the final size to compensate for bayer interpolation...
--
Erik Krause
Alfred
You are correct. Choosing F36 was a bad decision. I was trying to maximize
the DOF, yet that caused three problems: 1. Diffraction become a more
visible problem; 2. Must use longer exposure time, that's more subject to
vibration due to wind; 3. Lower light and higher ISO rendered more noise.
These were the problems, I am thinking about use prime lens only in future
project and may apply focus stacking technique (taking 2 or 3 shots with
different focal plane at the same location and repeat it throughout the
panorama. That will probably get me the best resolution.
Cheers,
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
hd_pano
Sent: Friday, December 24, 2010 6:17 AM
To: PTGui Support
Subject: [PTGui] Re: Stitch large panorama in PTGUI
Hi Alfred,
cheers hd_pano
--
Alfred
Despite the harsh language, F36 was a mistake. However, I don't see a
problem as hard as you described, the individual image still look very
reasonable. It lost the sharpness I desired, but still not as bad. I think
it was partly due to diffraction only affect the immediate neighbor pixels,
in the longer range neighbor, it diminish fast. It does reduce the
sharpness.
Cheers,
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
--
Erik Krause
--
Alfred
Alfred
image diffraction compensation. I think it can be modeled accurately and
used to correct some diffraction problem.
Cheers,
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
Erik Krause
Sent: Friday, December 24, 2010 9:30 AM
To: pt...@googlegroups.com
--
Erik Krause
--
Erik Krause
> Despite the harsh language, F36 was a mistake. However, I don't see a
> problem as hard as you described, the individual image still look very
> reasonable.
May be if you don't zoom in too far. However, if 112GP is the resolution
based on a calculation like Tom suggested the image carries just as much
information as a 3.4GP image. Even if you applied the 70% rule 5GP would
have been sufficient.
It is a bitter fact that only fast prime lenses wide open use the full
resolution of an 18MP APS-C-sensor...
--
Erik Krause
Erik Krause
> One more question for the forum, does anyone know any existing literature on
> image diffraction compensation.
you can't compensate for diffraction it's a physical property of light.
You only can use an optimum aperture for a given DOF:
http://www.kenrockwell.com/tech/focus.htm
--
Erik Krause
Alfred
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
Erik Krause
Sent: Friday, December 24, 2010 10:24 AM
To: pt...@googlegroups.com
Subject: Re: [PTGui] Re: Stitch large panorama in PTGUI
--
Erik Krause
--
Alfred
pixel's color and intensity along with the model to calculate the effect on
the neighbor pixels, this needs to be optimized against some objective
function (yet to decide). Based on the result, we can adjust the pixel color
and intensity to regain the sharpness. There are similar successful cases on
haze reduction. I think that's something worth looking into.
Of course, taking the image with the optimal setting is a better way,
however, there are circumstance we need to balance.
Cheers,
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
Erik Krause
Sent: Friday, December 24, 2010 10:27 AM
To: pt...@googlegroups.com
Subject: Re: [PTGui] Re: Stitch large panorama in PTGUI
--
Erik Krause
--
Sacha
Especially for the version ii of canon's tele-extender.
It'd be a good rainy day project to verify this wild claim.
Erik Krause
> Since diffraction can be mathematically modeled, we can use the center
> pixel's color and intensity along with the model to calculate the effect on
> the neighbor pixels, this needs to be optimized against some objective
> function (yet to decide).
The details of the Airy Disk (see wikipedia f.e.) are well known.
> Based on the result, we can adjust the pixel color
> and intensity to regain the sharpness. There are similar successful cases on
> haze reduction. I think that's something worth looking into.
You can simply sharpen the image. You can use sophisticated
deconvolution algorithms to recover part of the information (MTF graphs
show the contrast loss at a certain resolution, and contrast can be
regained). But there is a point where information is completely lost due
to diffraction (contrast is below the noise level). It is impossible to
recover details that are smaller than that limit.
--
Erik Krause
Tom Sharpless
computer. Reshoot your pano at f/8 instead.
-- Tom
Alfred
Sure, when I get another chance. But it is quite expensive to even do once,
travel, hotel, time, etc. That shot cost me over $5000. I am in artificial
intelligence research, not a photographer. Shooting photos is only a hobby,
I did made some honest mistake in choosing the F-stop. Learned from error.
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
Tom Sharpless
Sent: Friday, December 24, 2010 4:40 PM
To: PTGui Support
-- Tom
--
Sacha
to date are far and away exaggerated when trying to calculate the
actual detail/information contained.
For a true count we need to know the
1. Exact field of view collected
2. Overalp Shot at
3. Average MTF (maximum theoretical detail capturing ability) of the
Lens+adapter of area of the lens being used in the final image.
For instance the center of the image will be sharper and the edges
less sharp, but in using overlap we can eliminate some of these less
sharp pixels in the final calculation.
So we can use this values to further replace focal length, aperture to
get some more realistic numbers and most likely save some bandwidth /
server space.
It's probably too late to go back and recalculate a lot of the epic
gigapans out there.
But I'd be interested to know which are some of the best in class for
resolution.
Or what about we we forget all of this.. Isn't there a way we can
analyze the images themselves for detail resolving power and calculate
that against the field of view captured?
Erik Krause
> Or what about we we forget all of this.. Isn't there a way we can
> analyze the images themselves for detail resolving power and calculate
> that against the field of view captured?
Yes there is. Take a relevant crop of the image and reduce it's size by
half. Then blow it up again to the original size. Now overlay the
original and the two times resized version in difference mode. If there
is no difference at pixel level any information in the larger version is
in the half size version as well. You can repeat this using larger
dividers (quarter, eighth...) until you see some difference. This way
you'll find the highest resolution that carries all the displayed
information.
best regards
--
Erik Krause
Joergen Geerds
> Yes there is. Take a relevant crop of the image and reduce it's size by
> half. Then blow it up again to the original size. Now overlay the
> original and the two times resized version in difference mode. If there
> is no difference at pixel level any information in the larger version is
> in the half size version as well. You can repeat this using larger
> dividers (quarter, eighth...) until you see some difference. This way
> you'll find the highest resolution that carries all the displayed
> information.
a 8x or 16 factor involved with alfred's pano.
We talked on facebook about this project (and the Sevilla pano), and I
wanted to thank tom, sacha and erik for all the info here.
joergen
Erik Krause
> I am in artificial
> intelligence research, not a photographer. Shooting photos is only a hobby,
> I did made some honest mistake in choosing the F-stop. Learned from error.
I wrote a small wiki page on diffraction with some formulas mentioning
the resolution problem:
http://wiki.panotools.org/Diffraction
Alfred
http://sepwww.stanford.edu/~morgan/texturematch/paper_html/node3.html . For
a more empirical comparison, you can see it at
http://www.gigapan.org/gigapans/60750/ . That's a 1:100 (expand 10 times on
the side). I am pretty sure, for all the large panoramas, using this method
won't reveal much difference. I mean even with the higher quality classical
ones such as Dresden 26G, Paris 26G.
I proposed to many people regarding forming an international community to
evaluate the large panoramas. Large panorama should be evaluated in terms of
information carried. Many people use empty space (sky, water surface, forest
etc) to pack pixels. These are not very interesting.
Entropy measure based on the Lapalacian pyramid can be a way to do it. But
we need to carry out some experiment and need other team to collaborate on
such experiment.
Cheers,
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
Joergen Geerds
Sent: Saturday, December 25, 2010 3:34 PM
To: PTGui Support
joergen
--
Alfred
are we saying they are all about the same in terms of information amount due
to resolution? Will be interested to see an apple to apple comparison just
based on their public information.
But there is another factor not included in this formula, many subjects are
very close to the camera, if set to a large aperture (say F8 or even F5.6),
then dof is very small, it blurs all the remote or the close subjects, then
there is another problem.
Have any of you guys taking large landscape panorama with focus stacking?
Any recommendations?
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
--
gerald.d
Why would you swap out the 7D for a 1DS Mk III?
Regards,
Gerald.
Joergen Geerds
> Why would you swap out the 7D for a 1DS Mk III?
joergen
gerald.d
myself.
But I don't think it's what Alfred would be looking for for his
panoramas.
After all, he'd get around half the number of pixels out of it for the
same lens/FoV combo than he does with his 7D.
Joergen Geerds
> Don't get me wrong - it is of course a great camera.
> same lens/FoV combo than he does with his 7D.
there is a huge difference between the 550D, the 60D or the 7D, they
all use more or less the same sensor, and much of the other components/
algorithms, at least when it comes to projects like a gigapixel pano.
joergen
Alfred
Yes, you did have one. You shoot Dubai in 7D. Was that for the same purpose.
This was public knowledge for the past half year.
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
gerald.d
Sent: Sunday, December 26, 2010 9:27 PM
To: PTGui Support
Subject: [PTGui] Re: Stitch large panorama in PTGUI
--
gerald.d
This is getting beyond a joke. I asked a simple question following
your earlier post here.
This is your earlier post:
"I use 7D with 400mm F5.6 + 2X tele-converter. I plan to change to 1DS
III
with 800mm F5.6 for the next project."
"Why would you swap out the 7D for a 1DS Mk III?"
to getting one, and then all the rumours about the 1Ds Mk IV started
flying about, so I held off).
This is NOT about me.
It was a really, really simple question just asking you why you would
consider doing your next panorama with the 1Ds Mk III rather than the
7D.
And what do I get?
Bombarded with direct emails where you again totally ignore the
question, and attempts to connect with me on both Facebook and
LinkedIn.
Frankly, I've had enough of this nonsense.
Have a good day.
Alfred
I had enough of those attack and "jokes" from you both on Gigapan and PTGUI
group. Why you been such a mean person? I was trying to be friendly with you
and set a technical dispute aside. Therefore sending you a facebook and
Linkedin invitation to be a friend. Don't be joke yourself.
You shoot Dubai in 7D and you wrote "Don't get me wrong - it is of course a
great camera. I wish I had one myself".
In order to get the highest sensor density and pack more pixel in the Dubai
image, you did the same thing as many other groups did. Now you are
pretending never had this intention nor action. This showed a very dark side
of you.
The reason I mentioned 1DS III and 800mm prime lens, of course, this camera
is not going to get the highest density, but it is going to have higher
quality. I think quality is something truly important when people get over
with the pure number competition. What was wrong?
Now the group is watching how you act!
gerald.d
I assume and respect that English isn't your first language Alfred,
but taking the feedback here and on Gigapan as "attacks and jokes" is
beyond ridiculous.
It's pretty clear now that, for whatever reason, you simply haven't
been able to follow the thread of the conversation, and if as a result
of that you have taken offence, it was certainly not of my doing.
The comment you quote below was made in reference to the 1Ds Mk III,
and not the 7D. There is simply no way that it can be interpreted as
referring to the 7D.
Just think about what you're implying here - that I shot with the 7D,
told a few million people that I shot with the 7D, and then I'm daft
enough to deny I've ever owned one?
Alfred
read out of context and taking defense. This is my final post on this
thread. Please do not waste people's time on such unfruitful fire exchange
anymore. PERIOD!
Joergen Geerds
I stumbled today about an older article by R. N. Clarke about
teleconverters and image resolution. The article itself is pretty
outdated, since his reference camera was a canon 10d, but the general
idea and methodology might help you understand why your shanghai
panorama didn't turn out perfect:
http://www.clarkvision.com/articles/relative-lens-sharpness/index.html
joergen
Hans
On Dec 24, 11:39 pm, Tom Sharpless <tksharpl...@gmail.com> wrote:
> It's a waste of time to try to recover lost resolution on the
> computer. Reshoot your pano at f/8 instead.
We are talking about an 800mm lens.
If this was a real super quality 800mm lens it woud be no problem
using it at F36 or what ever was the case.
My old APO Ronar 480 cm is fine down to F64 and goes to f90
The longer lens the smaller aperture you can use.
Using an 800mm at F8 would mean that you had to focus at 2500 meters
to get DOF from 1200 - infinity
Hans
Joergen Geerds
> > It's a waste of time to try to recover lost resolution on the
> > computer. Reshoot your pano at f/8 instead.
> to get DOF from 1200 - infinity
detail due to many portions of the tile being OOF isn't really worth
the effort unless you focus stack (which adds another layer of
computational problems). f11-f18 are usually a good compromise.
> If this was a real super quality 800mm lens it woud be no problem
> using it at F36 or what ever was the case.
killed all the details (see tom's or sacha's earlier post).
what "killed" alfreds panorama is a combination of diffraction and
motion blur (2.4 sec/shot doesn't give the rig any time to dampen any
movement).
> My old APO Ronar 480 cm is fine down to F64 and goes to f90
> The longer lens the smaller aperture you can use.
portrait lens, and as far as i know, the resolution requirements for
8x10 are far different than for a modern dslr, which require a circle
of confusion of roughly 3x the pixel size (0.019mm or 19µm for the
7d), while a CoC of 0.2mm (200µm) is quite acceptable for a 8x10 neg.
this is the reason why you can stop down a LF lens. not because it
obeys different optical laws.
and focal length has nothing to do with the f-stop in terms of
"resolution". diffraction is the same.
joergen
Erik Krause
> But there is another factor not included in this formula, many subjects are
> very close to the camera, if set to a large aperture (say F8 or even F5.6),
> then dof is very small, it blurs all the remote or the close subjects, then
> there is another problem.
Yes, that's true. You must trade DOF for diffraction. How to do that is
described by Ken Rockwell: http://www.kenrockwell.com/tech/focus.htm
> Have any of you guys taking large landscape panorama with focus stacking?
> Any recommendations?
Me not. But I'd only do it if I could set the focus by software. Since
several lens parameters change by focusing I'd use the shots at
hyperfocal distance as a basis and align all other shots to them
allowing for a different FoV and even some lens parameters. Experiments
would show what exactly changes between focus positions. Then I'd merge
the focus stacks using enfuse contrast criterion or tufuse. The results
should be stitchable as normal.
--
Erik Krause
Peter Wilson
gets smaller. f 2.8 , minimal DoF
F8 is a small aperture, not a large aperture. Anything below F5.6 is
going to get you into trouble. I have the Sunex, too, which I seldom
use. It is a fixed f5.6 ( Any buyers?)
I shoot many panos in HDRI at f 8 == 0EV , specifically to guarantee
all the near field stuff is within the DoF. I usually have furnished
rooms to contend with, and may have the camera stand ( my own design)
or the tripod actually on a dining table. The near field objects need
to be in focus. F8 == fire and forget, or. If in doubt, use F8.
Peter Wilson
loveb...@gmail.com
--
Erik Krause
--
Alfred
The post was referring to a different context. People were discussing the
technical lessons learned from http://www.gigapan.org/gigapans/66626/ . I
guess you are referring to fisheye lens and 360 spherical panorama shooting.
Cheers,
Alfred
-----Original Message-----
From: pt...@googlegroups.com [mailto:pt...@googlegroups.com] On Behalf Of
Peter Wilson
Sent: Tuesday, December 28, 2010 3:55 PM
To: pt...@googlegroups.com
Subject: Re: [PTGui] Re: Stitch large panorama in PTGUI
Joergen Geerds
> Don't quite see the logic here. The DoF diminishes as the F number
> gets smaller. f 2.8 , minimal DoF
> F8 is a small aperture, not a large aperture. Anything below F5.6 is
> going to get you into trouble.
> I have the Sunex, too, which I seldom use. It is a fixed f5.6 ( Any buyers?)
investment.
> I shoot many panos in HDRI at f 8 == 0EV , specifically to guarantee
> all the near field stuff is within the DoF. I usually have furnished
> rooms to contend with, and may have the camera stand ( my own design)
> or the tripod actually on a dining table. The near field objects need
> to be in focus. F8 == fire and forget, or. If in doubt, use F8.
function of focal length and f-stop:
http://www.dofmaster.com/dofjs.html
i very much doubt it that you shoot your interiors at f8/400mm. i am
sure your f8 rule works great for the fisheye or UWA you are probably
using.
joergen
Peter Wilson
followed the whole thread! Mea culpa!
Peter Wilson
loveb...@gmail.com
Peter Wilson
Normal for me is the Nikkor 10.5 interiors / exteriors 360 , and the
Tokina 11-16 for non-spherical
With respect to semantics, if we don't all agree on the terminology in
use, them communication ceases to happen. Every discipline has its
jargon.
Yes, I understand the optical principles very well, although I do
struggle with the maths in ' Fundamentals of Optics ( Jenkins, F A ;
White, H E; ISBN 13: 978-0070323308, ), no matter how many times I
refer to it.
Peter Wilson
loveb...@gmail.com
joergen
--
Hans
While it may start at F10 on a short lens it may not start until
f22-32 on longer lenses.
It has nothing to do with how large the circle of confusion is.
Most longer lenses has the optimal performance at F16 or f22 while
many short wide angles already has its optimal performance (at center)
at f4-5.6
Hans
>
> joergen
Joergen Geerds
> Diffraction depends on the physical size of the aperture.
> While it may start at F10 on a short lens it may not start until
> f22-32 on longer lenses.
> many short wide angles already has its optimal performance (at center)
> at f4-5.6
sorry to warm-up an old thread, but I believe your claim above isn't
completely true.
While I agree that the peak performance varies from lens to lens
(usually somewhere between f4 and f10), diffraction is the same
problem for every lens past f11.
if you look at the review for a long lens (not the longest), you see
that the resolution drops quickly if you go past f11 (f16-22):
http://www.dpreview.com/lensreviews/canon_70-200_2p8_is_usm_ii_c16/page5.asp
exactly the same happens with wide-angle lenses:
http://www.dpreview.com/lensreviews/sony_24_2_m15/page4.asp
diffraction (loss of resolution) is a problem for every lens, no
matter how long. I haven't seen a single lens review that would show
great resolution (no diffraction) for f32 on any lens (LF/MF lens
reviews and resolution charts are rare).
joergen
DennisS
> but that's a ridiculously small aperture. You get a diffraction circle
> of about 0.025mm diameter no matter what lens you use. This reduces the
> effective resolution of a 7D to about 600x920 pixel (0.55 megapixel)!
>
> Even the largest aperture (f/5.6) of the lens gives an effective
> aperture of f/11 with the tele converter. This results in a diffraction
> circle of 0.007mm diameter or 2150x3300 pixels (7 megapixels) on the
> APC-C sensor. This are the theoretical maximum values. Real lenses will
> have less resolution.
I am confused about the above statements. I thought the megapixel
count is determined by the number of pixels on the sensor, not the
chosed f stop. Doesn't the f stop limit the amount of light hitting
the sensor? I can understand if, due to defraction, you have a
smaller area that is sharp and usable. Please be gentle in your
reply. I have never done any gig panoramas and still have a lot to
learn.
John Houghton
double the size of the image in Photoshop, but there will be no more
detail available in the image. Diffraction blurrs the whole image and
thus reduces the level of detail captured. When shooting at f/32, you
could resize the image down considerably before there was any
significant loss of detail.
John
DennisS
I am asking about this statement in particular:
(0.55 megapixel)!"
image? I think I get it, I just want to make sure.
Dennis
Sacha
DennisS
"This reduces the effective resolution of a 7D to about 600x920 pixel
(0.55 megapixel)!"
the lens? Still confused.........
Sacha
led to diffraction causing a loss of detail and that this image would
have been equivalent to one 600x920 upsized to 18 megapixels. Whether
or not this is exactly the case is open to debate given the nature of
aperatures and telephotos. What's not being debated is that there
certainly is some smaller size when upsized would yield roughly the
same detail.
The larger debate is whether or not Gigapixel/megapixel image size
should have any meaningful relation to image quality.
One camp believes and image width x image height = your megapixels and
nuff said.
The other camp (including me) says that your final megapixel/gigapixel
tally should say something meaningful about how detailed the image is
otherwise what's the point of even publishing the number. There are a
lot of factors that affect the detail of the image.
If you want to test software limits, repetitive pictures of your cat
should be fine enough to make the world record gigapixel/terapixel
image.
Joergen Geerds
> The whole point was that shooting at the smallest possible aperture
> led to diffraction causing a loss of detail and that this image would
> have been equivalent to one 600x920 upsized to 18 megapixels. Whether
> or not this is exactly the case is open to debate given the nature of
> aperatures and telephotos. What's not being debated is that there
> certainly is some smaller size when upsized would yield roughly the
> same detail.
> The larger debate is whether or not Gigapixel/megapixel image size
> should have any meaningful relation to image quality.
> One camp believes and image width x image height = your megapixels and nuff said.
for publicity sake for their "record breaking pano". in contrast, Eric
and Greg from xrez have been doing great super rez panos for years,
and never done fake pixel resolutions, but rather nice and crisp and
clean work and is enjoyable to see.
But maybe it is the gigapan software that (in comparison to PTGui)
doesn't do such a great job, and leads people to believe whatever
fantasy numbers it is producing.
> The other camp (including me) says that your final megapixel/gigapixel
> tally should say something meaningful about how detailed the image is
> otherwise what's the point of even publishing the number. There are a
> lot of factors that affect the detail of the image.
"maxpixelzoom=1.0" in their panos.
on the other hand, the gigapan player doesn't seem to know what
maxpixelzoom actually is.
> If you want to test software limits, repetitive pictures of your cat
> should be fine enough to make the world record gigapixel/terapixel image.
480x240kpx, and then bask in the glory of 115 hogwash gigapixels.
here is a link to a simple 1.2 gigapixel pano that is just big enough
IMO:
http://luminous-newyork.com/img/2010-09-Inn-of-Chicago/2010-09-Inn-of-Chicago.html
joergen
Sacha
business trip.
Nice SHOT!
> here is a link to a simple 1.2 gigapixel pano that is just big enough
>
> joergen
Erik Krause
>> > Diffraction depends on the physical size of the aperture.
>> > While it may start at F10 on a short lens it may not start until
>> > f22-32 on longer lenses.
>> > Most longer lenses has the optimal performance at F16 or f22 while
>> > many short wide angles already has its optimal performance (at center)
>> > at f4-5.6
> Hans,
> sorry to warm-up an old thread, but I believe your claim above isn't
> completely true.
If you look at http://en.wikipedia.org/wiki/Rayleigh_criterion you see
that the smallest angular separation depends on the wavelength of light
and the physical diameter of the entrance pupil only. The physical
diameter of the entrance pupil is the focal length divided by the
f-number. So for the same wavelength you get the same *angular*
resolution for 20mm f/4, 40mm f/8, 80mm f/16 or 160mm f/32.
However, this does not mean that you get equally blurred images if you
use this lenses at the indicated f-stops on your camera. Since the Field
of View of the 160mm lens on a given format is far smaller than the 20mm
the same angle covers much more of the image and diffraction is much
more magnified.
But if you use formats where the FoV is comparable (an 8 times larger
format for the 160mm lens than for the 20mm one) you get a similar
amount of diffraction blur for above combinations. That's why large
format cameras can be stopped down to f/64 or even f/128.
Now for stitched images, where format is independent from focal length,
this applies same and like Hans said. If you shoot the same panorama
with same FoV with a 20mm and a 160mm lens you get the same diffraction
blur at the indicated f-stops. http://wiki.panotools.org/Diffraction
has a table where this is clearly visible, too.
--
Erik Krause
http://www.erik-krause.de
Sacha
sharpest focal length will always be wide open, and now that we can
correct for aberration, we should be able to realize this.
Looks like its time for some real world tests. I know this generally
accurate off the bat for primes.
On Jan 20, 2:37 pm, Erik Krause <erik.kra...@gmx.de> wrote:
> Am 19.01.2011 18:00, schrieb Joergen Geerds:
>
> >> > Diffraction depends on the physical size of the aperture.
> >> > While it may start at F10 on a short lens it may not start until
> >> > f22-32 on longer lenses.
> >> > Most longer lenses has the optimal performance at F16 or f22 while
> >> > many short wide angles already has its optimal performance (at center)
> >> > at f4-5.6
> > Hans,
> > sorry to warm-up an old thread, but I believe your claim above isn't
> > completely true.
>
> and the physical diameter of the entrance pupil only. The physical
> diameter of the entrance pupil is the focal length divided by the
> f-number. So for the same wavelength you get the same *angular*
> resolution for 20mm f/4, 40mm f/8, 80mm f/16 or 160mm f/32.
>
> However, this does not mean that you get equally blurred images if you
> use this lenses at the indicated f-stops on your camera. Since the Field
> of View of the 160mm lens on a given format is far smaller than the 20mm
> the same angle covers much more of the image and diffraction is much
> more magnified.
>
> But if you use formats where the FoV is comparable (an 8 times larger
> format for the 160mm lens than for the 20mm one) you get a similar
> amount of diffraction blur for above combinations. That's why large
> format cameras can be stopped down to f/64 or even f/128.
>
> Now for stitched images, where format is independent from focal length,
> this applies same and like Hans said. If you shoot the same panorama
> with same FoV with a 20mm and a 160mm lens you get the same diffraction
Erik Krause
> Interesting, according whomever wrote this wiki page, (you?), the
> sharpest focal length will always be wide open,
In terms of diffraction, yes. Unfortunately there are other sources of
image blur which are worst wide open. Most lens aberrations of course
and depth of field.
> and now that we can
> correct for aberration, we should be able to realize this.
You can correct for transversal chromatic aberration, but that's the
only one. You can't correct for longitudinal chromatic aberration, coma,
spherical aberration, astigmatism and field curvature. And there are
more, f.e. due to decentered lens elements, density gradients in the
glass and so on. All of them cause blur, especially wide open.
> Looks like its time for some real world tests.
Yes, indeed.
Joergen Geerds
> Thanks for reminding me that I need to bring my bathrobe on my next
> business trip.
> Nice SHOT!
thanks for your kind words.
here is another example I've been working on... It might be that I am
gonna change the WB to make it a bit warmer... but it is about 6
gigapixel (and about 1.5 of them are "meaningful"):
Blog post: http://newyorkpanorama.com/2010/07/26/brooklyn-bridge-2/
krpano full rez: http://luminous-newyork.com/img/2010-07_brooklyn_bridge/2010-07_brooklyn_bridge.html
This is about as big as I will ever shoot (with current technology).
comments welcome, but by all means enjoy
joergen
Sacha
compression must have been set to around 2. I hate to think that
someone not familiar with jpg might think the banding's in your image.
Its a beautiful piece of work. May it bring a lot of income!
On Jan 23, 11:56 pm, Joergen Geerds <jgee...@gmail.com> wrote:
> On Jan 20, 1:35 pm, Sacha <sachagriffin...@gmail.com> wrote:
>
> > Thanks for reminding me that I need to bring my bathrobe on my next
> > business trip.
> > Nice SHOT!
>
> hehe...
> thanks for your kind words.
> here is another example I've been working on... It might be that I am
> gonna change the WB to make it a bit warmer... but it is about 6
> gigapixel (and about 1.5 of them are "meaningful"):
> Blog post:http://newyorkpanorama.com/2010/07/26/brooklyn-bridge-2/