PanoTools NG is a Restricted Group with 1942 members.
 PanoTools NG

 Restricted Group,
 1942 members
converting cylindrical images to equirectangulars
Expand Messages
 Hi all,
there is an interesting discussion on the PTgui mailing list on how to
convert cylindrical images to equirectangular.
John Houghton points out that if the cylindrical image has been cropped
non symmetrically, then some attempts should be made to restore the
cropped areas, albeit blank.
Which leads to my question: is there a way to identify where the horizon
runs in a cropped cylindrical image, if possible programmatically, so
that it is possible to restore those "cropped areas" before translating
to equirect?
Yuv

Copyright © 2007 Yuval Levy
Verbatim copying and distribution on other medium than YahooGroup
strictly forbidden.  Hi all,
there is an interesting discussion on the PTgui mailing list on how to
convert cylindrical images to equirectangular.
John Houghton points out that if the cylindrical image has been cropped
non symmetrically, then some attempts should be made to restore the
cropped areas, albeit blank.
Which leads to my question: is there a way to identify where the horizon
runs in a cropped cylindrical image, if possible programmatically, so
that it is possible to restore those "cropped areas" before translating
to equirect?
Yuv

Copyright © 2007 Yuval Levy
Verbatim copying and distribution on other medium than YahooGroup
strictly forbidden.  Hi all,
there is an interesting discussion on the PTgui mailing list on how to
convert cylindrical images to equirectangular.
John Houghton points out that if the cylindrical image has been cropped
non symmetrically, then some attempts should be made to restore the
cropped areas, albeit blank.
Which leads to my question: is there a way to identify where the horizon
runs in a cropped cylindrical image, if possible programmatically, so
that it is possible to restore those "cropped areas" before translating
to equirect?
Yuv

Copyright © 2007 Yuval Levy
Verbatim copying and distribution on other medium than YahooGroup
strictly forbidden.  Hi all,
there is an interesting discussion on the PTgui mailing list on how to
convert cylindrical images to equirectangular.
John Houghton points out that if the cylindrical image has been cropped
non symmetrically, then some attempts should be made to restore the
cropped areas, albeit blank.
Which leads to my question: is there a way to identify where the horizon
runs in a cropped cylindrical image, if possible programmatically, so
that it is possible to restore those "cropped areas" before translating
to equirect?
Yuv

Copyright © 2007 Yuval Levy
Verbatim copying and distribution on other medium than YahooGroup
strictly forbidden.  Hi all,
there is an interesting discussion on the PTgui mailing list on how to
convert cylindrical images to equirectangular.
John Houghton points out that if the cylindrical image has been cropped
non symmetrically, then some attempts should be made to restore the
cropped areas, albeit blank.
Which leads to my question: is there a way to identify where the horizon
runs in a cropped cylindrical image, if possible programmatically, so
that it is possible to restore those "cropped areas" before translating
to equirect?
Yuv

Copyright © 2007 Yuval Levy
Verbatim copying and distribution on other medium than YahooGroup
strictly forbidden.  Hi all,
there is an interesting discussion on the PTgui mailing list on how to
convert cylindrical images to equirectangular.
John Houghton points out that if the cylindrical image has been cropped
non symmetrically, then some attempts should be made to restore the
cropped areas, albeit blank.
Which leads to my question: is there a way to identify where the horizon
runs in a cropped cylindrical image, if possible programmatically, so
that it is possible to restore those "cropped areas" before translating
to equirect?
Yuv

Copyright © 2007 Yuval Levy
Verbatim copying and distribution on other medium than YahooGroup
strictly forbidden.   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
totally! The horizon should be exactly in the middle.
> Hi all,
>
> there is an interesting discussion on the PTgui mailing list on how to
> convert cylindrical images to equirectangular.
>
> John Houghton points out that if the cylindrical image has been cropped
> non symmetrically, then some attempts should be made to restore the
> cropped areas, albeit blank.
>
> Which leads to my question: is there a way to identify where the horizon
Find the horizon (the only way to do this is to inspect the image).
> runs in a cropped cylindrical image, if possible programmatically, so
> that it is possible to restore those "cropped areas" before translating
> to equirect?
>
All you need is a point. Then compute the distance to the top and
bottom edges. If it is not the same, enlarge the smaller side so it
matches. Now you have a "complete" cylindrical that you can easily map
to equirectangular.
dmg   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
totally! The horizon should be exactly in the middle.
> Hi all,
>
> there is an interesting discussion on the PTgui mailing list on how to
> convert cylindrical images to equirectangular.
>
> John Houghton points out that if the cylindrical image has been cropped
> non symmetrically, then some attempts should be made to restore the
> cropped areas, albeit blank.
>
> Which leads to my question: is there a way to identify where the horizon
Find the horizon (the only way to do this is to inspect the image).
> runs in a cropped cylindrical image, if possible programmatically, so
> that it is possible to restore those "cropped areas" before translating
> to equirect?
>
All you need is a point. Then compute the distance to the top and
bottom edges. If it is not the same, enlarge the smaller side so it
matches. Now you have a "complete" cylindrical that you can easily map
to equirectangular.
dmg   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
totally! The horizon should be exactly in the middle.
> Hi all,
>
> there is an interesting discussion on the PTgui mailing list on how to
> convert cylindrical images to equirectangular.
>
> John Houghton points out that if the cylindrical image has been cropped
> non symmetrically, then some attempts should be made to restore the
> cropped areas, albeit blank.
>
> Which leads to my question: is there a way to identify where the horizon
Find the horizon (the only way to do this is to inspect the image).
> runs in a cropped cylindrical image, if possible programmatically, so
> that it is possible to restore those "cropped areas" before translating
> to equirect?
>
All you need is a point. Then compute the distance to the top and
bottom edges. If it is not the same, enlarge the smaller side so it
matches. Now you have a "complete" cylindrical that you can easily map
to equirectangular.
dmg   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
totally! The horizon should be exactly in the middle.
> Hi all,
>
> there is an interesting discussion on the PTgui mailing list on how to
> convert cylindrical images to equirectangular.
>
> John Houghton points out that if the cylindrical image has been cropped
> non symmetrically, then some attempts should be made to restore the
> cropped areas, albeit blank.
>
> Which leads to my question: is there a way to identify where the horizon
Find the horizon (the only way to do this is to inspect the image).
> runs in a cropped cylindrical image, if possible programmatically, so
> that it is possible to restore those "cropped areas" before translating
> to equirect?
>
All you need is a point. Then compute the distance to the top and
bottom edges. If it is not the same, enlarge the smaller side so it
matches. Now you have a "complete" cylindrical that you can easily map
to equirectangular.
dmg   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
totally! The horizon should be exactly in the middle.
> Hi all,
>
> there is an interesting discussion on the PTgui mailing list on how to
> convert cylindrical images to equirectangular.
>
> John Houghton points out that if the cylindrical image has been cropped
> non symmetrically, then some attempts should be made to restore the
> cropped areas, albeit blank.
>
> Which leads to my question: is there a way to identify where the horizon
Find the horizon (the only way to do this is to inspect the image).
> runs in a cropped cylindrical image, if possible programmatically, so
> that it is possible to restore those "cropped areas" before translating
> to equirect?
>
All you need is a point. Then compute the distance to the top and
bottom edges. If it is not the same, enlarge the smaller side so it
matches. Now you have a "complete" cylindrical that you can easily map
to equirectangular.
dmg   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
totally! The horizon should be exactly in the middle.
> Hi all,
>
> there is an interesting discussion on the PTgui mailing list on how to
> convert cylindrical images to equirectangular.
>
> John Houghton points out that if the cylindrical image has been cropped
> non symmetrically, then some attempts should be made to restore the
> cropped areas, albeit blank.
>
> Which leads to my question: is there a way to identify where the horizon
Find the horizon (the only way to do this is to inspect the image).
> runs in a cropped cylindrical image, if possible programmatically, so
> that it is possible to restore those "cropped areas" before translating
> to equirect?
>
All you need is a point. Then compute the distance to the top and
bottom edges. If it is not the same, enlarge the smaller side so it
matches. Now you have a "complete" cylindrical that you can easily map
to equirectangular.
dmg   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
totally! The horizon should be exactly in the middle.
> Hi all,
>
> there is an interesting discussion on the PTgui mailing list on how to
> convert cylindrical images to equirectangular.
>
> John Houghton points out that if the cylindrical image has been cropped
> non symmetrically, then some attempts should be made to restore the
> cropped areas, albeit blank.
>
> Which leads to my question: is there a way to identify where the horizon
Find the horizon (the only way to do this is to inspect the image).
> runs in a cropped cylindrical image, if possible programmatically, so
> that it is possible to restore those "cropped areas" before translating
> to equirect?
>
All you need is a point. Then compute the distance to the top and
bottom edges. If it is not the same, enlarge the smaller side so it
matches. Now you have a "complete" cylindrical that you can easily map
to equirectangular.
dmg   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
totally! The horizon should be exactly in the middle.
> Hi all,
>
> there is an interesting discussion on the PTgui mailing list on how to
> convert cylindrical images to equirectangular.
>
> John Houghton points out that if the cylindrical image has been cropped
> non symmetrically, then some attempts should be made to restore the
> cropped areas, albeit blank.
>
> Which leads to my question: is there a way to identify where the horizon
Find the horizon (the only way to do this is to inspect the image).
> runs in a cropped cylindrical image, if possible programmatically, so
> that it is possible to restore those "cropped areas" before translating
> to equirect?
>
All you need is a point. Then compute the distance to the top and
bottom edges. If it is not the same, enlarge the smaller side so it
matches. Now you have a "complete" cylindrical that you can easily map
to equirectangular.
dmg   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
totally! The horizon should be exactly in the middle.
> Hi all,
>
> there is an interesting discussion on the PTgui mailing list on how to
> convert cylindrical images to equirectangular.
>
> John Houghton points out that if the cylindrical image has been cropped
> non symmetrically, then some attempts should be made to restore the
> cropped areas, albeit blank.
>
> Which leads to my question: is there a way to identify where the horizon
Find the horizon (the only way to do this is to inspect the image).
> runs in a cropped cylindrical image, if possible programmatically, so
> that it is possible to restore those "cropped areas" before translating
> to equirect?
>
All you need is a point. Then compute the distance to the top and
bottom edges. If it is not the same, enlarge the smaller side so it
matches. Now you have a "complete" cylindrical that you can easily map
to equirectangular.
dmg   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
totally! The horizon should be exactly in the middle.
> Hi all,
>
> there is an interesting discussion on the PTgui mailing list on how to
> convert cylindrical images to equirectangular.
>
> John Houghton points out that if the cylindrical image has been cropped
> non symmetrically, then some attempts should be made to restore the
> cropped areas, albeit blank.
>
> Which leads to my question: is there a way to identify where the horizon
Find the horizon (the only way to do this is to inspect the image).
> runs in a cropped cylindrical image, if possible programmatically, so
> that it is possible to restore those "cropped areas" before translating
> to equirect?
>
All you need is a point. Then compute the distance to the top and
bottom edges. If it is not the same, enlarge the smaller side so it
matches. Now you have a "complete" cylindrical that you can easily map
to equirectangular.
dmg  dmgalpha wrote:
> Find the horizon (the only way to do this is to inspect the image).
wouldn't there be a way to find it programmatically, e.g. through the
curves in nonvertical lines?
Yuv

Copyright © 2007 Yuval Levy
Verbatim copying and distribution on other medium than YahooGroup
strictly forbidden.  dmgalpha wrote:
> Find the horizon (the only way to do this is to inspect the image).
wouldn't there be a way to find it programmatically, e.g. through the
curves in nonvertical lines?
Yuv

Copyright © 2007 Yuval Levy
Verbatim copying and distribution on other medium than YahooGroup
strictly forbidden.  Actually you don't have to modify the cylindrical image; instead you
should be able to adjust the vertical shift (e parameter) until the
horizon is in the middle. I think this can be done in all guis, in PTGui
it's on the Lens Settings tab.
John is right: if not done properly, and the resulting equirect is
loaded in a panorama viewer, straight lines will become curved.
Joost
dmgalpha wrote:>  In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>> Hi all,
>>
>> there is an interesting discussion on the PTgui mailing list on how to
>> convert cylindrical images to equirectangular.
>>
>> John Houghton points out that if the cylindrical image has been cropped
>> non symmetrically, then some attempts should be made to restore the
>> cropped areas, albeit blank.
>>
>
> totally! The horizon should be exactly in the middle.
>
>
>> Which leads to my question: is there a way to identify where the horizon
>> runs in a cropped cylindrical image, if possible programmatically, so
>> that it is possible to restore those "cropped areas" before translating
>> to equirect?
>>
>
> Find the horizon (the only way to do this is to inspect the image).
> All you need is a point. Then compute the distance to the top and
> bottom edges. If it is not the same, enlarge the smaller side so it
> matches. Now you have a "complete" cylindrical that you can easily map
> to equirectangular.  On 29 Mar 2007, at 01:17, Yuval Levy wrote:
> dmgalpha wrote:
I would think setting a series of vertical line points around the
>> Find the horizon (the only way to do this is to inspect the image).
>
> wouldn't there be a way to find it programmatically, e.g. through the
> curves in nonvertical lines?
>
> Yuv
scene and optimising vertical offset would do the job.
Ian  Ian Wood wrote:
> On 29 Mar 2007, at 01:17, Yuval Levy wrote:
that's still manual. I rephrase my wish: a way to feed a ton of
>> dmgalpha wrote:
>>> Find the horizon (the only way to do this is to inspect the image).
>>>
>> wouldn't there be a way to find it programmatically, e.g. through the
>> curves in nonvertical lines?
>
> I would think setting a series of vertical line points around the
> scene and optimising vertical offset would do the job.
asymmetrically cropped cylinders into some piece of code that would find
out how where the horizon is and fill in the blank?
Yuv

Copyright © 2007 Yuval Levy
Verbatim copying and distribution on other medium than YahooGroup
strictly forbidden.   In PanoToolsNG@yahoogroups.com, Yuval Levy <yahoo06@...> wrote:
>
Do you want the theoretical answer? I suspect the answer is yes. You
> Ian Wood wrote:
> > On 29 Mar 2007, at 01:17, Yuval Levy wrote:
> >> dmgalpha wrote:
>
> >>> Find the horizon (the only way to do this is to inspect the image).
> >>>
> >> wouldn't there be a way to find it programmatically, e.g. through the
> >> curves in nonvertical lines?
> >
> > I would think setting a series of vertical line points around the
> > scene and optimising vertical offset would do the job.
>
> that's still manual. I rephrase my wish: a way to feed a ton of
> asymmetrically cropped cylinders into some piece of code that would find
> out how where the horizon is and fill in the blank?
>
> Yuv
might have to do feature extraction, determine potential horizontal
lines, and from their curvature estimate the location of the horizon.
How good will it work? I don't know.
For me the practical answer is to create a simple user interface where
a user can choose the point where the horizon is and move to the next
image.
This task is not one that is very likely to be needed by other
individuals, so the programming effort to solve it automatically might
be larger than the nonskill effort needed to pinpoint the horizon.
dmg  dmgalpha wrote:
> This task is not one that is very likely to be needed by other
as so often, you are probably right.
> individuals, so the programming effort to solve it automatically might
> be larger than the nonskill effort needed to pinpoint the horizon.
Yuv

Copyright © 2007 Yuval Levy
Verbatim copying and distribution on other medium than YahooGroup
strictly forbidden.   In PanoToolsNG@yahoogroups.com, Joost Nieuwenhuijse <imim@...> wrote:
>
PTGui
> Actually you don't have to modify the cylindrical image; instead you
> should be able to adjust the vertical shift (e parameter) until the
> horizon is in the middle. I think this can be done in all guis, in
> it's on the Lens Settings tab.
Using panotools or similar tools is like using a hammer to kill a fly.
>
> John is right: if not done properly, and the resulting equirect is
> loaded in a panorama viewer, straight lines will become curved.
>
> Joost
You don't need to remap the image. All you need is to fill the empty
space.
dmg  dmgalpha wrote:
>  In PanoToolsNG@yahoogroups.com, Joost Nieuwenhuijse <imim@...> wrote:
But the question was about remapping from cylindrical to equirect.. So
>> Actually you don't have to modify the cylindrical image; instead you
>> should be able to adjust the vertical shift (e parameter) until the
>> horizon is in the middle. I think this can be done in all guis, in
> PTGui
>> it's on the Lens Settings tab.
>>
>> John is right: if not done properly, and the resulting equirect is
>> loaded in a panorama viewer, straight lines will become curved.
>>
>> Joost
>
> Using panotools or similar tools is like using a hammer to kill a fly.
> You don't need to remap the image. All you need is to fill the empty
> space.
you can do it in one go, without editing the source image.
Joost   In PanoToolsNG@yahoogroups.com, Joost Nieuwenhuijse <imim@...> wrote:
>
You are combining two problems into 1.
> dmgalpha wrote:
> >  In PanoToolsNG@yahoogroups.com, Joost Nieuwenhuijse <imim@> wrote:
> >> Actually you don't have to modify the cylindrical image; instead you
> >> should be able to adjust the vertical shift (e parameter) until the
> >> horizon is in the middle. I think this can be done in all guis, in
> > PTGui
> >> it's on the Lens Settings tab.
> >>
> >> John is right: if not done properly, and the resulting equirect is
> >> loaded in a panorama viewer, straight lines will become curved.
> >>
> >> Joost
> >
> > Using panotools or similar tools is like using a hammer to kill a fly.
> > You don't need to remap the image. All you need is to fill the empty
> > space.
>
> But the question was about remapping from cylindrical to equirect.. So
> you can do it in one go, without editing the source image.
>
> Joost
>
Problem one: finding the horizon in the original image. You suggest to
use PToptimizer to do this. In my opinion this is an overkill.
Problem two. Remap from cylindrical to equirectilinear. This requires
PTmender (or similar).
dmg  On 30 Mar 2007, at 09:55, dmgalpha wrote:
> Problem one: finding the horizon in the original image. You suggest to
Actually, that was my suggestion not Joost's.
> use PToptimizer to do this. In my opinion this is an overkill.
Ian
Your message has been successfully submitted and would be delivered to recipients shortly.