| View previous topic :: View next topic |
| Author |
Message |
ravi nooka
Guest
|
 Posted: Mon Nov 03, 2008 7:56 am Post subject: seamless view of images |
 |
|
dear all,
I want to have a seamless view of 10 by 10 ,ETM+, PAN images. but
because of the radiometric difference, i am not getting seamless
view , what is the best way for processing(automatically) each image
so that we can get a seamless view. |
|
| |
|
 Back to top |
ImageAnalyst
Guest
|
| Posted: Mon Nov 03, 2008 1:12 pm Post subject: Re: seamless view of images |
|
|
On Nov 3, 2:56 am, ravi nooka <nooka.r...@gmail.com> wrote:
[quote]dear all,
I want to have a seamless view of 10 by 10 ,ETM+, PAN images. but
because of the radiometric difference, i am not getting seamless
view , what is the best way for processing(automatically) each image
so that we can get a seamless view.
------------------------------------------------------------------------------------------------------[/quote]
ravi:
You can do a global leveling of the average intensity in the image.
Or, if there is some overlap, you might try doing a graduated,
weighted average as you go from one side of the overlap to the other.
Regards,
ImageAnalyst |
|
| |
|
| Back to top |
ImageAnalyst
Guest
|
| Posted: Mon Nov 03, 2008 6:32 pm Post subject: Re: seamless view of images |
|
|
On Nov 3, 2:56 am, ravi nooka <nooka.r...@gmail.com> wrote:
[quote]dear all,
I want to have a seamless view of 10 by 10 ,ETM+, PAN images. but
because of the radiometric difference, i am not getting seamless
view , what is the best way for processing(automatically) each image
so that we can get a seamless view.
[/quote]
-------------------------------------------------------------
Related interesting video of using motors to scan a regular camera to
make gigapixel images. Then they need to get stitched together.
They>ll undoubtedly have the same problems as you.
http://blog.makezine.com/archive/2008/11/take_a_gigapixel_photo_wi.html?CMP=OTC-0D6B48984890 |
|
| |
|
| Back to top |
Martin Leese
Guest
|
| Posted: Tue Nov 04, 2008 12:01 am Post subject: Re: seamless view of images |
|
|
ravi nooka wrote:
[quote]dear all,
I want to have a seamless view of 10 by 10 ,ETM+, PAN images. but
because of the radiometric difference, i am not getting seamless
view , what is the best way for processing(automatically) each image
so that we can get a seamless view.
[/quote]
If I have understood your problem, you want
to match the images for graytone. You need
to use "histogram matching".
Note the warning in this post:
http://groups.google.ca/group/sci.image.processing/browse_thread/thread/eb7b1a3a0c614292/df6fba9c926b2913?hl=en&lnk=st&q=#df6fba9c926b2913
then look at this thread for code:
http://groups.google.ca/group/sci.image.processing/browse_frm/thread/e4d95a591536d9a6/5bafd57e29e20ece?q=insubject:%22histogram+insubject:match+insubject:code+insubject:required%22&rnum=8&hl=en#5bafd57e29e20ece
I am not sure making the process completely
automatic is a good idea. But you can
certainly make it semi-automatic without
too much danger of it going wrong.
The other approach is the one used in making
black & white photo mosacis. You first use
a dodging printer. This performs a gentle
high-pass filter, and so makes all the
photo prints uniformly gray.
--
Regards,
Martin Leese
E-mail: please@see.Web.for.e-mail.INVALID
Web: http://members.tripod.com/martin_leese/ |
|
| |
|
| Back to top |
aruzinsky
Guest
|
| Posted: Tue Nov 04, 2008 4:38 pm Post subject: Re: seamless view of images |
|
|
On Nov 3, 1:56 am, ravi nooka <nooka.r...@gmail.com> wrote:
[quote]dear all,
I want to have a seamless view of 10 by 10 ,ETM+, PAN images. but
because of the radiometric difference, i am not getting seamless
view , what is the best way for processing(automatically) each image
so that we can get a seamless view.
[/quote]
I am not sure what "ETM+, PAN images" are but seamless stitching of
regular images is usually implemented by seamful stitching of wavelet
or Laplace pyramids and then taking the inverse. I recommend Laplace
pyramid because wavelets are relatively useless for anything but
compression. |
|
| |
|
| Back to top |
Martin Leese
Guest
|
| Posted: Wed Nov 05, 2008 3:44 am Post subject: Re: seamless view of images |
|
|
aruzinsky wrote:
[quote]On Nov 3, 1:56 am, ravi nooka <nooka.r...@gmail.com> wrote:
dear all,
I want to have a seamless view of 10 by 10 ,ETM+, PAN images. but
because of the radiometric difference, i am not getting seamless
view , what is the best way for processing(automatically) each image
so that we can get a seamless view.
I am not sure what "ETM+, PAN images" are but seamless stitching of
regular images is usually implemented by seamful stitching of wavelet
or Laplace pyramids and then taking the inverse. I recommend Laplace
pyramid because wavelets are relatively useless for anything but
compression.
[/quote]
The Enhanced Thematic Mapper plus (ETM+) is
a sensor on the Landsat 7 remote sensing
satellite. The images are delivered as
170 x 183 km scenes. Band 8 of the sensor
produces 15 metre resolution panchromatic
(greytone) images. (The size of each image
is therefore about 11,000 x 12,000 pixels.)
The original poster suggested he was trying
to produce a mosaic of 10 x 10 of these.
The sensor actually images a continuous
183 km wide swath, so the first trick is to
ensure he obtains images taken on the same
satellite path. Mosaicing in the
North/South direction then simply consists
of removing the overlap between scenes
(which is artificially inserted on the
ground.) This guarantees perfect geometric
and radiometric matching in one direction.
What is now left is the task of mosaicing 10
strips in the East/West direction.
Radiometric differences can be reduced by
obtaining strips of images close in time.
(Even this does not guarantee a good match
because, if it rained, then damp soil will
have a different tone from dry soil.)
How to proceed really depends on the terrain
being mosaiced. If it is consistent
(although this is unlikely given the large
area covered) then starting with a strip in
the centre and matching each adjacent strip
in turn might work. If the terrain varies
widely then he will be reduced to some sort
of high-pass filtering. In my previous post
I gave suggestions for each of these
approaches.
I have never used pyramids, but wonder
whether they are practical given the number
of pixels involved.
--
Regards,
Martin Leese
E-mail: please@see.Web.for.e-mail.INVALID
Web: http://members.tripod.com/martin_leese/ |
|
| |
|
| Back to top |
aruzinsky
Guest
|
| Posted: Wed Nov 05, 2008 4:22 pm Post subject: Re: seamless view of images |
|
|
On Nov 4, 3:44 pm, Martin Leese <ple...@see.Web.for.e-mail.INVALID>
wrote:
[quote]aruzinsky wrote:
On Nov 3, 1:56 am, ravi nooka <nooka.r...@gmail.com> wrote:
dear all,
I want to have a seamless view of 10 by 10 ,ETM+, PAN images.. but
because of the radiometric difference, i am not getting seamless
view , what is the best way for processing(automatically) each image
so that we can get a seamless view.
I am not sure what "ETM+, PAN images" are but seamless stitching of
regular images is usually implemented by seamful stitching of wavelet
or Laplace pyramids and then taking the inverse. I recommend Laplace
pyramid because wavelets are relatively useless for anything but
compression.
The Enhanced Thematic Mapper plus (ETM+) is
a sensor on the Landsat 7 remote sensing
satellite. The images are delivered as
170 x 183 km scenes. Band 8 of the sensor
produces 15 metre resolution panchromatic
(greytone) images. (The size of each image
is therefore about 11,000 x 12,000 pixels.)
The original poster suggested he was trying
to produce a mosaic of 10 x 10 of these.
The sensor actually images a continuous
183 km wide swath, so the first trick is to
ensure he obtains images taken on the same
satellite path. Mosaicing in the
North/South direction then simply consists
of removing the overlap between scenes
(which is artificially inserted on the
ground.) This guarantees perfect geometric
and radiometric matching in one direction.
What is now left is the task of mosaicing 10
strips in the East/West direction.
Radiometric differences can be reduced by
obtaining strips of images close in time.
(Even this does not guarantee a good match
because, if it rained, then damp soil will
have a different tone from dry soil.)
How to proceed really depends on the terrain
being mosaiced. If it is consistent
(although this is unlikely given the large
area covered) then starting with a strip in
the centre and matching each adjacent strip
in turn might work. If the terrain varies
widely then he will be reduced to some sort
of high-pass filtering. In my previous post
I gave suggestions for each of these
approaches.
I have never used pyramids, but wonder
whether they are practical given the number
of pixels involved.
--
Regards,
Martin Leese
E-mail: ple...@see.Web.for.e-mail.INVALID
Web:http://members.tripod.com/martin_leese/
[/quote]
On further consideration, I don>t think pyramids are popular for
stitching. This paper, http://people.csail.mit.edu/alevin/papers/eccv04-blending.pdf
, suggests gradient domain blending and I have used this for seamless
carving and cloning. It is slow unless some sophisticated numerical
method, e.g., http://research.microsoft.com/vision/visionbasedmodeling/publications/Szeliski-SG06.pdf
, is used. I have experience with Laplace pyramids in the context of
focus stacking and stacking for quasi-tonemapped HDR ,
http://research.edm.uhasselt.be/~tmertens/papers/exposure_fusion_reduced.pdf
.. Laplace pyramids are surprisingly fast and effective in this
application. With minor modifications, this method can be applied
to stitching (Just pretend that each tile is the focused part of a
larger image to be stacked). |
|
| |
|
| Back to top |
|
