Hello Tom!

I am very glad to hear that You like my proposal!

There is a great chance that any implementing difficulties will not
occur. This is because Thin Plate Spline is not mesh-based algorithm.
The TPS algorithm exactly calculates new position of a pixel in
morphed image and do it fast. I think that using grids will not be
necessary at all.

I assume that the next step to become GSoC student is to send You a
timeiline. This will not be a problem because I am very familiar with
TPS algorithm and I exactly know how to implement it. Two months ago I
have written it in c# (the images
http://dl.dropbox.com/u/13127408/koala.jpg presenting the results of
TPS are generated by this program).

Please give me further instructions to become Hugin's GSoC student.

Best Regards,
Lukasz Maliszewski



W dniu 9 kwietnia 2011 19:18 użytkownik Tom Sharpless
<[email protected]> napisał:
> Hello Lukasz
>
> On Apr 8, 3:56 pm, Łukasz Maliszewski <[email protected]>
> wrote:
>> My name is Lukasz Maliszewski. I am a last year Informatics student
>> from the Technical University of Gdansk in Poland, Faculty of
>> Electronics, Telecommunications and Informatics.
>>
>> My GsoC proposal:
>>
>> My idea is to implenet Thin Plate spline.
>>
>> Thin Plate Spline lets to deform image in an unlimited way. We
>> describe how we want to bend an image by defining set of control
>> points and their destination positions. The algorithm bends an image
>> in a way that the control points are moved to the desired positions.
>>
>> The big advantage of the algorithm is that the Image after
>> transformation looks very natural. Example can be seen 
>> here:http://dl.dropbox.com/u/13127408/koala.jpg
>>
>> I think that such functionality would be useful for panorama
>> application. It is quite often that the component images are deformed
>> and control points do not overlaps. Thin plate spline could bend image
>> in the most natural way so that the control points would overlap.
>>
>
> It would be a very useful improvement to Hugin, to be able to
> designate a set of control points as "morph-to-fit" and have the
> warper do that.  This was a feature of PanoTools, but never fully
> developed or made available through the GUIs.  I foresee some
> difficulty in implementing it, since libpano does not have a general
> mesh-based warper.  It does interpolate in a mesh, but at present only
> a regular rectangular one, obtained by decimating the output pixel
> grid.  So ideally some basic architectural changes might be needed.
> You should be able to use the existing rectangular grid, since the
> thin plate spline can handle interpolation at those mesh points as
> well as any others; but the existing data structures might not be
> sufficient.
>
> If such a facility were available, it would enable us to develop
> several kinds of useful tools, for example to fix small parallax
> errors between source images, or to improve the perspective illusion
> by straightening selected lines in the final panorama.
>
> So I would encourage you to pursue this project.
>
> Regards, Tom
>
>
>
>> I have submitted this proposal on GSoC site.
>> I am aware that my proposition is incomplete and it is very late for
>> GSoC proposal but let me know if such functionality is desired by You.
>>
>> Best Regards,
>>
>> Lukasz Maliszewski
>>
>> Computer Science,
>>
>> Technical University of Gdansk in Poland.
>
> --
> You received this message because you are subscribed to the Google Groups 
> "Hugin and other free panoramic software" group.
> A list of frequently asked questions is available at: 
> http://wiki.panotools.org/Hugin_FAQ
> To post to this group, send email to [email protected]
> To unsubscribe from this group, send email to 
> [email protected]
> For more options, visit this group at http://groups.google.com/group/hugin-ptx
>

-- 
You received this message because you are subscribed to the Google Groups 
"Hugin and other free panoramic software" group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/hugin-ptx

Reply via email to