Prev: Set Focus on a specific cell in a ui table
Next: fuction "primarykey" does not work
From: Lulu Ardiansyah on 18 Jan 2010 10:15
I try to perform image, using algorithms as listed on Wikipedia (http://en.wikipedia.org/wiki/Image_registration). There were some questions about it, namely:
[-] I managed to find the delta x and delta y, then what should I do with that information?
[-] How do I combine several images into one image in MATLAB?
[-] If I use three images or more. how to register your image?
Previously I thank you for your response.
From: Rob Campbell on 18 Jan 2010 11:11
Firstly, if you want specific and detailed help you need to be more explicit in what you're trying to achieve. Your current questions are rather vague. More specific details on what you're trying to achieve would help.

You say you calculated the deltas. I guess these refer to translation shifts. In which case you could create a matrix of zeros larger than your images and insert your images into the center plus the offset due to the deltas. Also see help padarray. May be useful if you want to overlay one image onto another but with an offset.

> [-] How do I combine several images into one image in MATLAB?
I'll assume the images are all the same size and that you want to overlay them rather than tile them. In which case you could make sure they are all normalised the same way then add them up and divide by the number of images. Very simple.

> [-] If I use three images or more. how to register your image?
I would choose one image as the baseline and register everything with respect to that. If noise is a concern you could choose the mean of several images and register to that.

Finally, you will probably find this useful:
www.mathworks.fr/matlabcentral/fileexchange/21451

You don't say what sort of images these are. You may find this useful:
www.mathworks.fr/matlabcentral/fileexchange/26322-mat2im
From: Lulu Ardiansyah on 19 Jan 2010 02:16
"Rob Campbell" <matlab(a)robertREMOVEcampbell.removethis.co.uk> wrote in message <hj216m$kj7$1(a)fred.mathworks.com>...
> Firstly, if you want specific and detailed help you need to be more explicit in what you're trying to achieve. Your current questions are rather vague. More specific details on what you're trying to achieve would help.
>
> You say you calculated the deltas. I guess these refer to translation shifts. In which case you could create a matrix of zeros larger than your images and insert your images into the center plus the offset due to the deltas. Also see help padarray. May be useful if you want to overlay one image onto another but with an offset.
>
> > [-] How do I combine several images into one image in MATLAB?
> I'll assume the images are all the same size and that you want to overlay them rather than tile them. In which case you could make sure they are all normalised the same way then add them up and divide by the number of images. Very simple.
>
> > [-] If I use three images or more. how to register your image?
> I would choose one image as the baseline and register everything with respect to that. If noise is a concern you could choose the mean of several images and register to that.
>
> Finally, you will probably find this useful:
> www.mathworks.fr/matlabcentral/fileexchange/21451
>
> You don't say what sort of images these are. You may find this useful:
> www.mathworks.fr/matlabcentral/fileexchange/26322-mat2im

Sorry if my question is not clear, but I mean the same as your assumptions.
Yes, I want to find the value of translational shift in the existing method (http://en.wikipedia.org/wiki/Image_registration) and Extension of phase correlation to subpixel registration. (Foroosh, H., Zerubia, Berthod J. B.).

Maybe I'll ask again, now I will try your suggestion.

Oh, one more may be on the out-off-topic, am I allowed to upload the results of the implementation of a paper (Extension of phase correlation to subpixel registration (Foroosh, H., Zerubia, Berthod J. B.)) here? Because I am not sure of the implementation, and whether not violate rights? I think if I upload the implementation will be in response or even improved by others.

You have been very helpful, thank you very much. :)
From: Rob Campbell on 19 Jan 2010 08:00
"
> Oh, one more may be on the out-off-topic, am I allowed to upload the results of the implementation of a paper (Extension of phase correlation to subpixel registration (Foroosh, H., Zerubia, Berthod J. B.)) here? Because I am not sure of the implementation, and whether not violate rights? I think if I upload the implementation will be in response

Research papers are public domain and if you've generated an implementation of an idea, then I'd say you're free to do as you like. The code sounds like it could be of general use so go ahead and send it to the file exchange.
From: ImageAnalyst on 19 Jan 2010 08:32
Like Rob says, if someone publishes something then they pretty much
expect people to implement their idea. Otherwise, why publish? - just
keep it a trade secret. They might even hope that people do or even
be excited to help you. So even though their actual paper is
copyrighted, you own your implementation of that algorithm. People
test other people's algorithms and publish results all the time -
virtually every paper out there compares their work to the work of
other prior papers. There are very few cases where someone has
patented an algorithm, for examples, GIF, LZ compression (both now
expired), SIFT - http://en.wikipedia.org/wiki/Scale-invariant_feature_transform),
etc. And some pundits are questioning whether those will survive
after the Bilski case now in the US Supreme Court (http://
en.wikipedia.org/wiki/Bilski). Just check with the authors if you're
unsure.
 |  Next  |  Last
Pages: 1 2
Prev: Set Focus on a specific cell in a ui table
Next: fuction "primarykey" does not work