parzen window - mutual information - comp.soft-sys.matlab

"Matt J" wrote in message <jtk1ai$slm$1@newscl01ah.mathworks.com>... > "eli " <elaheh_kohan@yahoo.com> wrote in message <jtk0k0$pet$1@newscl01ah.mathworks.com>... > > Hello all > > Did any one use the parzen window method to calculate mutual information between two images? > ============ > > These guys did: > > http://dl.acm.org/citation.cfm?id=1522648.1522649 Matt thank you for replying. I did not succeed to down load this article. Can you just tell me how I choose the samples from the bistogram? this means what is my samples to make from them the density function? do you now any code to make me an idea or an example? Thank you alot. Ella

"eli " <elaheh_kohan@yahoo.com> wrote in message <jtmf84$g0k$1@newscl01ah.mathworks.com>... > "Matt J" wrote in message <jtk1ai$slm$1@newscl01ah.mathworks.com>... > > "eli " <elaheh_kohan@yahoo.com> wrote in message <jtk0k0$pet$1@newscl01ah.mathworks.com>... > > > Hello all > > > Did any one use the parzen window method to calculate mutual information between two images? > > ============ > > > > These guys did: > > > > http://dl.acm.org/citation.cfm?id=1522648.1522649 > Matt > thank you for replying. I did not succeed to down load this article. > Can you just tell me how I choose the samples from the bistogram? > this means what is my samples to make from them the density function? > do you now any code to make me an idea or an example? ============== I haven't worked with MI-registration directly, so there's limited practical advice I can give. There seem to be some relevant code contributions on the File Exchange, but I haven't tried any of them: http://www.mathworks.com/matlabcentral/fileexchange/?term=mutual+information Also, here's a link to the PDF of the article I mentioned earlier, which you should have less problems downloading http://www.iipl.is.ritsumei.ac.jp/2007/IS-2-02-P0101.pdf I'll add that I haven't read this article either, but it looks like it would contain the information you're looking for.

Matt thank you for yur helps it helps me . but it proceed toooooo slow.it means in each iteration the function's value is reduced by 0.0001. Whould you please help me what I can do to solve this problem? Thanks just as always. Ella "Matt J" wrote in message <jtmo53$np4$1@newscl01ah.mathworks.com>... > "eli " <elaheh_kohan@yahoo.com> wrote in message <jtmf84$g0k$1@newscl01ah.mathworks.com>... > > "Matt J" wrote in message <jtk1ai$slm$1@newscl01ah.mathworks.com>... > > > "eli " <elaheh_kohan@yahoo.com> wrote in message <jtk0k0$pet$1@newscl01ah.mathworks.com>... > > > > Hello all > > > > Did any one use the parzen window method to calculate mutual information between two images? > > > ============ > > > > > > These guys did: > > > > > > http://dl.acm.org/citation.cfm?id=1522648.1522649 > > Matt > > thank you for replying. I did not succeed to down load this article. > > Can you just tell me how I choose the samples from the bistogram? > > this means what is my samples to make from them the density function? > > do you now any code to make me an idea or an example? > ============== > > I haven't worked with MI-registration directly, so there's limited practical advice I can give. There seem to be some relevant code contributions on the File Exchange, but I haven't tried any of them: > > http://www.mathworks.com/matlabcentral/fileexchange/?term=mutual+information > > Also, here's a link to the PDF of the article I mentioned earlier, which you should have less problems downloading > > http://www.iipl.is.ritsumei.ac.jp/2007/IS-2-02-P0101.pdf > > I'll add that I haven't read this article either, but it looks like it would contain the information you're looking for.

"eli " <elaheh_kohan@yahoo.com> wrote in message <ju8quf$qrn$1@newscl01ah.mathworks.com>... > Matt thank you for yur helps > it helps me . but it proceed toooooo slow ========== What is "it"? Which code did you download? If you think it should be faster, maybe you can contact the author of the registration routine?

"eli " <elaheh_kohan@yahoo.com> wrote in message <jtk0k0$pet$1@newscl01ah.mathworks.com>... > Hello all > Did any one use the parzen window method to calculate mutual information between two images? Ella: Hello! I'm a graduate student and I searched the internet that I found you had solved the problem of the calculation of normalized mutual information(NMI) between two images by using the parzen window method. I saw your conversation with matt on the matlab site . I have read the paper "differentiable normalized mutual information based on parzen window method" which you mentioned in the conversation.And I got some problems during the programming,could you help me please? The questions are below: 1. I used the method in the paper above to calculate the NMI , I used two images , they are the same one . And in theory the NMI of two exactly the same images should be 2, but the method didn't give me the right answer,it's just 1.6 . so I'm wondering how is your program? 2. During the gradient caculation of NMI, in eq13. of the paper above , PM(n|u) and p(m,n|u) could be zero and so how to do with the log2(PM(n|u)) and log2(P(m,n|u))? 3. In eq14 there is a derivative of h(t) ,but h(x) has no derivative when x=0, so how to do with this problem? 4. Could you give me your program about this for reference please? I'm sorry to bother you but I want to konw the answers very much. I'm very glad that you could see and reply my message! I expect your reply. Thank you very much! Xiav