I am doing offline signature verification for my project. While researching online I found that tha 1D log Gabor method can be used to find Hamming distance of an image. I got stuck in understanding 1D log Gabor. Wikipedia have some information. My question is how can we find the value of $f$, $f_0$ and $\sigma$ of an image given in Wikipedia 1D formula ?

Here is the link where I got idea about log Gabor can be one method for verification.


You're misunderstanding:

  • Wikipedia describes $G(f)$, the frequency response of the Log-Gabor filter. So you don't "find" $f$, it's your free variable.
  • $f_0$ and $\sigma$ are parameters of your filter, which you chose to maximize the "distinguishability" of "different" pictures. You need to come up with what "different" and "similar" means for you, and then you can optimize these parameters to make the Gabor transform of picture reflect these definitions. Notice that this is totally dependent on the image content and application you're doing.
  • $\begingroup$ So f0 and σ is the value that is determined by hit and trail ? or there is any specific mechanism for that ? $\endgroup$
    – ascc
    Aug 19 '16 at 9:26
  • $\begingroup$ @ascc trial and error is a specific mechanism :) but yes, often, for thinghs that humans find similar or not, humans are the only measure of how good a parametrization is. Usually, you'd assume that "close" values of $\sigma$ and $f_0$ yield nearly the same results, so you'd just write a program that tries let's say 64 combination of possible $\sigma$ and $f_0$ values and based on these, compares different pictures. Then you manually select which similarity measures are the best, and refine your search. $\endgroup$ Aug 19 '16 at 9:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.