search

ShurenQi / MomentToolbox

Matlab code for the paper "A survey of orthogonal moments for image representation: Theory, implementation, and evaluation"
https://github.com/ShurenQi/MomentToolbox
27 stars 5 forks source link

Request support about the image feature descriptor extraction using FJFM_D file #1

Open Ruiwen-HE opened 3 years ago

Ruiwen-HE commented 3 years ago

Dear authors, Thank you very much for sharing your code with the community.

I would like to extract a feature vector of my input image using the FJFM_D file (I suppose D stands for descriptor?). Could you please inform me if it is correct to consider the X matrix as a feature vector or is there better way to do that ? I also noticed a special format of the values included in this matrix, for example the first value X[0][0] = -4.3893 - 1.0388i It seems to not being a simple float value ?

Thank you in advance for your support.

Best regards Ruiwen

ShurenQi commented 3 years ago

Dear He, Thank you for your attention.

In fact, image moments are usually complex-valued. You can get the descriptor that is rotation invariant by taking the absolute value of X, as I did in the "3. Pattern Recognition" folder. For more information about FJFM, you can refer to my previous paper: Robust and discriminative image representation: Fractional-order Jacobi-Fourier Moments. Pattern Recognition (PR), 2021.

Good luck and have fun.

Kind regards, Shuren Qi