How i remove repetitive structure like line or circle using notch filter in image?

Hi,I am want to remove certain repetitive structure from image in Fourier domain FFT using some filter like notch filer. In our image circle or line is the repetitive structure how can I remove in FFT domain and later reconstruct again using IFFT. Please guide me and give simple code for notch filter. thanks

Answer 1

[Image Analyst]

You can use a Fourier filter. See attached demo and adapt as needed.

[MUHAMMAD ADNAN]

Thank you for your reply. But can you guide me how can I remove line pattern from image. see like these kind of image

[Image Analyst]

See those bright spikes and lines? Zero those out. Try thresholding like I did in my demo. The diagonal line is due to the periodic sequence of diagonal lines. The "grid lines" are due to the really fine texture (of fabric?). The spots are due to the spots. If you just wanted to remove the diagonal lines, you might do better with a non-linear spatial domain filter, like a modified median filter or something.

[MUHAMMAD ADNAN]

1)But I want to remove diagonal lines in Ft domain Can you tell me any way to do that? 2)It is possible we apply Hough transform in FT domain to detect these line or not?

[Image Analyst]

It's apparent you don't really have a good intuitive feel for Fourier transform. Perhaps I do because I got a Ph.D. in Optics and the Fourier transform is the diffraction pattern, and I spend a lot of time in the lab creating diffraction patterns.

Basically, if you have a plane wave hitting a pattern, the light scatters in a direction perpendicular to the pattern in inverse to the spatial dimension. So a line will give a sinc function perpendicular to the line. And a set of lines, which is like a line convolved with a comb function, will give a set of since functions. Now because the line is really long in the spatial dimension, it will be really thin in the Fourier domain. That's why the line is a series of dots (really sinc functions) that are in a thin line perpendicular to the lines in the spatial dimension. So if you get rid of the line going from upper left to lower right, you should get rid of (or reduce visibility of) the lines in the spatial domain once inverse transformed. Sorry, I know it's confusing but it's hard to impart years of hands on intuition in a short post. Try to study up on Fourier patterns some more.