I recently started designing a Polyphase FIR filter in with a lanczos kernel interpolator to resample an image in Python. The goal is to downscale the image using these methods. However, I think I don't understand the intuition behind these processes.

I have a RGB image with dimensions (1000,1910,3) which is sent through the polyphase filter with the lanczos kernel, and I can manage to upsample the image. But downsampling is giving me nightmares. My grand theory is that I just don't understand the basic principle of what interpolation and downscaling actually means when applied with the polyphase filter. It is mostly the tap-parameter in the filter, and its impact, that I just don't seem to get. Is there anyone who could explain the polyphase filters function in regards to image processing simple words?

I am quite new to applied signal processing, so I apologize for my noobines.

  • 1
    $\begingroup$ good news: the math is exactly the same for 1D as for 2D signals, so simply look at how decimators with polyphase kernel/core filters are implemented. $\endgroup$ – Marcus Müller Jan 24 at 13:55
  • $\begingroup$ If you have trouble understanding down/upsampling I would start with a normal structure of the filter and once you understand how this works advance to polyphase filters. $\endgroup$ – Irreducible Jan 25 at 9:59
  • $\begingroup$ Turns out, as Marcus Müller pointed out, that the math was in fact exactly the same! Thank you. $\endgroup$ – ewol Feb 4 at 9:06

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.