I am running a stationary wavelet transformation on a brain image. I can't understand why it shifts with each level so that the image is no longer centered. You can see that the X component shifts to the right, the Y to the bottom, the diags up and to the left. This makes it hard to reach my goal of using the SWT to detect edge activity at each scale, as it looks like I will have to re-align the images to do the calculation. Why is it like this?

enter image description here

  • $\begingroup$ Can we get the image you are using ? $\endgroup$
    – Sektor
    Jun 11 '14 at 16:25
  • $\begingroup$ SWT's filters might be "causal" filters. A causal filter causes some delay. But I am not an expert in Wavelets. $\endgroup$
    – Sohail Si
    Jul 10 '18 at 18:37
  • $\begingroup$ A bit late here, but ... I have experienced this same thing in 1D time-series analysis. I believe it is due to the fact that the SWT is non - orthonormal. The implication being that information 'leaks' between adjacent space (or in my case time) steps. It would be nice if someone could add to this with a slightly more detailed explanation. $\endgroup$
    – L Bitton
    Jan 2 '19 at 14:32

Your Answer

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

Browse other questions tagged or ask your own question.