search

PyWavelets / pywt

PyWavelets - Wavelet Transforms in Python
http://pywavelets.readthedocs.org
MIT License
2.04k stars 470 forks source link

Another reconstruct question #564

Closed rolfjl closed 4 years ago

rolfjl commented 4 years ago

Hi developers,

I have a question regarding waverecn and the output dimension. Do you suggest that I trim the output to get the same dimention as the original signal? Or is this a bug?

import numpy import pywt input_data=numpy.ones((22,155,128)) wdec=pywt.wavedecn(input_data,'db2','symmetric',2) output_data = pywt.waverecn(wdec,'db2','symmetric')
output_data.shape

The code returns: (22, 156, 128)

Best regards Rolf Lorentzen

grlee77 commented 4 years ago

Yes, you can discard the final, extra coefficient on axis 1 in your example. If I recall correctly, such coefficients are automatically trimmed at intermediate stages of a multilevel reconstruction, but there is no reference signal to know the size unambiguously at the last level and it can be up to 1 element larger than the starting signal. The user currently has to discard this extra sample.

I will close this is something we cannot easily fixed without switching to something like a class-based API where a WaveletTransform object could store the original shape and do the trimming for the user.

This has been a common question in the past, so we should probably add a FAQ or something explaining the issue.