In CWT code, the wavelet is first integrated:
int_psi, x = integrate_wavelet(wavelet, precision=precision)
then convolved:
conv[n, :] = np.convolve(data[n], int_psi_scale)
and then finite difference of convolution is taken:
coef = - np.sqrt(scale) * np.diff(conv, axis=-1)
which is equivalent to not integrating the wavelet in the first place.
Why is it done in such a way?
In CWT code, the wavelet is first integrated:
int_psi, x = integrate_wavelet(wavelet, precision=precision)then convolved:
conv[n, :] = np.convolve(data[n], int_psi_scale)and then finite difference of convolution is taken:
coef = - np.sqrt(scale) * np.diff(conv, axis=-1)which is equivalent to not integrating the wavelet in the first place.
Why is it done in such a way?