Convolution for filter bank about bruges HOT 3 CLOSED

I have implemented something for this now... it's in the bleeding edge bruges, under filters.convlve(). I tried lots of ways, and found map to be a bit faster than apply_along_axis, so that's what I used. But it's basically the same as what you're doing. With a bit of fiddling with shapes, I made it work for any dimensions -- so you can convolve a 3D reflectivity volume with a 2D wavelet bank, for example.

See https://github.com/agile-geoscience/bruges/blob/develop/bruges/filters/convolve.py#L12-L31

By the way, you don't have to construct the wavelet bank with a list comprehension or anything. You can just do:

w, t = bruges.filters.ricker(0.128, 0.001, f=np.linspace(5, 100, 20), return_t=True)

Then w will have shape (20, 128).

from bruges.

Hi Diego

Are you working on this? I'd started playing around with it last year, taking off from Matt's post x lines of Python: synthetic wedge model and came up with this: Making many synthetic seismic models in Python.
Perhaps vectorizing the code in cell 17: synth = np.array([np.apply_along_axis(lambda t: np.convolve(t, w, mode='same'), axis=0, arr=r) for r in rc for w in ws ])
What do you think? Then you already have a notebook example that I can clean-up and update for Python 3.

from bruges.

Fixed with f4fc0c1

from bruges.