Hello all,
Is anyone aware of a linear convolution implementation?
The Base.conv and Base.conv2 functions are implemented with fft which makes
them circular convolution functions (as far as I know).
I'm looking for something alike Matlabs conv2 or SciPys signal.convolve2d.
Should be
Both conv and conv2 are linear convolutions but the implementations use the
fft. Maybe the documentation could be more clear on that.
2014-03-04 13:01 GMT+01:00 Oliver Lylloff oliverlyll...@gmail.com:
Thanks Tim.
Can't believe I missed that - been working with Images.jl all day. Nice
job by
Yes, with sufficient padding, you can compute a linear convolution (of
finite length vectors) exactly using a circular convolution. The FFT might
introduce a little noise in the result, but that is all.
On Tuesday, 4 March 2014 13:12:48 UTC+1, Oliver Lylloff wrote:
Well ok,
Maybe I
Tim,
a little bit offtopic question but might it make sense to break of the
algorithmic parts of Images.jl and put it into some signal processing
package?
I know that the imagemagick dependency is a soft one but all the filtering
stuff is IMHO so basic that it belongs to base, or rather into
I'm fine with that. Do you want to start it?
--Tim
On Tuesday, March 04, 2014 05:32:02 AM Tobias Knopp wrote:
Tim,
a little bit offtopic question but might it make sense to break of the
algorithmic parts of Images.jl and put it into some signal processing
package?
I know that the
There is DSP.jl: http://dspjl.readthedocs.org/en/latest/index.html
On Tue, Mar 4, 2014 at 10:22 AM, Tobias Knopp
tobias.kn...@googlemail.comwrote:
I don't want to give a definate yes to it but will think a little bit how
such a package could look like.
My Cartesian macro foo is currently
We have some standard DSP stuff in https://github.com/JuliaDSP/DSP.jl. Our
idea is to put everything DSP-related that is not application-specific in
there, and then make other application-specific packages depend on it.
--
João Felipe Santos
On Tue, Mar 4, 2014 at 10:22 AM, Tobias Knopp
I was worried about getting a little noise in the result, so I ran a
quick test in Matlab and Julia, and got almost exactly the same error. This
is the Matlab code:
Ts=0.01;
t=-10:Ts:10;
s=sinc(t);
sc=Ts*conv(s,s);
sc=sc(1000:3000);
sum((sc-s).*(sc-s))
ans =
0.3695
So, at least for
Thanks for the hint. This seems to be a good start although the available
functions already seem to be quite specialized.
The ndimage module from scipy
(http://docs.scipy.org/doc/scipy/reference/ndimage.html#module-scipy.ndimage.filters)
goes into a direction what in my mind could cover a
This depends a lot on the input sizes. For full length convolutions, the
fft approach should be more accurate because of less additions. For very
short kernels this does not hold anymore. But in practice these kinds of
errors are mostly negligable.
Am Dienstag, 4. März 2014 16:42:33 UTC+1
That's funny, because in my opinion the functions in scipy.ndimage.filters
are also specialized, as most of the time they seem to be used in image
processing (but also in time-series processing) :)
In DSP.jl we mainly have functions for one-dimensional signal processing,
though, and it would be
11 matches
Mail list logo