Problem incorporating a Python image filter in Matlab

Hi,

I'm trying to run a python median filter in Matlab (because Matlab does not have a good tool to omit nan's in large window median filters).

Below is the code I'm running as a test. It runs but the output from the python medianfilter is incorrect (if you run the code and look at the plots you'll notice). I suppose the problem might be in the way the inital array needs to be turned into a vector and reshaped back to its original shape again. This needs to be done because numpy.array won't convert a matlab matrix directly. But I am unable to pick the exact problem here. Maybe someone can help?

% load modules
py.importlib.import_module('silx');
py.importlib.import_module('numpy');

I = imread('ngc6543a.jpg');    % load image
I = I(:,:,3);                  % keep only one layer of the image as test
Im = medfilt2(I,[20,20]);      % matlab median filter

I = double(I);                % convert to double top np-array
npI = py.numpy.array(I(:).'); % convert to np-array --> vector
Ishape = py.tuple({uint16(size(I,1)), uint16(size(I,2))});   % tuple of original matrix shape
npI = py.numpy.reshape(npI, Ishape,'C');                     % reshape the matrix back to original 
pyIm = py.silx.math.medianfilter.medfilt(npI,uint8(21),0);   % apply median filter
pyIm = double(pyIm);          % convert back to matlab array

% plot and compare results
subplot(2,2,1)
imagesc(I)
title('original image')
subplot(2,2,2)
imagesc(Im)
title('matlab filter')
subplot(2,2,3)
imagesc(pyIm)
title('python filter')

Answer 1

[Kojiro Saito]

I'm not sure about "numpy.array won't convert a matlab matrix directly", but py.numpy.array(I) works fine, so I don't think it's need to create a vector and reshape.

The following modified code will work fine.

% load modules
py.importlib.import_module('silx');
py.importlib.import_module('numpy');
I = imread('ngc6543a.jpg');    % load image
I = I(:,:,3);                  % keep only one layer of the image as test
Im = medfilt2(I,[20,20]);      % matlab median filter
I = double(I);                % convert to double top np-array
npI = py.numpy.array(I); % convert to np-array
npI2 = npI.copy(); % copy np-array to avoid "must be a C_CONTIGUOUS numpy array" error
pyIm = py.silx.math.medianfilter.medfilt(npI2,uint8(21),0);   % apply median filter
pyIm = double(pyIm);          % convert back to matlab array
% plot and compare results
subplot(2,2,1)
imagesc(I)
title('original image')
subplot(2,2,2)
imagesc(Im)
title('matlab filter')
subplot(2,2,3)
imagesc(pyIm)
title('python filter')

[Richard Ott]

I think I was just having Python version compabilit issues. That I resovled now.