CODEWITHSUNDEEP
pythonimage-processingnumpypython-imaging-library
Clippit
Clippit

Reputation: 906

How to improve efficiency in this numpy iterating?

I'm working on an assignment about converting a grayscale image to 1-bit binary image by dithering. I'm trying a simple 4x4 matrix that will make the image 16 times larger than original.

dithering_matrix = array([[ 0,  8,  2, 10],
                          [12,  4, 14,  6],
                          [ 3, 11,  1,  9],
                          [15,  7, 13,  5]], dtype=uint8)
split_num = dithering_matrix.size + 1

I read a 512x512 image to im ndarray and did following things:

output = list()
for row in im:
    row_output = list()
    for pixel in row:
        pixel_matrix = ((pixel / (256 / split_num)) > dithering_matrix) * 255
        row_output.append(pixel_matrix)
    output.append( hstack( tuple(row_output) ) )
output_matrix = vstack( tuple(output) )

I found it took 8-10s to output and I think the loop of im above spent much time. In some software the same operation was usually done in a flash. So is it possible to improve the efficiency?

UPDATE: @Ignacio Vazquez-Abrams I'm not vert fimiliar with profiler:( I tried cProfile and the result is strange.

         1852971 function calls (1852778 primitive calls) in 9.127 seconds

   Ordered by: internal time
   List reduced from 561 to 20 due to restriction <20>

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1    6.404    6.404    9.128    9.128 a1.1.py:10(<module>)
      513    0.778    0.002    0.778    0.002 {numpy.core.multiarray.concatenate
}
   262144    0.616    0.000    1.243    0.000 D:\Python27\lib\site-packages\nump
y\core\shape_base.py:6(atleast_1d)
   262696    0.260    0.000    0.261    0.000 {numpy.core.multiarray.array}
   262656    0.228    0.000    0.487    0.000 D:\Python27\lib\site-packages\nump
y\core\numeric.py:237(asanyarray)
      515    0.174    0.000    1.419    0.003 {map}
   527019    0.145    0.000    0.145    0.000 {method 'append' of 'list' objects
}

The line 10 of a1.1.py is the first line from numpy import * (all comments before that) which really puzzles me up.

Upvotes: 2

Views: 572

Answers (1)

NPE
NPE

Reputation: 500663

If you use the Kronecker product to turn every pixel into a 4x4 submatrix, that'll enable you to get rid of the Python loops:

im2 = np.kron(im, np.ones((4,4)))
dm2 = np.tile(dithering_matrix,(512,512))
out2 = ((im2 / (256 / split_num)) > dm2) * 255

On my machine this is roughly 20x faster than your version.

Upvotes: 8

Related Questions