Python / opencv：使图像像素为最接近的4像素的平均值

Question

I need to reduce the actual resolution of a picture so that each pixel in 4-pixel rectangle is the mean value of those 4 pixels.

ie

p1 p2 p6 p7         a1 a1 a2 a2 
p3 p4 p8 p9  ... -> a1 a1 a2 a2 where a1 = average(p1, p2, p3, p4), a2 = average(p6, p7, p8, p9) etc

Is it possible to efficiently do that using python/opencv not manipulating individual pixels? Does bilinear/nearest neighbor rescaling down and back to the original resolution produce the requested result(not just the visual effect)? The pictures are monochrome.

Answer 1

If you are interested in transparent numpy/indexing based solution try something like that:

I use some dummy data to test the solution:

img = np.array([range(0,4), range(4, 8), range(8, 12), range(12, 16)])

This gives us following 'image':

[[ 0  1  2  3]
 [ 4  5  6  7]
 [ 8  9 10 11]
 [12 13 14 15]]

Then I create four subimages by taking every second element:

subimg1 = img[::2, ::2]
subimg2 = img[1::2, ::2]
subimg3 = img[::2, 1::2]
subimg4 = img[1::2, 1::2]

Now you can average the corresponding samples using numpy:

average = np.average([subimg1, subimg2, subimg3, subimg4], 0)

And create the output:

output = np.zeros(img.shape)
output[::2, ::2] = average
output[1::2, ::2] = average
output[::2, 1::2] = average
output[1::2, 1::2] = average

And the output for given test data is:

[[  2.5   2.5   4.5   4.5]
 [  2.5   2.5   4.5   4.5]
 [ 10.5  10.5  12.5  12.5]
 [ 10.5  10.5  12.5  12.5]]