Python Image Processing Cookbook

How to do it...

Follow the steps listed to implement the Wiener filter and apply it to restore images corrupted with blur/noise:

1. Read the image and convert it into grayscale:

im = color.rgb2gray(imread('images/cactus.png'))

2. Define convolve2d() to implement the convolution in the frequency domain (by the convolution theorem, this implementation works much faster than the one in the spatial domain). Construct the 7x7 average (box-blur) point-spread function and use it as convolution-kernel to blur the image, with the following code:

def convolve2d(im, psf, k):
    M, N = im.shape
    freq = fp.fft2(im)
    # assumption: min(M,N) > k > 0, k odd    psf = np.pad(psf, (((M-k)//2,(M-k)//2+1), \                      ((N-k)//2,(N-k)//2+1)), mode='constant') 
    freq_kernel = fp.fft2 ...