Python Image Processing Cookbook

Image inpainting with convex optimization

Follow these steps to pose the inpainting problem as a convex optimization problem and solve it using pyunlocbox library's solvers:

1. Import the required library functions, load the input cat image, and convert it into grayscale:

#!pip install pyunlocboxfrom pyunlocbox import functions, solversim_original = rgb2gray(imread('images/cat.png'))

2. Generate a random masking matrix that masks 80% of the pixels, apply it to the image, and obtain the corrupted image:

np.random.seed(1) # Reproducible results.mask = np.random.uniform(size=im_original.shape)mask = mask > 0.8g = lambda x: mask * xim_corrupted = g(im_original)

3. The prior objective to minimize is defined by f₁(x)=∥x∥_TV. Express it using the toolbox ...