import numpy as npdef svt(mat, tau):u, s, v = np.linalg.svd(mat, full_matrices = 0)vec = s - tauvec[vec < 0] = 0return np.matmul(np.matmul(u, np.diag(vec)), v)def LRMC(sparse_mat, dense_mat, rho, maxiter):pos_train = np.where(sparse_mat != 0)pos_test = np.where((sparse_mat == 0) & (dense_mat != 0))binary_mat = sparse_mat.copy()binary_mat[pos_train] = 1X = sparse_mat.copy()Z = sparse_mat.copy()T = sparse_mat.copy()rse = np.zeros(maxiter)for it in range(maxiter):Z = svt(X + T / rho, 1 / rho)X = Z - T / rhoX[pos_train] = sparse_mat[pos_train]T = T - rho * (Z - X)rse[it] = (np.linalg.norm(X[pos_test] - dense_mat[pos_test], 2) / np.linalg.norm(dense_mat[pos_test], 2))return X, rse

import imageio
import matplotlib.pyplot as pltlena = imageio.imread('z6.bmp')/255.0
sparse_lena=imageio.imread('z6ss.bmp')/255.0
print('The shape of the image is {}.'.format(lena.shape))dim1, dim2,dim3 = lena.shape
mask = np.round(np.random.rand(dim1, dim2,dim3))  # Generate a binary mask.
mask1 = np.round(np.random.rand(dim1, dim2,dim3))
# mask2 = np.round(np.random.rand(dim1, dim2))plt.figure(figsize=(15,12))
plt.imshow(lena)
plt.title('The original Lena')
plt.axis('off')plt.figure(figsize=(15,12))
plt.imshow(sparse_lena)
plt.title('The incomplete Lena')
plt.axis("off")plt.show()

import timestart = time.time()
rho = 0.005
maxiter = 50
mat_hat, rse_svt = LRMC(sparse_lena[:,:,0], lena[:,:,0], rho, maxiter)
print('Running time: %d seconds.'%(end - start))
start = time.time()
mat_hat1, rse_svt1 = LRMC(sparse_lena[:,:,1], lena[:,:,1], rho, maxiter)
print('Running time: %d seconds.'%(end - start))
start = time.time()
mat_hat2, rse_svt2 = LRMC(sparse_lena[:,:,2], lena[:,:,2], rho, maxiter)
print('Running time: %d seconds.'%(end - start))#修复完把三张图拼接在一起
c=[]
for i in range(dim1):c.append([])for j in range(dim2):c[i].append([mat_hat[i][j],mat_hat1[i][j],mat_hat2[i][j]])

plt.figure(figsize=(20,15))
# plt.imshow(mat_hat)
plt.imshow(c)
# plt.imshow(mat_hat2)
plt.savefig("lbk.png")
plt.axis('off')
plt.show()