Content-Aware Compressive Sensing Recovery Using Laplacian Scale Mixture Priors and Side Information

Algorithm 1

CS image recovery via SI-LSM-AMP.

Input: , , , , .

for    to    do

(a) Approximate the Onsager correction term via MC.

(b) Update the residual .

(c) Obtain the noisy image .

(d) Calculate the proximal operator (i.e., solve (25))

for    to    do

(I) Construct the low-rank matrix .

(II) Distinguish irregular structures from regular structures with the similarity measure , and set

and via (26).

(III) Perform the SVD on to get the singular value vector .

(IV) Estimate the expectations of scale parameters and via (16), and the noise variance

.

(V) Compute the global optimums of coefficients and via (22) and (23).

(VI) If , recover the whole image by aggregating all recovered pixels.

end for

end for