Linear Total Variation Approximate Regularized Nuclear Norm Optimization for Matrix Completion

Figure 1

The recovered results with 60% random mask and word mask for 𝛾=0, 0.5 and 1 by LTVNN.

(a) Original image (256 × 256)

(b) Random masked image

(c) 𝛾=0, PSNR: 8.934

(d) 𝛾=0.5, PSNR: 29.183

(e) 𝛾=1, PSNR: 8.593

(f) Word masked image

(g) 𝛾=0, PSNR: 14.706

(h) 𝛾=0.5, PSNR: 33.421

(i) 𝛾=1, PSNR: 14.759