Bayesian Image Restoration Using a Large-Scale Total Patch Variation Prior

Figure 1

(a) Original “Lena” image. (b) Degraded “Lena” image (SNR = 6.65) with 9×9 uniform blur and additive white Gaussian noise (variance = 0.1663). (c) Wiener deconvolution. (d) TV prior Bayesian restoration. (e) Nonquadratic Huber prior Bayesian restoration. (f) The ForWaRD restoration in [11]. (g) The SA prior Bayesian restoration proposed in [21]. (h) The proposed LS-TPV prior Bayesian restoration (The left arrows: block-suppressing illustration, the right arrows: edge-preserving illustration).

(a) Original image

(b) Degraded image

(c) Wiener restoration

(d) TV prior restoration

(e) Huber prior restoration

(f) ForWaRD restoration [11]

(g) SA prior restoration [21]

(h) LS-TPV prior restoration