Total Variation Regularization Algorithms for Images Corrupted with Different Noise Models: A Review

Figure 3

Denoising and deconvolution with the Poisson TV via [108] (a spatially adaptive algorithm).

(a) Grayscale “Cameraman” image corrupted with Poisson noise (𝑀=30). SNR: 5.8 dB., SSIM: 0.31

(b) Blurred grayscale “Cameraman” image corrupted with Poisson noise (𝑀=30). SNR: 5.4 dB., SSIM: 0.24

(c) Blurred color “Peppers” image corrupted with Poisson noise (𝑀=30). SNR: 10.4 dB

(d) Poisson-TV denoising. SNR: 15.26 dB, SSIM: 0.66

(e) Poisson-TV deconvolution. SNR: 14.88 dB, SSIM: 0.61

(f) Poisson-TV deconvolution. SNR: 15.46 dB