Local Approximations in Signal and Image Processing
Anisotropic Deblurring results
| Experiment 1 - Cameraman 256x256, blur: 9x9 uniform "box-car" blur, noise: BSNR 40dB | ||
 |  |  |
| Blurred and noisy observation | Regularized inverse | Anisotropic LPA-ICI RWI estimate, ISNR 8.29dB, PSNR 29.06dB |
| Experiment 2 - Cameraman 256x256, blur: 15x15 1/(x²+y²), x,y=-7,...,7, kernel, noise: sigma²=2 | ||
 |  |  |
| Blurred and noisy observation | Regularized inverse | Anisotropic LPA-ICI RWI estimate, ISNR 7.86dB, PSNR 30.09dB |
| Experiment 3 - Cameraman 256x256, blur: 15x15 1/(x²+y²), x,y=-7,...,7, kernel, noise: sigma²=8 | ||
 |  |  |
| Blurred and noisy observation | Regularized inverse | Anisotropic LPA-ICI RWI estimate, ISNR 6.00dB, PSNR 28.17dB |
| Experiment 4 - Lena 512x512, blur: 5x5 separable [1, 4, 6, 4, 1]/16 filter, noise: BSNR 15.93dB | ||
 |  |  |
| Blurred and noisy observation | Regularized inverse | Anisotropic LPA-ICI RWI estimate, ISNR 3.95dB, PSNR 32.44dB |
Tampere University of Technology - Institute of Signal Processing - Transforms and Spectral Methods Group