Savitzky-Golay |
Very low signal distortion, short computational time |
SNR gain lower than one order of magnitude, two parameters optimization |
Fourier-Transform |
Very low signal distortion, easy to implement, short computational time, easy to optimize |
SNR gain lower than one order of magnitude |
PCA |
Significant SNR gain and reasonable signal distortion |
Medium difficulty algorithm, time and memory consuming computations |
MNF |
Significant SNR gain and reasonable signal distortion |
Difficult algorithm, hard to implement, time and memory consuming computations |
Wavelets |
Very low signal distortion |
SNR gain around one order of magnitude, time consuming calculation and optimization |
Spatial denoising technique
|
PROS
|
CONS
|
Fourier-Transform |
Good pSNR gain, high SSIM, easy to implement, reasonable computation time, easy to optimize |
Image artifacts |
Mean Filter |
Good SSIM, easy to implement, low computational time |
Mild pSNR gain |
Median Filter |
Good SSIM, easy to implement, low computational time |
Mild pSNR gain |
Gauss Filter |
Good SSIM, easy to implement, low computational time |
Mild pSNR gain |
Weighted Mean Filer |
Good SSIM, easy to implement, low computational time |
Mild pSNR gain |
Wavelets |
Noticeable pSNR gain |
Low pSNR gain and SSIM, time consuming calculation and optimization, image artifacts |
Deep Neural Networks |
Reasonable pSNR and SSIM |
Difficult algorithm to train |