- Photoeletronic noise: low-light, hight-light image. Use frame averaging. If available, average N frames of same object If noise is independent frame-to-frame, variance will be reduced by \delta/N
- Uniform Noise: X_n[n] = X[n] + U[n], where U[n] is (0,1) or (-A, A). Its mean function equals to zero. Use midpoint filter (max+min /2)
- Gaussian Noise: X_n[n] = X[n] + G[n], where G[n]: {mean 0, var = \delta ^ 2}. Use Linear Smoothers to reduce Gaussian noise. Also you can use midpoint filter
- Salt and Pepper Noise:

X_n[n] = { X[n], (1-p)% of time

= { 1, (p/2)% of time

= { 0, (p/2)% of time

Use median filters to reduce this kind of noise. - Mixed Gaussian: X_n[n] = X[n] + MG[n], where

MG[n] = { (0, \delta ^2 , (1-p)% of time

= { other, (p)% of time - Dust and Scratches: damage associated with photographs. Lines are scratches, and dots are dust. Use median filter or median filter windows to reduce this kind of noise.
- Structured noise: it is periodic noise. Commonly caused by interference between electronic components. Periodicity means noise power is isolated into a few frequencies. Use band-reject filter. Another kind of structured noise is JPEG noise (aperoidic)

## Tuesday, July 29, 2008

### Noise models

Subscribe to:
Post Comments (Atom)

## No comments:

## Post a Comment