TEM and FESEM are widely used charachterization techniques in the field of material science and we often get noisy dataset while studying the physical behaviour using image analysis and I need to know that which filters other that gaussian blur.
For highly noisy data I recommend using a median filter. In Matlab medfilt2([3 3]), where [3 3] defines the domain in which the denoising is performed. You can apply this filter in iteration ode, or use a larger domain, but you need to be careful to not affect your data "too much" because the filter is passive and the operator needs to decide when the image is sufficiently clean.
In SEM and in many cases in TEM you can do image averaging or increase the integration time to get the less noise. Increasing beam current is also one of possible solutions but may decrease the resolution. I would first try to talk to operators if they can provide better images and only if not try to use the filters.
With a suitable filter, you can eliminate Gaussian noise in your image.
For more information filter the content:DOL:10.13140/RG.2.2.30211.22564/1
(Gaussian noise removal with proper filter)
See.
In most cases noisy image is a fault of an operator. Proper acquisition gives (almost) nose free image. Noise acceptable only for imaging at highest magnifications or for drifting (beam sensitive) specimens. As was suggested by Grzegorz Cios you may want to talk to operator.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling