The concept of noise is very intuitive in communication systems. When signals are transmitted through a channel, various sources such as thermal noise, shot noise corrupt the data. The noise is usually modelled as additive white noise with Gaussian statistics.
I want to know what noise sources, if any, originate in the field of image processing. What do we mean when we say (if we do say) that an image is buried in noise? What are the sources?
A related concept is the distortion. Technically both distortion and noise modify the data. However, we can always compensate for the distortion (meaning it is not random..but deterministic) but cannot do so for noise. Blurring of images comes to my mind as source of distortion. Why does blurring occur?
An interesting point regarding 2D z-transforms. In 1D, a polynomial of degree N can always be factored in terms of its zeros, like
f(z) = A(z-z0)(z-z1)(z-z2).....(z-zN).
Surprisingly we can't do this for 2D functions such as f(u,v), where u and v are in general complex. This is supposed to be linked to concept of vector space projections. Any idea what this means ?
Subscribe to:
Post Comments (Atom)
1 comment:
I was studying the book, Fourier analysis and imaging by Bracewell (yeah, I am a newbie to image processing:)) .. In his last chapter he explains what noise means for an image.. Now I've some more knowledge in this :).... BTW, it's a very good book to understand the role of FT in image processing.
A question though..with the use of wavelets increasing..do people still use FT to analyse images ?
Post a Comment