Catalogue of Artificial Intelligence Techniques

   

Jump to: Top | Entry | References | Comments

View Maths as: Images | MathML

Fourier Transform

Keywords: frequency domain, linear systems theory

Categories: Pattern Recognition and Image Processing


Author(s): Fritz Seytter

A method that transforms a (spatial or temporal) signal into the frequency domain by exploiting the fact that a signal can be represented as the sum (or integral) of a (usually infinite) series of weighted sinusoidal functions. The Fourier transform is useful for analysing and understanding the influence of linear operations on temporal or spatial signals. A good qualitative understanding about how a certain operation influences the signal in the frequency domain often helps to estimate limitations or possibilities that are otherwise not so easily recognised. This is most important for low-level vision applications. Fourier transforms are the most important tool for linear systems theory. The Fourier transform is quite tedious computationally: i.e., the computational complexity of most one-dimensional transforms is O(Nlog(N)) and two-dimensional transforms is O(N2log(N)) where typically N>250 . Hence, it is usually impractical to implement the transform in software, although multi-processor systems are now being used. Alternatives include use of special-purpose hardware or optical techniques.


References:


Comments:

Add Comment

No comments.