DFT Implementation of FIR filters

• 
• 
• is the circular convolution of x and h. It is a spatially aliased version of .
• If the DFT size is large enough to contain , then

  

  

  

  

• Computation:

  

• The FFT implementation can require considerably less computation, but more storage and I/O than a Direct Implementation.
• Block Convolution implementations are also possible.