🚀 go-pugleaf

In article <1992Sep22.163256.15935@versyss.com>, keithd@versyss.UUCP (Keith Doyle) writes...
>I'm interested in using FFT for noise reduction of images, and while I
>have worked with FFT in 1D (for audio) and have an assortment of FFT
>source code for 1D, I don't really know how to apply it to the 2D case.
>Is it simply an FFT of all the rows combined with an FFT of all the columns
>of pixels in the image?  

>>    Here is the basic idea: Start with a 2-d array (matrix) of pixels. First,
>>take the 1-d FFT of each row vector, forming an intermdiate matrix from the 
>>result. Now take the 1-d FFT of each column vector in this matrix. The columns
>>of the final output matrix are formed from these transforms. Therefore, the
>>calculation of a 2-d FFT for an NxN matrix requires calculation of N^2 1-D
>>FFT's.
>>
>>
>>                  Hope this is useful info,
>>
>>
>>                                Dave Hall

Dave's description of the 2-D FFT is accurate.  However, the process requires 2N 1-D FFT's NOT N^2.  Taking the FFT of each row requires N 1-D FFT's.  Then, the FFT of the columns requires an additional N 1-D FFT's.  Resulting in 2N 1-D FFT's.

John Arens
johna@comm.mot.com