2D circular convolution (part 2)
Two 2D sequences, one is 3 x 4 points & the other is 4 x 3 points in extent, are circularly convolved using (6 x 6)-point 2-D DFTs (Discrete Fourier Transforms). Which samples of the (6 x 6)-output array are identical to the samples of the linear convolution of the two input arrays & which are different??
......see http://mathworld.wolfram.com/ConvolutionTheorem.html at bottom of page eqn 7, it applies in discrete domain as well.
ie f**g = F^(-1)( F(f)F(g) )
ie f**g = F^(-1)( F(f)F(g) )
ASKER
>> are circularly convolved using (6 x 6)-point 2-D DFTs
I mean by that zeros are padded to each input (A & B) before making the circular convolution (i.e. before taking the DFTs of A & B and multiplying them, and then taking the IDFT of the product)
>> then the outputs are the same as A**B for all inputs A,B because of the convolution theorem
Yes you are right (I checked it out), for any inputs A & B of the specifications above, the circular convolution & the linear convolution are the same (if the DFT is applied on a 6x6 basis - after padding zeros). However, 2D circular convolution (part 1) wouldn't have the same answer........
I mean by that zeros are padded to each input (A & B) before making the circular convolution (i.e. before taking the DFTs of A & B and multiplying them, and then taking the IDFT of the product)
>> then the outputs are the same as A**B for all inputs A,B because of the convolution theorem
Yes you are right (I checked it out), for any inputs A & B of the specifications above, the circular convolution & the linear convolution are the same (if the DFT is applied on a 6x6 basis - after padding zeros). However, 2D circular convolution (part 1) wouldn't have the same answer........
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
No sorry....this time it's different....please do read the question, these are two completely different questions!!!!!
Each have a separate answer, and could have been answered by a different expert......
Each have a separate answer, and could have been answered by a different expert......
"are circularly convolved using (6 x 6)-point 2-D DFTs (Discrete Fourier Transforms)"
do you mean convolving A and B but taking thier DFT multiplying and then doing an inverse DFT ? If you you do then the outputs are the same as A**B for all inputs A,B because of the convolution theorem.