Is there any advantage of using option 2? Option 2 looks more complicated than option 1, why do we still need option 2?
username
I guess it is because option 1 is in spatial domain, which requires convolution. While option 2 needs multiplication in frequency domain, which is more efficient than convolution.
THWG
Computation in the frequency domain is highly parallelizable, thus better for a modern processor (GPU, ASIC etc)
Is there any advantage of using option 2? Option 2 looks more complicated than option 1, why do we still need option 2?
I guess it is because option 1 is in spatial domain, which requires convolution. While option 2 needs multiplication in frequency domain, which is more efficient than convolution.
Computation in the frequency domain is highly parallelizable, thus better for a modern processor (GPU, ASIC etc)