comparing hardware architecture

Currently, I have two algorithm to consider.

one using FFT, one using DCT the DCT is using NEDA for FFT, I am not quite sure which one to use

Do you guys have any suggestions for FFT and DCT regarding implementation ? I am now reading on NEDA and Radix.

Besides, how would you guys implement DoG (Difference of Gaussian) in hardware ?

Thanks !

Do you have any requirements for these functions other than "they need to be implemented"? What data throughput do you require? What data representation do you need use? What type of device will it be implemented on?

You are considering algorithms which is fine, but the choice of algorithm is often impacted by the requirements.