Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
16 years ago

Implement Phase Correlation algorithm

I'm putting a Phase Correlation algorithm in FPGA. Two images in, window them with 2D Hamming. 2D FFT on each. Complex multiply the matrices with one being conjugated first. The magnitude taken of ...
Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
16 years ago

As your final goal is finding a peak on FFT's product square-magnitude.

e.g., FFT data output could be scaled down by discarding 16 least significant bits (height*width), so your result is 22 bits, further, complex product is 44 bits, again it could be scaled down to 23 bits (bit growth of addition) by using a rounder circuit, same could be done for square magnitude calculation. You must run several simulations in order to determine the number of bits to discard.

Division operation can be replaced by a multiplier if a search of a peak (p=G(xm,ym) ) that satisfies

p>=threshold_percent*G(0,0)

is performed.

Where threshold_percent could be 0.8, 0.9, it depends on SNR.