SI-FPGAIP-PHASE: Overview
Sheldon Instruments offers a Phase Correlation IP that is efficiently designed with a highly parallel and pipelined architecture, enabling a throughput of several thousands of frames per second making it ideal for image recognition and target tracking applications.

Phase Correlation is the same as cross correlation, but with a different computation path. Instead of performing a convolution of two time domain signals (that may represent images) to obtain the cross correlation, with phase correlation the two signals are first transformed into the frequency domain via an FFT, and then multiplied together element wise. From this multiplication, the resultant magnitude is normalized, which is finally reverted back to the time domain by applying an inverse FFT.

The result is a peak (or peaks if there are multiple instances of the template in the search signal) at the location where the two signals correlate. This peak indicates the translative offset between the two images. The advantages of using a Phase Correlation over a Cross Correlation is the significantly faster processing times, and the results are less susceptible to certain types of noise, occlusions, and other defects typical of medical or satellite images.

Key Features

• Highly integrated into the FFT architectures, making use of the same FPGA resources (memories and complex multipliers) to perform the multiplication stage of the algorithm.
• 1D and 2D dimensionality.
• Flexibility to handle binarily expandable data sizes.