Adaptive frequency domain filtering for phase plane correlation

摘要

In a Phase Plane Correlation (PPC) process, using adaptive frequency domain filtering to aid in generating candidate motion vectors. It is determined when it is beneficial to pre-filter an input image, prior to a PPC process. This results in more reliable and consistent PPC surfaces than otherwise. The filter is applied in the frequency domain where time-domain convolution becomes a much more efficient component-wise multiplication with an in-place window. An energy measure of the high-frequency content in the computed Fourier surfaces gauges the degree of high frequency content in the image. First, the Fourier transform of the two images is computed. Then, the high-frequency content is estimated from the Fourier surfaces. A window function is computed as a function of the high-frequency energy. The window is applied to the Fourier surfaces. Then, the modified Fourier surfaces are fed into the PPC process.