Compute the "periodic" FFT of an image.
Compute the "periodic" FFT of an image. This suppresses border effects with minimal impact on filters.
The input image to be transformed with the periodic FFT.
The transformed image in the frequency domain of the (periodic, smooth) components.
Compute the boundary image as part of the task of constructing the smooth component.
Compute the boundary image as part of the task of constructing the smooth component. This should be made into a GPU kernel for speed.
Implements Moisan's periodic FFT. This is modeled on Peter Kovesi's Matlab implementation of the algorithm. See the paper "Periodic plus Smooth Image Decomposition," Journal of Mathematical Imaging and Vision, vol 39:2, pp. 161-179, 2011.
The input is an image that is suitable for the FFT (number of rows and columns are powers of 2). The output is the FFT of the "periodic" component of the input image with border effects suppressed.
NOTE: This could be sped up by writing a GPU kernel for
computeBoundaryImage
.