linumpy.gpu.fft_ops#

GPU-accelerated FFT operations for linumpy.

Provides GPU versions of FFT-based operations including phase correlation for image registration and stitching.

Functions#

phase_correlation(vol1, vol2[, n_peaks, use_gpu])

GPU-accelerated phase correlation for finding translation between images.

fft2(image[, use_gpu])

GPU-accelerated 2D FFT.

ifft2(spectrum[, use_gpu])

GPU-accelerated 2D inverse FFT.

Module Contents#

linumpy.gpu.fft_ops.phase_correlation(vol1, vol2, n_peaks=8, use_gpu=True)[source]#

GPU-accelerated phase correlation for finding translation between images.

Parameters:

  • vol1 (np.ndarray) – Fixed image (2D or 3D)

  • vol2 (np.ndarray) – Moving image (2D or 3D)

  • n_peaks (int) – Number of peaks to sample for refinement

  • use_gpu (bool) – Whether to use GPU acceleration

Returns:

  • list – Translation [dx, dy] or [dx, dy, dz] of vol2 relative to vol1

  • float – Cross-correlation score

Return type:

Any

linumpy.gpu.fft_ops.fft2(image, use_gpu=True)[source]#

GPU-accelerated 2D FFT.

Parameters:

  • image (np.ndarray) – Input 2D image

  • use_gpu (bool) – Whether to use GPU

Returns:

FFT result (complex)

Return type:

np.ndarray

linumpy.gpu.fft_ops.ifft2(spectrum, use_gpu=True)[source]#

GPU-accelerated 2D inverse FFT.

Parameters:

  • spectrum (np.ndarray) – Input spectrum (complex)

  • use_gpu (bool) – Whether to use GPU

Returns:

Inverse FFT result

Return type:

np.ndarray