pivuq.disparity

pivuq.disparity.ilk(image_pair, U, window_size=16, prefilter=True, window='gaussian', velocity_upsample_kind='linear', warp_direction='center', warp_order=1, warp_nsteps=1)[source]

Disparity map calculation using iterative Lucas Kanade (“ilk”).

Parameters:

  • image_pair (np.ndarray) – Image pairs \(\mathbf{I} = (I_0, I_1)^{\top}\) of size (2 x rows x cols).

  • U (np.ndarray) – Sparse or dense 2D velocity field \(\mathbf{U} = (u, v)^{\top}\) of (2 x U_rows x U_cols).

  • window_size (int, default: 16) – Window size around the pixel to consider the disparity for optical flow estimator.

  • window ({"gaussian", "tophat"}, default: "gaussian") – Windowing kernel type for integration around the pixel.

  • prefilter (bool, default: True) – Whether to prefilter the estimated optical flow before each image warp. When True, a median filter with window size 3 along each axis is applied. This helps to remove potential outliers.

  • velocity_upsample_kind ({"linear", "cubic", "quintic"}, default: "linear") – Velocity upsampling kind for spline interpolation scipy.interpolation.interp2d.

  • warp_direction ({"forward", "center", "backward"}, default: "center") – Warping direction.

  • warp_order (1-5, default: 1) – The order of interpolation for skimage.transform.warp.

  • warp_nsteps (int, default: 5) – Number of sub-steps to use for warping to improve accuracy.

Returns:

  • X, Y (np.ndarray) – x and y coordinates of disparity map.

  • D (np.ndarray) – pixel-wise 2D disparity map \(\mathbf{D} = (d_x, d_y)^\top\) of size (2 x rows x cols).

See also

skimage.registration.optical_flow_ilk

Coarse to fine optical flow estimator.

skimage.transform.warp

Warp an image according to a given coordinate transformation.

pivuq.disparity.sws(image_pair, U, window_size=16, grid_size=4, window='gaussian', radius=1, sliding_window_subtraction=True, ROI=None, velocity_upsample_kind='linear', warp_direction='center', warp_order=1, warp_nsteps=1)[source]

Python implementation of Sciacchitano-Wieneke-Scarano algorithm of PIV Uncertainty Quantification by image matching [1].

Parameters:

  • image_pair (np.ndarray) – Image pairs \(\mathbf{I} = (I_0, I_1)^{\top}\) of size (2 x rows x cols).

  • U (np.ndarray) – Sparse or dense 2D velocity field \(\mathbf{U} = (u, v)^{\top}\) of (2 x U_rows x U_cols).

  • window_size (int, default: 16) – Window size around the pixel to consider the disparity ensemble.

  • grid_size (int, default: 4) – Disparity ensemble grid resolution in pixels.

  • window ({"gaussian", "tophat"}, default: "gaussian") – Window type for the disparity statistics.

  • radius (int, default: 1) – Search radius for particle peak position.

  • sliding_window_subtraction (bool, default: False) – Whether to use the sliding window subtraction before disparity vector calculation.

  • ROI (tuple, default: None) – Region of interest to use for calculating the disparity ensemble (i_min, i_max, j_min, j_max).

  • velocity_upsample_kind ({"linear", "cubic", "quintic"}, default: "linear") – Velocity upsampling kind for spline interpolation scipy.interpolation.interp2d.

  • warp_direction ({"forward", "center", "backward"}, default: "center") – Warping direction.

  • warp_order (1-5, default: 1) – The order of interpolation for skimage.transform.warp.

  • warp_nsteps (int, default: 5) – Number of sub-steps to use for warping to improve accuracy.

Returns:

  • X, Y (np.ndarray) – x and y coordinates of disparity map.

  • delta (np.ndarray) – Instantaneous error estimation map of size \(2 \times N \times M\) defined by Eq. (3) [1].

  • N (np.ndarray) – Number of peaks inside the window.

  • mu (np.ndarray) – Mean disparity map of size \(2 \times N \times M\) defined by Eq. (3) [1].

  • sigma (np.ndarray) – Standard deviation disparity map of size \(2 \times N \times M\) defined by Eq. (3) [1].

References