pandora.matching_cost.matching_cost

This module contains functions associated to the cost volume measure step.

Module Contents

Classes

AbstractMatchingCost

Abstract Matching Cost class

class pandora.matching_cost.matching_cost.AbstractMatchingCost

Abstract Matching Cost class

__metaclass__

matching_cost_methods_avail: Dict

_subpix

_window_size

cfg

_band

_step_col

_method

_WINDOW_SIZE = 5

_SUBPIX = 1

_BAND

_STEP_COL = 1

schema

margins

classmethod register_subclass(short_name: str, *args)

Allows to register the subclass with its short name

Parameters:

• short_name (string) – the subclass to be registered

• args – allows to register one plugin that contains different methods

desc() → None

Describes the matching cost method :return: None

instantiate_class(**cfg: str | int) → None

Parameters:

cfg (dictionary) – optional configuration, {‘window_size’: int, ‘subpix’: int, ‘band’: str}

Returns:

None

check_conf(**cfg: Dict[str, str | int]) → Dict

Add default values to the dictionary if there are missing elements and check if the dictionary is correct

Parameters:

cfg (dict) – matching cost configuration

Return cfg:

matching cost configuration updated

Return type:

dict

check_band_input_mc(img_left: xarray.Dataset, img_right: xarray.Dataset) → None

Check coherence band parameter between inputs and matching cost step

Parameters:

• img_left (xarray.Dataset) –

  left Dataset image containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

• img_right (xarray.Dataset) –

  right Dataset containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

Returns:

None

abstract compute_cost_volume(img_left: xarray.Dataset, img_right: xarray.Dataset, cost_volume: xarray.Dataset) → xarray.Dataset

Computes the cost volume for a pair of images

Parameters:

• img_left (xarray.Dataset) –

  left Dataset image containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

• img_right (xarray.Dataset) –

  right Dataset image containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

• cost_volume (xr.Dataset) – an empty cost volume

Returns:

the cost volume dataset , with the data variables:

• cost_volume 3D xarray.DataArray (row, col, disp)

Return type:

xarray.Dataset

grid_estimation(img: xarray.Dataset, cfg: Dict[str, dict] | None, disparity_grids: Tuple[numpy.ndarray, numpy.ndarray]) → xarray.Dataset

Estimation of the grid xarray dataset that will store the cost volume.

Parameters:

• img (xarray.Dataset) – left Dataset image

• cfg (dict) – user configuration

• disparity_grids (Tuple[np.ndarray, np.ndarray]) – Tuple of disparity grids min and max

Returns:

a grid with coordinates and a attributes list with:

• indexs of columns to compute

• size

• sampling interval

Return type:

xarray.Dataset

allocate_cost_volume(image: xarray.Dataset, disparity_grids: Tuple[numpy.ndarray, numpy.ndarray], cfg: Dict = None) → xarray.Dataset

Create a cost_volume dataset.

Parameters:

• image (xr.Dataset) – Image to compute cost volume from

• cfg (Dict) – user configuration

• disparity_grids (Tuple[np.ndarray, np.ndarray]) – Tuple of disparity grids min and max

Returns:

a empty grid

Return type:

xr.Dataset

static get_disparity_range(disparity_min: int, disparity_max: int, subpix: int) → numpy.ndarray

Build disparity range and return it.

Parameters:

• disparity_min (int) – minimum disparity

• disparity_max (int) – maximum disparity

• subpix – subpixel precision = (1 or 2 or 4)

Returns:

disparity range

Return type:

np.ndarray

point_interval(img_left: xarray.Dataset, img_right: xarray.Dataset, disp: float) → Tuple[Tuple[int, int], Tuple[int, int]]

Computes the range of points over which the similarity measure will be applied

Parameters:

• img_left (xarray.Dataset) –

  left Dataset image containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

• img_right (xarray.Dataset) –

  right Dataset image containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

• disp (float) – current disparity

Returns:

the range of the left and right image over which the similarity measure will be applied

Return type:

tuple

static masks_dilatation(img_left: xarray.Dataset, img_right: xarray.Dataset, window_size: int, subp: int) → Tuple[xarray.DataArray, List[xarray.DataArray]]

Return the left and right mask with the convention :

• Invalid pixels are nan

• No_data pixels are nan

• Valid pixels are 0

Apply dilation on no_data : if a pixel contains a no_data in its aggregation window, then the central pixel becomes no_data

Parameters:

• img_left (xarray.Dataset) –

  left Dataset image containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

• img_right (xarray.Dataset) –

  right Dataset image containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

• window_size (int) – window size of the measure

• subp (int) – subpixel precision = (1 or 2 or 4)

Returns:

the left mask and the right masks:

• left mask : xarray.DataArray msk 2D(row, col)

• right mask : xarray.DataArray msk 2D(row, col)

• right mask shifted : xarray.DataArray msk 2D(row, shifted col by 0.5)

Return type:

tuple (left mask, list[right mask, right mask shifted by 0.5])

static get_min_max_from_grid(disp_min: numpy.ndarray, disp_max: numpy.ndarray) → Tuple[int, int]

Find the smallest disparity present in disp_min, and the highest disparity present in disp_max

Parameters:

• disp_min (np.ndarray) – minimum disparity

• disp_max (np.ndarray) – maximum disparity

Returns:

dmin_min: the smallest disparity in disp_min, dmax_max: the highest disparity in disp_max

Return type:

Tuple(int, int)

find_nearest_multiple_of_step(value: int) → int

In case value is not a multiple of step, find nearest greater value for which it is the case.

Parameters:

value – Initial value.

Type:

value: int

Returns:

nearest multiple of step.

Return type:

int

cv_masked(img_left: xarray.Dataset, img_right: xarray.Dataset, cost_volume: xarray.Dataset, disp_min: numpy.ndarray, disp_max: numpy.ndarray) → None

Masks the cost volume :

• costs which are not inside their disparity range, are masked with a nan value

• costs of invalid_pixels (invalidated by the input image mask), are masked with a nan value

• costs of no_data pixels, are masked with a nan value. If a valid pixel contains a no_data in its

  aggregation window, then the cost of the central pixel is masked with a nan value

Parameters:

• img_left (xarray.Dataset) –

  left Dataset image containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

• img_right (xarray.Dataset) –

  right Dataset image containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

• cost_volume (xarray.Dataset) –

  the cost_volume DataSet with the data variables:

  • cost_volume 3D xarray.DataArray (row, col, disp)

• disp_min (np.ndarray) – minimum disparity

• disp_max (np.ndarray) – maximum disparity

Returns:

None

allocate_numpy_cost_volume(img_left: xarray.Dataset, disparity_range: numpy.ndarray | List) → numpy.ndarray

Allocate the numpy cost volume cv = (disp, col, row), for efficient memory management

Parameters:

• img_left (xarray.Dataset) –

  left Dataset image containing :

  • im: 2D (row, col) or 3D (band_im, row, col) xarray.DataArray float32

  • disparity (optional): 3D (disp, row, col) xarray.DataArray float32

  • msk (optional): 2D (row, col) xarray.DataArray int16

  • classif (optional): 3D (band_classif, row, col) xarray.DataArray int16

  • segm (optional): 2D (row, col) xarray.DataArray int16

• disparity_range (np.ndarray) – disparity range

• offset_row_col (int) – offset in row and col

Returns:

the cost volume dataset , with the data variables:

• cost_volume 3D xarray.DataArray (row, col, disp)

Return type:

xarray.Dataset

static crop_cost_volume(cost_volume: numpy.ndarray, offset: int = 0) → numpy.ndarray

Return a cropped view of cost_volume.

If offset, do not consider border position for cost computation. :param cost_volume: cost volume to crop :type cost_volume: np.ndarray :param offset: offset used to crop cost volume :type offset: int :return: cropped view of cost_volume. :rtype: np.ndarray