pandora.refinement.refinement

This module contains classes and functions associated to the subpixel refinement step.

Module Contents

Classes

AbstractRefinement

Abstract Refinement class

class pandora.refinement.refinement.AbstractRefinement

Abstract Refinement class

__metaclass__

subpixel_methods_avail: Dict

_refinement_method_name

cfg

margins

subpixel_refinement(cv: xarray.Dataset, disp: xarray.Dataset) → None

Subpixel refinement of disparities and costs.

Parameters:

• cv (xarray.Dataset) –

  the cost volume dataset with the data variables:

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

  • confidence_measure 3D xarray.DataArray (row, col, indicator)

• disp (xarray.Dataset) –

  Dataset with the variables :

  • disparity_map 2D xarray.DataArray (row, col)

  • confidence_measure 3D xarray.DataArray (row, col, indicator)

  • validity_mask 2D xarray.DataArray (row, col)

Returns:

None

approximate_subpixel_refinement(cv_left: xarray.Dataset, disp_right: xarray.Dataset) → xarray.Dataset

Subpixel refinement of the right disparities map, which was created with the approximate method : a diagonal search for the minimum on the left cost volume

Parameters:

• cv_left (xarray.Dataset) –

  the left cost volume dataset with the data variables:

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

  • confidence_measure 3D xarray.DataArray (row, col, indicator)

• disp_right (xarray.Dataset) –

  right disparity map with the variables :

  • disparity_map 2D xarray.DataArray (row, col)

  • confidence_measure 3D xarray.DataArray (row, col, indicator)

  • validity_mask 2D xarray.DataArray (row, col)

Returns:

disp_right Dataset with the variables :

• disparity_map 2D xarray.DataArray (row, col) that contains the refined disparities

• confidence_measure 3D xarray.DataArray (row, col, indicator) (unchanged)

• validity_mask 2D xarray.DataArray (row, col) with the value of bit 3 ( Information: calculations stopped at the pixel step, sub-pixel interpolation did not succeed )

• interpolated_coeff 2D xarray.DataArray (row, col) that contains the refined cost

Return type:

xarray.Dataset

classmethod register_subclass(short_name: str)

Allows to register the subclass with its short name

Parameters:

short_name (string) – the subclass to be registered

abstract desc() → None

Describes the subpixel method :return: None

static loop_refinement(cv: numpy.ndarray, disp: numpy.ndarray, mask: numpy.ndarray, d_min: int, d_max: int, subpixel: int, measure: str, method: Callable[[numpy.ndarray, numpy.ndarray, numpy.ndarray, numpy.ndarray, str], Tuple[int, int, int]]) → Tuple[numpy.ndarray, numpy.ndarray, numpy.ndarray]

Apply for each pixels the refinement method

Parameters:

• cv (3D numpy array (row, col, disp)) – cost volume to refine

• disp (2D numpy array (row, col)) – disparity map

• mask (2D numpy array (row, col)) – validity mask

• d_min (int) – minimal disparity

• d_max (int) – maximal disparity

• subpixel (int ( 1 | 2 | 4 )) – subpixel precision used to create the cost volume

• measure – the measure used to create the cot volume

• measure – string

• method – the refinement method

• method – function

Returns:

the refine coefficient, the refine disparity map, and the validity mask

Return type:

tuple(2D numpy array (row, col), 2D numpy array (row, col), 2D numpy array (row, col))

abstract static refinement_method(cost: numpy.ndarray, disp: float, measure: str) → Tuple[float, float, int]

Return the subpixel disparity and cost

Parameters:

• cost (1D numpy array : [cost[disp -1], cost[disp], cost[disp + 1]]) – cost of the values disp - 1, disp, disp + 1

• disp (float) – the disparity

• measure (string = min | max) – the type of measure used to create the cost volume

Returns:

the refined disparity (disp + (sub_disp/subpix)), the refined cost and the state of the pixel ( Information: calculations stopped at the pixel step, sub-pixel interpolation did not succeed )

Return type:

float, float, int

static loop_approximate_refinement(cv: numpy.ndarray, disp: numpy.ndarray, mask: numpy.ndarray, d_min: int, d_max: int, subpixel: int, measure: str, method: Callable[[numpy.ndarray, numpy.ndarray, numpy.ndarray, numpy.ndarray, str], Tuple[int, int, int]]) → Tuple[numpy.ndarray, numpy.ndarray, numpy.ndarray]

Apply for each pixels the refinement method on the right disparity map which was created with the approximate

method : a diagonal search for the minimum on the left cost volume

Parameters:

• cv (3D numpy array (row, col, disp)) – the left cost volume

• disp (2D numpy array (row, col)) – right disparity map

• mask (2D numpy array (row, col)) – right validity mask

• d_min (int) – minimal disparity

• d_max (int) – maximal disparity

• subpixel (int ( 1 | 2 | 4 )) – subpixel precision used to create the cost volume

• measure (string = min | max) – the type of measure used to create the cost volume

• method (function) – the refinement method

Returns:

the refine coefficient, the refine disparity map, and the validity mask

Return type:

tuple(2D numpy array (row, col), 2D numpy array (row, col), 2D numpy array (row, col))