Pandora

Added

- Multi wheel cibuildwheel in github CI for release
- get disp min and disp max in datasets

Changed

- Numba function refactored to C++ functions.
- Census refactored to c++.
- Lower and upper bounds from risk added.
- Clean and simplify README

Fixed

- Confidence measure robust to zncc matching cost measure.
- properly manage multiband inputs in census

Added

- Using the numba cache. [413]
- Modification of check_conf to compare the disparity interval with the image size. [411]
- Addition of a spline order parameter to the shift_right_img method. [420]
- Add possibility to not use numba cache. [421]

Changed

- Update criteria documentation. [410]
- Changing the after callback to before in the state machine. [407]
- Updating the filtering.rst, a documentation file. [414]
- Remove numpy warning. [408]

Fixed

- Solving problems displaying images in notebooks. [409]
- Move the _indicator parameter to remove the sphinx warning. [416]
- Risk normalisation by external global interval. [418]
- Use of ‘max’ measurement type with ambiguity. [404]

Changed

- Fix numpy version.

Fixed

- Fixed calculation of ambiguity and risk with variable disparity. [400]

Changed

- Ambiguity normalisation by external global interval. [402]

Added

- Add criteria documentation in Exploring the Field section. [380]
- Add a new method to add disparity grid in xarray dataset. [389]
- Add get method to margins. [394]
- Add interval_indicator on filetring documentation.

Fixed

- Fix link to notebooks in the documentation. [310]
- Update cv_masked to be used with a column step. [391]
- Fix step in get_coordinates method.
- Fix disparities bigger than images shape. [396]
- Update notebooks after evolution of pandora run_prepare method. [395]

Changed

- Uniformisation tests for cv_masked method. [372]
- Extract col_to_compute from grid_estimation method. [393]
- Update pandora after evolution of use_confidence in libsgm plugin. [397]