Some stereo matching methods such as Semi-Global Matching (SGM) might provide sparse disparity maps, meaning that some pixels will not have valid disparity values. In those cases, an interpolation of the predicted disparity map might be needed for evaluation against semi-dense ground-truth, as in the KITTI Stereo 2015 benchmark.
Usage example:
import disparity_interpolation
def interpolate_disparity(disparity_map: np.array) -> np.array:
"""Intepolate disparity image to inpaint holes.
The expected run time for a stereo image with 2056 × 2464 pixels is ~50 ms.
"""
# Set the invalid disparity values defined as "0" to -1.
disparity_map[disparity_map == 0] = -1
disparity_map_interp = disparity_interpolation.disparity_interpolator(disparity_map)
return disparity_map_interp