OpenCV Disparity map post-filtering using only one depth image

Question

OpenCV has a tutorial on post-filtering disparity maps here but it needs the left and right images to function properly. Is there a way to do this using only one depth image which we have acquired beforehand?

From this:

To This:

Answer 1

From the documentation :

Disparity map filter based on Weighted Least Squares filter (in form of Fast Global Smoother that is a lot faster than traditional Weighted Least Squares filter implementations) and optional use of left-right-consistency-based confidence to refine the results in half-occlusions and uniform areas.

I think you can use the filter method after converting your depth map to a disparity map.

However, I would recommend a bilateral filter to filter/clean the depth map as it is generally used in the literature for depth maps. Here is the OpenCV documentation page associated to the bilateral() function. There is also a CUDA version.

For example, the OpenCV odometry_evaluation uses a bilateral filter.