Align RGB image to Depth Image using Intrinsic and Extrinsic Matrix

Question

similar questions are solved many times. However, they generally maps depth coordinates to RGB coordinates, by following the next steps:

1. apply the inverse depth intrinsic matrix to the depth coordinates.
2. rotate and translate the 3d coordinates obtained using the rotation R and T matrixes that maps 3d depth coordinates to 3D RGB coordinates.
3. apply the RGB intrinsic matrix to obtain the image coordinates.

However, I want to do the reverse process. From a RGB coordinates obtain the depth coordinates. Then I can obtain an interpolated value from the depth map based on those coordinates.

The problem is that I don't know how can I define the z coordinate in the RGB image to make everything works.

The process should be:

1. obtain 3D RGB coordinates by applying the camera's inverse intrinsic matrix. How can I set the z coordinates? Should I define and estimated value? Set all the z coordinates to one?
2. rotate and translate the 3D RGB coordinates to the 3d coordinates.
3. apply the depth intrinsic matrix.

If this process cannot be done. How can I map RGB coordinates to depth coordinates instead of the other way around?

Thank you!