intrinsic about depth2normal HOT 2 OPEN

The Sobel function in OpenCV only supports a 2D Image, which is not applicable to computing a normal image from a depth map. In my opinion, if the goal is to compute normal image from depth map, the first step is to convert the depth map into a point cloud, given a set of camera parameters. And the second step is computing the vertex normal of the point cloud. Finally, we can use the correspondence between pixel coordinates and 3D points to generate a normal map. The best practice of the above operations in python is using the Open3D library instead of OpenCV.

from depth2normal.

hi, for different camera focal, we have different ways to explain depth map. so why not take the intrinsic in to consideration when generating normals?

I think so.

from depth2normal.