Transform image to get top-down view

Hi,

I have an image like this:

I need to transform the image in such a way that i could get a top-down view representing circles, triangles and squares. I also have the camera intrinsic and extrinsic parameters. How can i use these parameters to generate a top-down view ? Any help would be greatly appreciated.

Regards !

[Walter Roberson]

Is it single images, or is it stereo pairs that you have?

Answer 1

[Matt J]

If you have camera extrinsic parameters, it means you derived them from some world coordinate system. I'll assume this coordinate system is aligned with the blue mat shown in the image and the "top-down view" you are searching for is down the z-axis of this coordinate system.

Let P=K*R*[I,-C] be the camera projection matrix in this coordinate system, and its decomposition. Then you want to transform the image with the 3x3 homography given by H=K*R.'*inv(K)

This H has the effect of transforming the projection matrix such that H*P is a new projection matrix with R=eye(3), in other words so that the camera axes are now aligned with your world axes.