Z values way too large compared to x and y in reconstruction

I reconstruct x,y,z from disparity using triangulation formula .My problem is that x,y, and z values are in very different orders .For eg order of x is like 0.001 and similar for y but z is in the order of 10 .Because of this I see a straight line instead of seeing a face .Is there any way I could apply some transform preserving the structure of face but getting a better reconstruction.

EDITED: here is a sample L image and the disparity map ( normalized to 0-255 for visualization not the true values).My point 开发者_开发问答of giving this is to show that disparity comes out fairly decently.

Assuming that you are solving for the fundamental matrix using point correspondences between the left and right images, this is the expected result. Because the fundamental matrix is rank-deficient, it is only defined up to a scale factor. If you define everything in terms of pixel units, there is no way to reconstruct the scene in real-world units.

Solving this requires an additional piece of information: the relationship between a pair of corresponding points in a three-dimensional coordinate frame. For a stereo system, this is most often the baseline, the distance between the left and right camera centers.