You can do this using the Computer Vision System Toolbox.
If your cameras are not calibrated, then you can do uncalibrated stereo image rectification, and then use the disparity() function to compute stereo dispartiy, which is proportional to depth. Unfortunately, you cannot easily get depth in world units with an uncalibrated stereo pair, unless you know distances between some points in the world, and you can locate these points in the image.
A better approach is to use calibrated cameras. Please see this http://www.mathworks.com/help/vision/examples/stereo-calibration-and-scene-reconstruction.html of how to calibrate a stereo pair of cameras and reconstruct the scene using the Computer Vision System Toolbox.
