The depth information recovered by the system may be simply displayed in an intensity encoded form on an estimated cyclopean image. The intensity value for an image point will be inversely proportional to its relative depth from the camera. The performance illustrations show the depth normalised between the strongest edge which has an depth-encoded intensity value of 255 and, the weakest edge which is set to an intensity value of 5 (setting it to intensity 0 wouldn't produce any image output - white on white). Absolute depth recovery requires a strong edge to be identified and its actual depth from the camera, found by triangulation. Knowing this, and the normalisation scale, absolute depth values can be recovered throughout the image.

The final image output is a cyclopean approximation produced by averaging the positions of the left and right primitives in all the true matches. If the system makes any incorrect matches their effect on the final image can be reduced by averaging its disparity value with any true cyclopean match position it coincides with.

As an illustration of the application of stereo vision, a series of images are presented below. These represent respectively a stereo image pair, the thresholded edge data, and the cyclopean depth map.

Some problems of the stereo vision system are immediately apparent. First, the depth map is sparse since edge features are required in both images to produce points of correspondence. Second, the technique fails to extract depth data at points where the boundary feature aligns with line separating the camera geometry. Because of these limitations, there is currently some work in hand to combine stereo viewing with other techniques, notably conventional 2D image segmentation and shape from shading and texture methods.