Depth correction: Methods for approximating depth information in web camera depth maps

Stereo web cameras can be used to obtain depth information about a scene. The maps produced using this procedure are prone to missing depth values, more so than when using a commercial stereo web camera. Post-processing can be used to find approximations for unknown depth values. In this paper we describe four techniques for approximating missing depth values and compare their effectiveness. A large median filter which considered only known pixel values has achieved promising results showing an increase in the proportion of valued pixels and decrease in the number of connected regions. A large median filter and use of a close operator have also both decreased the number of connected regions although they both increased the proportion of unknown pixels.