Journal article

A Parallel Stereo Algorithm that Produces Dense Depth Maps and Preserves Image Features

We have developed a stereo algorithm that relies on grey level correlation followed by interpolation using an energy based technique. During the correlation phase the two images play a symmetric role and we use a validity criterion for the matches that eliminates the gross errors: when the images cannot be correlated reliably, due to lack of texture or occlusions for example, the algorithm does not produce wrong matches but a very sparse disparity map as opposed to a dense one when the correlation is successful and we argue that the density of the map is a good estimate of its reliability. To generate dense depth map, the information is then propagated across the featureless areas but not across discontinuities by an interpolation algorithm that takes image grey levels to preserve image features.

Related material