System and method for three-dimensional estimation based on image data

A method of determining three dimensional information from two dimensional images of a scene, comprising:

• forming, by a computer, a spatio-temporal manifold from two dimensional images of a scene;
• determining, by the computer, characteristics of the spatio-temporal manifold;
• decomposing, by the computer, the spatio-temporal manifold into cells using the characteristics of the spatio-temporal manifold;
• and
• densifying, by the computer, the decomposed spatio-temporal manifold;
• wherein determining characteristics of the spatio-temporal manifold comprises:
• determining segments formed of level lines in the spatio-temporal manifold, and selecting segments with higher meaningfulness;
• and
• wherein selecting segments with higher meaningfulness comprises:
• computing the meaningfulness of each segment,
• ordering the segments by meaningfulness, and
• for each segment, in order of meaningfulness, determining edgels of the segment that form a longest sub-segment of the segment excluding edgels of the segment that have already been determined as edgels of a segment of higher meaningfulness that form a longest sub-segment of the segment of higher meaningfulness.

A method of determining three dimensional information from two dimensional images of a scene, comprising:

• forming, by a computer, a spatio-temporal manifold from two dimensional images of a scene;
• determining, by the computer, characteristics of the spatio-temporal manifold;
• decomposing, by the computer, the spatio-temporal manifold into cells using the characteristics of the spatio-temporal manifold;
• and
• densifying, by the computer, the decomposed spatio-temporal manifold;
• wherein determining characteristics of the spatio-temporal manifold comprises:
• determining segments formed of level lines in the spatio-temporal manifold, and selecting segments with higher meaningfulness;
• and
• wherein each segment is formed of level lines having digital straight segments.

A method of determining three dimensional information from two dimensional images of a scene, comprising:

• forming, by a computer, a spatio-temporal manifold from two dimensional images of a scene;
• determining, by the computer, characteristics of the spatio-temporal manifold;
• decomposing, by the computer, the spatio-temporal manifold into cells using the characteristics of the spatio-temporal manifold;
• and
• densifying, by the computer, the decomposed spatio-temporal manifold;
• wherein determining characteristics of the spatio-temporal manifold comprises:
• determining segments formed of level lines in the spatio-temporal manifold, and selecting segments with higher meaningfulness;
• and
• wherein decomposing the spatio-temporal manifold into cells using the characteristics of the spatio-temporal manifold comprises:
• determining a continuous segment for each selected segment, each continuous segment including endpoints of the corresponding selected segment,
• determining intersections of the continuous segments, and
• modifying lengths of continuous segments that intersect with continuous segments having a greater slope.

A system for forming three dimensional information from two dimensional images of a scene, comprising:

• a camera on a stage;
• a motor for moving the camera;
• a computer configured to provide commands to the motor and to provide commands to the camera, and to receive image information from the camera;
• the computer being programmed to form a spatio-temporal volume from the image information, determine straight segments of level lines of the spatio-temporal volume, extend the straight segments to an intersection with a straight segment with a greater slope, decompose the spatio-temporal volume into cells using the extended straight segments of the spatio-temporal volume, and densify the decomposed spatio-temporal volume.

A system comprising:

• a camera;
• means for translating the camera in a first direction;
• and
• means for determining a spatio-temporal volume from images received from the camera, for determining meaningful straight segments of level lines of the spatio-temporal volume, for extending at least some of the straight segments until occlusion of the at least some of the straight segments, and for densifying the spatio-temporal volume.