摘要 |
PROBLEM TO BE SOLVED: To provide the computerized method for extracting three-dimensional shape information from a sequence of two-dimensional images of an object. SOLUTION: A three-dimensional model is extracted from a sequence of images including a reference image 203. The respective images in the sequence are arrayed with the reference image and have their features determined. The features of the images are used to restore structure and motion parameters 220 by using geometrical restrictions of the style of a wire frame mode. Outward appearances which are predicted for the respective images are generated 215 by using the restored structure and motion parameters. Then the respective predicted outward appearances are arrayed 211 with the corresponding images. The restoration, generation, and arraying stages are repeated until the differences (color or luminance) in mean pixel between the predicted outward appearances and corresponding images become smaller than a specific threshold. |