Depth measuring method and depth measuring apparatus

1. A depth measuring apparatus comprising:

an attribute-based searching section for acquiring a corresponding point in a second image, based on an attribute of a point of interest in a first image;

a depth computing section for computing a depth value of the point of interest for which the corresponding point is acquired, by applying the principle of triangulation;

a depth interpolating section for computing the depth value of the point of interest for which a corresponding point is not acquired, by interpolation from the depth values of a plurality of points of interest for each of which the depth value is determined;

a provisional corresponding point acquiring section for determining a provisional corresponding point for the point of interest for which a corresponding point is not acquired, from the computed depth value;

a corresponding point re-computing section for executing region-based search limited to the neighborhood of the provisional corresponding point so as to acquire a more accurate corresponding point than the provisional corresponding point, wherein

said depth computing section computes the depth value for the point of interest for which the more accurate corresponding point is acquired by said corresponding point re-computing section.

2. The depth measuring apparatus according to claim 1 , wherein the point of interest is an edge point and the attribute is a feature of an image region on both sides of the edge.

3. The depth measuring apparatus according to claim 2 , wherein the feature of the image region on both sides of the edge is color of the image region.

4. The depth measuring apparatus according to claim 2 , wherein the feature of the image region on both sides of the edge is texture of the image region.

5. The depth measuring apparatus according to claim 1 , wherein said depth interpolating section computes the depth value, based on spatial relationship between the plurality of points of interest for each of which the depth value is determined and the point of interest for which a corresponding point is not acquired.

6. The depth measuring apparatus according to claim 1 , wherein said corresponding point re-computing section executes region-based search on an epipolar line, in the second image, of the point of interest for which a corresponding point is not acquired, the search on the epipolar line being limited to the neighborhood of the provisional corresponding point.

7. The depth measuring apparatus according to claim 1 , wherein said attribute-based searching section detects the point of interest in the first image, by referring to a patterned light projected onto a subject.

8. A depth measuring method comprising the steps of:

acquiring a corresponding point in a second image, based on an attribute of a point of interest in a first image;

computing a depth value of the point of interest for which the corresponding point is acquired, by applying the principle of triangulation;

determining the depth value of the point of interest for which a corresponding point is not acquired, by interpolation from the depth values of a plurality of points of interest for each of which the depth value is determined;

determining a provisional corresponding point for the point of interest for which a corresponding point is not acquired, from the computed depth value;

executing region-based search limited to the neighborhood of the provisional corresponding point so as to acquire a more accurate corresponding point than the provisional corresponding point, and

computing the depth value for the point of interest for which the more accurate corresponding point is acquired by applying the principle of triangulation.

9. The depth measuring method according to claim 8 , further comprising the pre-processing steps of:

projecting a patterned light onto a subject; and

detecting the point of interest in the first image by referring to the patterned light projected onto the subject.

10. The depth measuring method according to claim 8 , wherein the attribute is a feature of an image region on both sides of the point of interest.

11. The depth measuring method according to claim 8 , wherein interpolation from the depth values of the plurality of points of interest for each of which the depth value is determined is executed based on spatial relationship between the plurality of points of interest for each of which the depth value is determined and the point of interest for which a corresponding point is not acquired.

12. The depth measuring method according to claim 8 , wherein region-based search is executed on an epipolar line, in the second image, of the point of interest for which a corresponding point is not acquired, the search being limited to a portion of the epipolar line in the neighborhood of the provisional corresponding point.

13. A depth measuring method comprising the steps of:

detecting an edge in a first image;

acquiring, based on an attribute of an edge point on the detected edge, a corresponding point in a second image corresponding to the edge point;

computing a depth value of the edge point for which the corresponding point is acquired, by applying the principle of triangulation;

determining the depth value of the edge point for which a corresponding point is not acquired, by interpolation from the depth values of a plurality of edge points for each of which the depth value is determined;

determining a provisional corresponding point for the edge point for which a corresponding point is not acquired, from the computed depth value;

executing region-based correlation search on an epipolar line, in the second image, of the edge point for which a corresponding point is not acquired, so as to acquire a more accurate corresponding point than the provisional corresponding point, the search being limited to a portion of the epipolar line in the neighborhood of the provisional corresponding point;

computing the depth value for the edge point for which the more accurate corresponding point is acquired, by applying the principle of triangulation; and

determining the depth values of points other than the edge points by interpolation from the depth values of the edge points.

14. The depth measuring method according to claim 13 , wherein said step of detecting the edge in the first image further comprises the step of:

projecting a patterned light onto a subject, wherein

the edge is detected by referring to the patterned light projected onto the subject.

15. The depth measuring method according to claim 13 , wherein the attribute is a feature of an image region on both sides of the edge.

16. The depth measuring method according to claim 13 , wherein interpolation from the depth values of the plurality of edge points for each of which the depth value is determined is executed based on spatial relationship between the plurality of edge points for each of which the depth value is determined and the edge point for which a corresponding point is not acquired.

17. The depth measuring method according to claim 13 , wherein the region-based correlation search is executed on a subpixel level.

18. A computer program causing a computer to execute each of the steps of claim 8 .

19. A computer program causing a computer to execute each of the steps of claim 13 .