Method and apparatus for feature-based stereo matching

1. A method for stereo matching of a reference image and at least one comparative image captured from different points of view using a computer device, comprising:

collecting the reference image and the at least one comparative image;

extracting feature points from the reference image;

tracking points corresponding to the feature points in the at least one comparative image using an optical flow technique; and

generating a depth map according to correspondence-point tracking results,

wherein the extracting of the feature points includes: extracting points within a fixed ranking by calculating unique values of a gradient matrix G in positions of the reference image and arranging points in descending order for a smaller unique value between two unique values produced as calculations results.

2. The method of claim 1 , wherein an approximate correspondence point is found through a number of iterations by applying a Newton-Raphson method to the optical flow technique.

3. The method of claim 1 , wherein:

the tracking of the points corresponding to the feature points includes configuring a feature-point tracking window for tracking points corresponding to the feature points in the reference image, and

a size, a shape, and a center position of the feature-point tracking window are changed according to previously provided information.

4. The method of claim 3 , wherein the previously provided information includes at least one of a relative position and an arrangement interval between at least two cameras.

5. The method of claim 1 , further comprising separating foregrounds and backgrounds in the reference image and the at least one comparative image,

wherein the reference image and the at least one comparative image are captured using an infrared illuminator, and the feature-point extraction and the correspondence-point tracking are performed for the foregrounds of the reference image and the at least one comparative image.

6. The method of claim 5 , wherein the separating of the foregrounds and the backgrounds includes separating an area having fixed luminance or more as the foreground in each image or separating a fixed ratio of each image as the foreground.

7. The method of claim 1 , wherein the reference image and the at least one comparative image are captured by at least two cameras from different points of view at the same time.

8. A method for stereo matching of a reference image and at least one comparative image captured from different points of view using a computer device, comprising:

collecting the reference image and the at least one comparative image;

generating at least one lower resolution version of the reference image and at least one comparative image, having successively lower pre-determined resolution than the corresponding reference image and at least one comparative image;

extracting first feature points from a version of the reference image;

tracking first correspondence points corresponding to the first feature points in a version of the at least one comparative image using an optical flow technique;

extracting second feature points from a version of the at least one comparative image;

tracking second correspondence points corresponding to the second feature points in a version of the reference image, using the optical flow technique; and

generating a depth map by synthesizing tracking results of the first and second correspondence points,

wherein the extraction of the first and second feature points and the tracking of the first and second correspondence points are started for images having lowest resolution and finally performed for images having original resolution.

9. The method of claim 8 , wherein the extracting of the first and second feature points includes extracting points within a fixed ranking by calculating unique values of a gradient matrix G in positions of the reference image and the at least one comparative image and arranging points in descending order for a smaller unique value between two unique values produced as calculation results.