Distance measure used for object detection. (Edge-based) templates are compared with the edge image of an input image at different locations by computing for each template edge pixel the distance to the nearest image edge pixel.

Then we compute the average of all these distances which gives the final matching score (distance value).

Variants:

- use different distance metrics than Euclidean, e.g., Manhattan distance (Taxicab geometry)
- compute maximum distance instead of sum of distances (“directed Hausdorff distance”)

So the Chamfer score is the average distance from template points to nearest image edge points. These nearest distances can be computed from the distance image.

H.G. Barrow, J.M. Tenenbaum, R.C. Bolles, and H.C. Wolf.

Parametric correspondance and chamfer matching: two techniques for image matching

In Proc. 5th International Joint Conference on Artificial Intelligence, pages 659-663, 1977

It seems to be integrated into OpenCV according to these notes.

public/chamfer_distance_measure.txt · Last modified: 2014/01/03 15:33 (external edit) · []