|
I
will describe a possible approach to face detection using a technique
that simultaneously finds the optimal correspondence and affine
transformation between a 2D model and image features. This technique
was developed by Gold, Rangarajan, Mjolsness and others at Yale.
The face model could be learned from examples by a clustering
method proposed by the same group.
The
papers that describe these techniques can be viewed at Anand Rangarajan's
home page, noodle.med.yale.edu/anand/profile.html
1.
Simultaneous correspondence and affine transformation:
Suguna
Pappu, Steven Gold and Anand Rangarajan, "A framework for
non-rigid matching and correspondence", Advances in Neural
Information Processing Systems 8, pp. 795-801, 1996 (short paper).
Steven
Gold, Anand Rangarajan, Chien-Ping Lu, Suguna Pappu and Eric Mjolsness,
"New Algorithms for 2D and 3D Point Matching: Pose Estimation
and Correspondence", in press, Pattern Recognition, 1997
(long paper).
2.
Learning a face model:
Steven
Gold, Anand Rangarajan and Eric Mjolsness, "Learning with
Preknowledge: Clustering with point- and graph-matching distance
measures", Neural Computation, 8(4):787-804, May 1996.
References
|
