Method and system for customizing facial feature tracking using precise landmark finding on a neutral face image

METHOD AND SYSTEM FOR CUSTOMIZING FACIAL FEATURE

TRACKING USING PRECISE LANDMARK FINDING ON A

NEUTRAL FACE IMAGE

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority under 35 U.S.C. §119(e)(l) and 37 C.F.R. § 1.78(a)(4) to U.S. provisional application serial number 60/220,288, entitled METHOD AND SYSTEM FOR CUSTOMIZING FACIAL FEATURE TRACKING USING PRECISE LANDMARK FINDING ON A NEUTRAL FACE IMAGE and filed My 24, 2000; and claims priority under 35 U.S.C. § 120 and 37 C.F.R. § 1.78(a)(2) as a continuation-in-part to U.S. patent application serial number 09/188,079, entitled WAVELET-BASED FACIAL MOTION CAPTURE FOR AVATAR ANIMATION and filed November 6, 1998. The entire disclosure of U.S. patent application serial number 09/188,079 is incorporated herein by reference.

BACKGROUND OF THE INVENTION The present invention relates to avatar animation, and more particularly, to facial feature tracking.

Virtual spaces filled with avatars are an attractive the way to allow for the experience of a shared environment. However, animation of a photo-realistic avatar generally requires robust tracking of an actor's movements, particularly for tracking facial features.

Accordingly, there exists a significant need for improved facial feature tracking. The present invention satisfies this need. SUMMARY OF THE INVENTION The present invention is embodied in a method, and related system, for customizing a visual sensor using a neutral face image of an actor. The method includes capturing a front neutral face image of an actor and automatically finding facial feature locations on the front neutral face image using elastic bunch graph matching. Nodes are automatically positioned at the facial feature locations on the front neutral face image of the actor. The node positions are then manually corrected on front neutral face image of the actor.

Further, the method may include generating a corrector graph based on the corrected node positions.

Other features and advantages of the present invention should be apparent from the following description of the preferred embodiments taken in conjunction with the accompanying drawings, which illustrate, by way of example, the principles of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a flow diagram for illustrating a method for customizing facial feature tracking using precise landmark finding on a neutral face image, according to the present invention.

FIG. 2 is an image of a visual sensor customization wizard having a camera image of an actor and a generic model image.

FIG. 3 is an image of a visual sensor customization wizard after automatic sensing and placement of node locations on a camera image of an actor's face.

FIG. 4 is an image of a visual sensor customization wizard having corrected node positions for generating a corrector graph, according to the present invention. FIG. 5 is a block diagram of a technique for generating a corrector graph using a neutral face image, according to the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention is embodied in a method and system for customizing a visual sensor for facial feature tracking using a neutral face image of an actor. The method may include generating a corrector graph to improve the sensor's performance in tracking an actor's facial features.

As shown in FIG. 1, the method captures a front face image of the actor (block 12) . The front neutral face image may be captured with the assistance of a visual sensor customization wizard 22, shown in FIG. 2. An example image 24 is shown to the actor to indicate the alignment of the captured image 26. Next, facial feature locations are automatically found using elastic bunch graph matching (block 14). Facial feature finding using elastic bunch graph matching is described in U.S. patent application number 09/188,079. In the elastic graph matching technique, an image is transformed into Gabor space using a wavelet transformations based on Gabor wavelets. The transformed image is represented by complex wavelet component values associated with each pixel of the original image. As shown in FIG. 3, nodes 28 are automatically placed on the front face image at the locations of particular facial features (block 16). Because of particular image characteristics of the actor, a facial feature graph placed over the actor's front face image may have nodes locations that are not properly placed on the front face image. For example, the four nodes for the actor's eyebrows are placed slightly above the eyebrows on the front face image.

The system operator may use the visual sensor customization wizard 22 to pick and move the nodes 28. The nodes are manually moved on the neutral face image 26 using a pointing device, such as a mouse, to select and drag a node to a desired location (block 18). For example, as shown in FIG. 4, node placement on the eyebrows of the actor's image has been adjusted to more closely aligned with the actor's eyebrows in accordance with the example image 24.

As shown in FIG. 5, after the nodes 28 for features, A through E, are correctly placed on the front neutral face image 24, image jets are recalculated for each facial feature and may be compared to corresponding jets in a gallery 32 of a bunch graph. The bunch graph gallery includes sub-galleries of a large number N of persons. Each person in the sub-gallery includes jets for a neutral face image 34 and for expressive facial images, 36 through 38, such as a smiling face or a face showing exclamation. Each feature jet from the corrected actor image 24 is compared with the corresponding feature jet from the neutral jets in the several sub-galleries. The sub-gallery neutral jet for a feature (i.e., feature A) that most closely matches the jet for the image feature A is selected for generating a jet gallery for the feature A of a corrector graph 40. As a more particular example, for the feature E, the sub-gallery for person N has a neutral jet for feature E that most closely corresponds to thejet for feature E from the neutral image 24. The corrector graph jets for facial feature E are generated using thejet for feature E from the neutral jets along with the jets for feature E from each of the expressive feature jets, 36 through 38, from the sub-gallery N. Accordingly, the corrector graph 40 is formed using the best jets, with respect to the neutral face image 24, from the gallery 32 forming the bunch graph.

The resulting corrector graph 40 provides a much more robust sensor for tracking node locations. A custom facial feature tracking sensor incorporating the corrector graph may provide a more photo-realistic avatar and an enhanced virtual space experience.

Although the foregoing discloses the preferred embodiments of the present invention, it is understood that those skilled in the art may make various changes to the preferred embodiments without departing from the scope of the invention. The invention is defined only by the following claims.