Facial image method and apparatus for semi-automatically mapping a face on to a wireframe topology

• US 6,031,539 A
• Filed: 03/10/1997
• Issued: 02/29/2000
• Est. Priority Date: 03/10/1997
• Status: Expired due to Term

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Digital Equipment Corporation (HP Inc.)
• Inventors: Kang, Sing Bing, Waters, Keith
• Primary Examiner(s): Powell, Mark R.
• Assistant Examiner(s): MANCUSO, HUEDUNG XUAN CAO
• Application Number:
  US08/815,981
• Time in Patent Office: 1,086 Days
• Field of Search: 345/419, 345/435, 345/425
• US Class Current: 345/419
• CPC Class Codes: G06T 15/04 Texture mapping

Abstract

A method for mapping a digitized image of a face on to a reference wireframe topology in a computer system is provided, the image composed of pixels, the wireframe composed of interconnected nodes, the method including the steps of determining facial features from the pixels, determining facial landmark points from the facial features, computing displacements for each one of the interconnected nodes, and manually finetuning the displacements in response to the step of computing.

Claims

1. A method for mapping a digitized image of a face on to a reference wireframe topology in a computer system, the image composed of a plurality of pixels, the wireframe composed of a plurality of interconnected nodes, the method comprising the steps of:

automatically determining a plurality of visible facial features from the plurality of pixels;

automatically determining a plurality of facial landmark points from the plurality of visible facial features using a contour of the image;

based on the determined facial landmark points, computing displacements for each one of the plurality of interconnected nodes, and

manually fine tuning the displacements in response to the step of computing.

2. The method for mapping a digitized image of a face on to a reference wireframe topology according to claim 1 wherein the step of determining the plurality of facial landmark points from the plurality of facial features comprises the steps of:

determining two eye locations and an approximate pupil location for each one of the two eye locations; and

determining an approximate mouth location.

3. The method for mapping a digitized image of a face on to a reference wireframe topology according to claim 2 wherein the step of determining the two eye locations comprises the steps of:

determining a template for each of two eyes; and

determining the location in a new face image corresponding to a best correlation with the two eye templates.

4. The method for mapping a digitized image of a face on to a reference wireframe topology according to claim 3 wherein the step of determining the mouth location comprises the steps of:

using an anthropometric ratio involving a separation of the two eyes to estimate an approximate mouth location;

determining the mouth location by finding a darkest horizontal line in a vicinity of the approximate mouth location.

5. The method for mapping a digitized image of a face on to a reference wireframe topology according to claim 4 further comprising the steps of:

computing a global transform of the plurality of facial landmark points;

determining a contour of the image by locating a plurality of strong edges parallel to an estimated tangent of an estimated contour; and

using the final distribution of the plurality of facial landmark points to compute a local displacement of each one of said plurality of reference wireframe nodes.

6. The method of claim 3 wherein further the step of determining the location in a new face of the eye templates is determined by minimizing an intensity error variance between the two eye templates and the digitized image of a face.

7. The method for mapping a digitized image of a face on to a reference wireframe topology according to claim 4 wherein the anthropometric ratio determines the approximate mouth location by drawing a box having a horizontal length approximately the separation of the pupils of the two eyes, and

the box having a horizontal and vertical midpoint located vertically below a midpoint of a line connecting the pupils by a distance approximately equal to the pupil separation.

8. The method for mapping a digitized image of a face on to a reference wireframe topology according to claim 4 wherein the finding of a darkest horizontal line in a vicinity of the approximate mouth location is performed by integral projection.

9. A method for mapping a digitized image of a face on to a reference wireframe topology in a computer system, the image composed of a plurality of pixels, the wireframe composed of a plurality of interconnected nodes, the method comprising the steps of:

determining a plurality of facial features from the plurality of pixels;

determining a plurality of facial landmark points from the plurality of facial features by determining a contour of the image by locating a plurality of strong edges parallel to an estimated tangent of an estimated contour to compute a global displacement factor;

determining a contour of the digitized image to compute local displacement factors;

computing displacements for each one of the plurality of interconnected nodes to transform the reference wireframe to the image, and

manually fine tuning the displacements from the step of computing.