Face Recognition and Expression Recognition using Morphable Models

<P>IN THIS THESIS, WE PROPOSE A MORPHABLE MODEL CONSTRUCTION FRAMEWORK THAT MINIMIZES BOTH THE GLOBAL DISTANCE BETWEEN SURFACES (GLOBAL ENERGY) AS WELL AS THE DISTANCE BETWEEN CORRESPONDING FACE FEATURES ON DIFFERENT FACES (LOCAL ENERGY). LOCALIZED FACE FEATURE POINTS SUCH AS EYE CENTERS, NOSE-TIP, MOUTH AND ETC ARE DETECTED AND HENCE THE FEATURES CAN BE USED TO CONSTRAIN THE MOVEMENT OF THE SURFACE DURING THE CONSTRUCTION OF THE MORPHABLE MODEL</P><P>IN OUR WORK, INTEGRATED INTER-FACE-MODEL (INTER-FACE), WHICH CONTAINS INTER-FACE VARIATIONS, IS USED TO PERFORM FACE IDENTITY RECOGNITION. TO ACCOUNT FOR EXPRESSION-RELATED INTRA-FACE VARIATIONS, INTRA-FACE-MODEL (INTRA-FACE) IS CREATED USING EXPRESSIVE FACES FOR EVERY SUBJECT IN THE DATABASE. WITH THE USE OF THE INTER-FACE AND INTRA-FACE MODEL, WE SHOW THAT A SUCCESS RATE OF OVER 85% IS ACHIEVED FOR BOTH IDENTITY AND EXPRESSION RECOGNITION.</P>

Author dc:creator

• WU SIN WAH

• Face recognition, expression recognition, morphable model