Three-dimensional face recognition using shape codes extracted from projected structured light patterns
D. Choudhury · 2006
A technique for three-dimensional (3D) human face recognition is presented that utilizes 3D shape codes derived from projected structured light patterns onto the face objects. The distortion-induced phase difference associated with the projected structured patterns, which is due to the 3D geometric shape variations of the face, is extracted and computed using a Fourier-transform analysis tool. The computed phase map is utilized as a shape code to synthesize a spatial harmonic function whose spatial frequency distribution is directly proportional to the computed phase values, and hence to the 3D shape of the original face. This spatial harmonic function may be used as a signature function for a particular face. Synthesized signature functions corresponding to the test faces are cross-correlated digitally with that due to the target face to produce high correlation peak for matching of the target face and no peak or noise for mismatch of the faces. The method does not require explicit reconstruction of the 3D shape which is a tedious job. Feasibility of the technique is demonstrated with experiments. Results are presented that show an excellent performance for 3D face recognition.