Multimodal 2D- 3D face recognition using structural context and pyramidal shape index

Combination of 2D and 3D face recognition approaches intensifies recognition accuracy. In this paper, we propose a new algorithm for face recognition by applying hybrid approach, structural context and pyramidal shape index. Proposed pyramidal local shape index descriptors are extracted in each level or scale of the Gaussian pyramid of range image. In this way, we can extract high contrast and reliable 3D face features. We extract Scale Invariant Feature Transform on pyramidal shape index image and histogram of structural context is used to find matched key points. A local descriptor structural context represents the structure of the image using SIFT. Structural context histogram is applied in both texture and range images to find SIFT matched points as 2D and 3D matching score respectively. Score level fusion using sum rule is applied to get final matching score. Experimental results on Face Recognition Grand Challenge (FRGC v2.0) database illustrate detection rate 98.8% and 98.5% at 0.1% false acceptance rate for All vs. All and ROC III experiments respectively. Comparing to the state of the art, these are the best results.