A new orientation-based face recognition method is proposed in this paper. The orientation analysis is performed with Gabor wavelet transform (GWT). The multispectral imagery includes the visible (RGB) and thermal (long-wave infrared) face images from the same group of subjects. The recognition performance of the new method is compared with that of three classical algorithms, principle component analysis, linear discriminant analysis, and elastic bunch graph matching. A score-level fusion of several algorithms versus multispectral images is explored and presented. Specifically, at each frequency band of GWT, an index number representing the strongest orientational response is selected, and then encoded in a binary number to favor the Hamming distance calculation. Multiple-band orientation codes are then organized into a face pattern byte (FPB) by using order statistics. With the FPB, Hamming distances are calculated and compared to achieve face identification. The FPB has the dimensionality of 8 bits per pixel and its performance will be compared to that of face pattern word (32 bits per pixel). The dimensionality of FPB can be further reduced down to 4 bits per pixel, called face pattern nibble. Experimental results with the multispectral faces of 96 subjects show that the proposed orientation-based face recognition method is very promising in contrast with three classical methods. Furthermore, the recognition performance with score-level fusion achieves 100% when tested on the entire multispectral database.