Among the appearance-based methods, we considered LBP, LDP, LDN, PTP, HOG, LPQ, Gabor, and LDTP. Furthermore, we compared against a method that uses a manifold based sparse representation (MSR)  and approaches dealing with intra-class variations , .
Person-independent expression recognition results on CK+ dataset by varying noise Descriptors Without Varying Noise Noise 0.08-0.16 0.16-0.32 LBP 85.84 72.09 69.09 LDP 88.07 78.28 57.44 LDN 88.58 76.45 52.75 LPTP 91.64 87.05 77.37 PTP 91.03 88.96 82.16 HOG 92.01 88.89 84.51 LDTP 93.58 89.91 86.24 HPED 93.74 91.12 86.93 Table 2.
The LDTP encodes intensity difference of the image in the first and second maximum directions, and thus it contains not only intensity information but also directional information.
To obtain the LDTP code, firstly computing the eight absolute edge response values [G.sub.i] of each pixel by Kirch masks [43, 49]:
Compared with LDTP, our proposed method with RBF kernel improves the recognition rate by approximately 2.7% in 6-class problem and 2.4% in 7-class problem.
In contrast to LDTP, our proposed method makes the average classification accuracy increased by approximately 3.6% on CK+ database and 2.4% on JAFFE database respectively.