Curvelet based CRLBP texture descriptor for facial expression recognition

In this paper, we proposed a new technique for facial expression recognition based on extraction of Complete Robust Local Binary Pattern (CRLBP) features from curvelet domain. The curvelet transform show evidence of improved multiscale directional capability, and a greater ability to localize distributed discontinuities such as edges along curves as compared to traditional multiscale transform such as wavelet transform. Hence, we transform original face images to frequency domain using curvelet transform. Noise and illumination invariant features are extracted from approximate sub-band using complete robust local binary pattern, which forms feature descriptor of facial expression. The proposed technique is assessed based on facial expression recognition carried out using a benchmark database such as JAFFE. Recognition of facial expression is achieved using a chi-square distance measure with a nearest neighbour classifier. Experimental results prove that our method outperforms compared to other popular LBP based methods.