An Expression Feature Extraction Method Based on Improved LBP with Wavelet Decomposition

Abstract

Facial expression feature extracted isn’t exquisite enough with the aid of original LBP (Local Binary Pattern) operator, merely focusing on differences between the center pixel gray value and the neighborhood pixel gray value. MLBP (Multiple Local Binary Patterns) is proposed in this paper, making some improvements to LBP algorithm. A novel approach to facial expression recognition based on MLBP operator with wavelet decomposition is put forward, effectively reducing the dimensionality of facial expression feature vector. Experimental results show that the method presented can efficaciously improve recognition rate.