Directional Edge Coding for Facial Expression Recognition System

Because the discriminative edges cannot currently be encoded, the local appearance-based texture descriptors used in facial expression recognition are only partially accurate. The main cause is the existence of noise-induced distortion and weak edges. We suggest a new local texture descriptor called as Weighted Directional Edge Coding for facial expression recognition system (DEC-FERS) to solve these issues instead of using the traditional local descriptors. DEC-FERS looks at neighboring pixels support for determining facial expression attributes including edges, corners, lines, and curved edges. DEC-FERS extracts weaker edge responses through edge detection masks and discards them after encoding only more robust edge responses. Robinson Compass Mask and Kirsch Compass Mask were two makes used for edge detection considered for the extraction of edge features. In addition, the DEC-FERS lessens the redundancy by removing redundant pixels those won’t have any contribution for center pixel.