Local Directional Weighted Threshold Patterns (LDWTP) for Facial Expression Recognition

Data Over the decade, automatic facial expression recognition system is commanding in various fields due to the instigation of biometric applications to overcome the security issues. Existing systems are suffering with some problems like illumination, texture, more calculation etc. which are used to extract the feature from the image. In this paper proposing local directional weighted threshold patterns for feature extraction to recognize the expression. Initially, calculated gradient magnitudes of every pixel in four directions later gradient image and original image is multiplied with weights to enhance the feature from the given image. Proposed method tested on CK+ database and performed well compared to other threshold methods like LBP, LTP etc.