Published Paper Details:

In real time, Facial expression recognition has always been a difficult task in the field Computer Vision and Image Analysis. Expressions are generated from very bendable features and differ in sexes, ages and cultures among people. Furthermore, varying illuminations, complicate surroundings and complex backgrounds makes the recognition more complex. The Local Binary Pattern (LBP) which acquires the local area characteristics from the images are widely for facial expression recognition. But still LBP are able to solve the few issues of the local regions such as occlusions, varying lighting conditions and so on. This study presents the local feature descriptor, Elliptical Local Binary Pattern (ELBP) for facial expression recognition. In ELBP, horizontal and vertical ellipse patterns are used to acquire micro facial features in both horizontal and vertical directions. For comparing the ELBP histograms, a template matching strategy called Chi-Square distance is suggested. The efficiency of the proposed method is evaluated on bench-marking datasets JAFFE.