Mathematical morphology based face segmentation and facial feature extraction for facial expression recognition

In this work, image processing and pattern recognition techniques are applied to extract human faces and facial features from color and gray images for facial expression recognition. Firstly, the image is segmented into skin and non-skin regions by using a color space model. Mathematical Morphology based techniques are applied for noise removal and hole filling. Then, ellipse region properties are used to determine whether a segmented skin region is a face or not. Secondly, the eyes and mouth areas are roughly located. Finally, eye features are extracted using the threshold values, and lip feature are extracted by using K-mean clustering. Both the above features are refined using morphological operations.