A New Approach To Face Detection Based On Binary Texture Extraction Algorithm

Face detection is the process to search all faces’area in the input image (if it exists), andit is a key part of the automatic face recognition system. Early face recognition research isfocused on the image of a face with strong constraints (such as the no-background image),and often assumes that the position of face is known or readily accessible, thus the facedetection problem is not taken seriously. In recent years, with the development of e-commerceand other applications, the researchers have been studying face detection as a separate subjectand many face detection methods have been appeared. However, due to lighting, faceexpression and posture, shelter and complex image background,face detection problem hasnot yet completely resolved.In this paper the author proposes a new face detection process based on the analysis ofmainstream face detection process and the overall complexity of the process. The author usesgrayscale as original input image in the process, introducing a new texture-based detectionalgorithm----binary texture extraction algorithm for image preprocessing. A lot ofexperimental results show the face parts have significantly protruded facial contour. Thisalgorithm is applicable to different light, different colors and complex background. Also, it’sable to extract the contours of the objects and texture features in the image accurately.Furthermore, the author designs a face coarse inspection method according to theprinciples of the image morphology.Finally, the author proves the applicability and robustness of the new face detectionprocess and the binary texture extraction algorithm with a set of experiments contained avariety of imaging environments and different colors test library. The author adopts Matlabtools and combines the binary texture extraction algorithm with the mainstream face detectiontechnology, such as BP neural network, Gabor and Adaboost.