AN ADAPTIVE REALTIME FACE DETECTING METHOD BASED ON TRAINING

摘要

the present invention, the face image database on the training of the horizontal / vertical size of the projection of the cross-correlation coefficient value the features of the steps of the learning through the Bayesian learning, constructing weak classifiers based on the rectangular features with the training image set, and made by a step of boosting (boosting) by learning the weak classifier for best strong classifiers jidoe, when the image is input, improving the quality of the input image for the face detection and histogram equalization for minimizing the light effect and (pretreatment step), using the Bayesian learning results, and detecting a face candidate from the input image (line detection step), the strong classifier obtained by using the boosting check whether the input face image to detect the best face area (state detecting step), duplicate detection for the same face in the scanning process to the real-time face detection the blocking provides a process for removing unwanted detection (single face candidate selection step) adaptive real-time face detection method based on the learning, characterized in that.