1. 2. Skin - color filtering

2. 3. Rejector Cascade
In this part, a rejection cascade is designed to reject a large number of non-face samples while detecting almost 100% of the faces. The cascade can reduce the computation time before more complex face classifiers

3. 3. Rejector Cascade

4. 4. Cascade of AdaBoost face classifier
For those windows that contain many skin-color pixels and are accepted by the rejector cascade, they should be further evaluated by the face detector. In this part, we present the algorithm to construct a strong classifier using AdaBoost algorithm based on Haar-like features

5. 4. Learning face detection
a. Haar-like features: are digital image features used in Object recognition. Each Haar –like feature consists of two or three jointed : black and white rectangles.
* Edge features
* Line features
* Center surround features

6. 4. Cascade of AdaBoost face classifier
The value of a Haar – like feature is the difference between the sum of the pixel gray level within black and white rectangular regions.
The rectangle Haar – like features can be computed repidly using “ integral image”.
Integral image at location of x, y contains the sum of the pixel values above and left of x, y, inclusive.

7. 3. Cascade of AdaBoost face classifier
b. AdaBoost (adaptive boost) is an iterative learning algorithm to construct a strong classifier using only a training set and weak learning algorithm. A weak classifier with the minimum classification error is selected by the learning algorithm at each iteration
Adaboost is adaptive in the sense that later classifiers are turned up in favor of those sub-windows misclassified by previous classifiers

8. 3. Learning face detection

9. 4. Cascade of AdaBoost face classifier
c. Cascade of Classifiers

10. 5. conclusion