Local Binary Patterns are used in face detection/recognition because of its fast performance and robustness.
LBP generally works by looking at each pixel and comparing it to its neighbors. If the center pixel's intensity is greater than the neighboring pixel, a 0 is given to the neighbor - otherwise, a 1 is given. The results are then concatenated to form a binary code for the center pixel, which can be changed to decimal format for simpler representation. This technique is used often for texture classification, and is robust to most lighting conditions.
Some algorithms for facial recognition involving LBP first divide an image into several smaller regions from which the features are extracted using LBP. This generates a feature histogram for each region which can be concatenated to form a visual descriptor of the face. This can be paired with other machine learning techniques like support vector machines for facial recognition.
Best Answer