Improved pruning algorithms in multiscale real-time object detection

Object detection is an important area of research in computer vision. One of the most popular approaches for object detection is based on combining many weak classifiers together to achieve one strong classifier through a technique called Boosting. A modified version of this technique for real-time face detection was developed by Viola and Jones, where a weak classifier is created by iteratively selecting a best single feature from a set of a very large number of potential features. During the detection process, there is a need to apply pruning techniques on the candidate results from different scales to eliminate the weak candidates and keep the most promising one. This paper presents improved pruning algorithms that result in reducing the number of false positives. For object detection, a complete framework is implemented based on Viola and Jones, then the proposed pruning algorithms are applied to obtain better detection results.