Robust multi-view face detection methods and apparatuses

A computer-readable medium having computer-implementable instructions for causing at least one processing unit to perform acts comprising:

• detecting possible human face image data within a digital image using a multiple stage face detection scheme that includes:
• a boosting filter stage to process a set of initial candidate portions of digital image data to produce a set of intermediate candidate portions using a plurality of boosting chain nodes and a boot strap function following each of the plurality of boosting chain nodes, the boot strap function to adjust a sample weight initialized for a current boosting classifier of a current boosting chain node based on a classification error rate of a previous boosting node;
• and
• a post-filtering stage configured to process the set of intermediate candidate portions to produce a set of final candidate portions, wherein the post-filter stage includes an image pre-processing process, a color-filtering process, and a support vector machine (SVM) filtering process, the SVM filtering process to perform wavelet transformation on the set of intermediate candidate portions.

An apparatus comprising:

• one or more processors;
• and
• memory having instructions including logic executable by the one or more processors to Perform a multiple stage face detection scheme to detect at least one human face within a digital image, the memory including:
• a boosting filter stage configured to process a set of initial candidate portions of digital image data using a boosting chain to produce a set of intermediate candidate portions, wherein the boosting chain includes a plurality of boosting chain nodes to identify candidate portions and a boot strap function following each of the plurality of boosting chain nodes, the boot strap function to use a weak learner of a previous boosting chain node in training another boosting chain node of the boosting chain, wherein the weak learner includes building a simple decision stump on a histogram of a Haar-like feature on a training set;
• and
• a post-filter stage configured to process the set of intermediate candidate portions to produce a set of final candidate portions, wherein at least one of the final candidate portions includes detected face image data.