YUV Face-Skin Detection

Recognition of face and skin areas is an important task in current video processing applications. Software for this purpose is designed to detect regions that probably belong to faces or to skin patches. The results can be used in a broad spectrum of applications, from modern video compressors (like H.264) to security surveillance systems. Our technology offers fast and efficient face and skin patch detection and tracking, along with the ability to make decisions that are based on motion and color information analysis. These characteristics allow for color video processing.

Unlike our technology, most other state-of-the-art classifier-based methods fail to sufficiently account for face orientation and for partial occlusion of faces. The drawback of our approach is a somewhat higher false alarm probability than that of properly trained, conventional classifier-based methods.

Primary Applications

• Digital camcorders
• Preprocessing for regions-of-interest selection algorithms
• Preprocessing for lossy video codecs to achieve better visual quality/compression ratio trade-off
• Security surveillance systems

Key Features

• One-pass real-time processing
• Fully automatic
• Non-frontal tilted face detection
• Skin patch detection
• Suitable for hardware implementation

Basic Deliverables

• Source code for a reference implementation in C
• C and assembly language source code for an implementation optimized for the PC (if required)
• Algorithm description
• Software description
• Verification instructions