Multi-face Location on Embedded DSP Image Processing System

This paper is devoted to the problem of Multi-face localization on embedded DSP image processing system. The application of face localization is becoming important in machine vision, systems video monitoring and security access system. Skin color acts as an important cue for segmentation, localization and tracking of face. We capture face image on DSP image processing system and build a skin-color model for face detection using the three components original data on YUV color space. Combining the character of the embedded DSP image processing system, a multi-face locating method based on the improved connected components labeling algorithm is presented in this paper. The experimental results show that the proposed approach not only has the features of high speed and high efficiency, but also has the ability to locate human faces regardless of the size, number, orientation and viewpoint of the faces. It can still get an ideal location effect in the inadequate lighting complicated background.