Abstract
This report proposes a simple yet robust face-parts tracking method aiming to optimum to consumer digital devices such as PDAs and Cell phones. Algorithms implemented on such consumer digital devices should be as efficient as possible because it should achieve both practical accuracy and real-time process under the lack of computational power and the limited amount of memory space. The proposed method detects face regions within video frames by employing a robust skin-color model which was developed in the field of Image Database. It then applies the Adaptive Histgram Equalization method towards the detected face regions for enabling to consistently detect such face-parts as eyes and a mouth which are darker than other parts. In addition, the method tracks those face-parts with simple logical arithmetic. Therefore, it is possible to guarantee robust and real-time process with the consumer device hardware. This report also presents experimental results.
