表情認識のための顔領域の高精細抽出

概要

論文の詳細を見る
A method for extracting a facial area from a natural background using variable interframe subtraction and color information is presented in this paper. In the previous method, the thresholds for extracting faces were decided according to the resolution and characteristics of the facial area. However, there still remain problems to be solved such as extraction accuracy and the evaluation of the facial area extracted. In terms of moving information, the variable inter-frame subtraction method is adopted for conforming to a natural office environment. When using this method, however, for extracting moving information, an adequate frame interval must be evaluated for subtracting between frames. The most adequate frame interval for obtaining differential images between frames is set in this method. In this system, a robust extraction system is proposed. From a computer simulation using 10438 images, 9185 images (95.12%) were correctly extracted.
2010-10-30