Application of Support Vector Machine Classifiers for Mouth Localization in Facial Images

摘要

An accurate mouth localization in facial images is a crucial preprocessing step for visual speech recognition. The aim is to define a rectangular region of interest comprising only the mouth; this is a non-trivial task especially in low-contrast gray level images. Here we propose to investigate the suitability of support vector machine classifiers in mouth localization, since they proved a powerful technique in other image processing tasks. We also propose a localization scheme based on support vector machines. The good localization results obtained show that support vector machines can be successfully used for mouth localization.