Modified Gradient Method for Face Localization
Г. А. Кухарев, Paweł Masicz, Piotr Masicz · Kluwer Academic Publishers eBooks · 2006
In this paper we present our research and its results in face localization in color images. We use modified gradient method with oval object detection. This method gives very good speed results and our improvement gives better accuracy than it was in original gradient methods. Our system works well in different lighting condition and without any manual interactions. We will show how the system works with various pictures delivered from digital cameras, camcorders and television and with multiple faces localization. The differences in head size, lighting and background are considerable. We achieved very good results in detection rate - more than 93% in poor conditions and reach 99% in good. Our system is implemented in C++ and works well and fast.