Facial Feature Extracting Based on Active Shape Model under Fewer Landmarks Condition
Zhen-hai Ji, Wenming Zheng, Cairong Zou, Li Zhao · 2006
The active shape model (ASM) is composed of two parts, global shape model and local texture model, which all derived from the enough landmarks around the shape. If we have only fewer landmarks for some reasons, then how to utilize the ASM efficiently is a challenging and meaningful task. Because the region, which is coved by the landmarks of face contour, occupies almost half area of a whole face, so the landmarks of face contour act as more important role than other landmarks. Under the consideration of this idea, the paper presents a strategy that combines the changes of gravity center of face contour with basic ASM to overcome the shortcoming of fewer landmarks. The experiments show our method consumes less time in searching target shape than basic ASM, and has slight wider capture range and same good performances in point location accuracy under fewer landmarks