The face detection algorithm based on local elastic potential energy feature

Abstract

Translate Article

Face detection is an important step in face recognition. Ineffective algorithm used for face detection will have a negative impact on the performance of face recognition. Face detection technology is not only a key step in face recognition technology, but also is an independent widely-used technology. It is important to design a suitable feature for face detection technology. This paper proposes a new face feature, which can be used in face detection. The face feature has the property of rotation and scaling invariant under some certain conditions. The algorithm regards an image as an elastic surface and regards the grey level as deformation. The algorithm calculates local elastic potential vectors and uses them as the input to the adaboost classifier. The output of the classifier is the final result. Theoretical analysis and experiments demonstrate that the feature proposed in the paper can improve accuracy to some certain extent and is robust to rotation and scaling transformation.