Feature improvement for classification of face images under varying light conditions using a hybird algorithm

Abstract

Image improvement is an important process for enhancing the quality of facial images under varying light condition in which shadow and light affects the performance of feature extraction and face recognition. This research proposed the development of image normalization for illumination, such as dark light and over light that creates some invisible face area and it is unable to use the normal face recognition process . This research uses self-quotient image as a main algorithm that to be hybridized with the weber, mean filter and wavelet methods. The standard dataset called Yale B database is used for demonstrating the performance of our proposed algorithm. The dataset is divided into 4 datasets. The self-quotient image together with weber face and mean filter creates the best result for reducing the illumination from shadow and light and helps improve the face recognition rate to reach 99.40%.

Keywords

• face image under varying light condition
• filter
• improvement face image
• face recognition
• pixel