Local Difference Sign-Magnitude Transform of Edge/Corner Features for Robust Face Recognition
Almabrok Essa Essa
In this research, a new appearance based feature descriptor, named Local Difference Sign-Magnitude Transform (LDSMT) is developed for robust face recognition, which efficiently summarizes the local structure of face images. LDSMT is a nonparametric descriptor that utilizes a combined edge/corner detection strategy. We obtain the information about corners and edges of the face image using the Frei and Chen edge detector, then for each pixel position there are two local differences to describe the relationship of pixels to their local neighborhood. The first one is using the sign (positive or negative) of the difference between the values of the central pixel and the neighboring pixel. The second one is using the magnitude of the difference between the central pixel and the neighboring pixel. Then a histogram is built for each component from each edge and corner map respectively. Finally, we concatenate these histograms together to form the final LDSMT feature vector. The performance evaluation of the proposed LDSMT algorithm is conducted on several publicly available databases and observed promising recognition rates.
Graduate Research - Graduate
Vijayan K Asari
Primary Advisor's Department
Electrical and Computer Engineering
Stander Symposium poster
"Local Difference Sign-Magnitude Transform of Edge/Corner Features for Robust Face Recognition" (2017). Stander Symposium Posters. 908.