A face recognition algorithm based on LLE-SIFT feature descriptors

• Published in: 2015 10th International Conference on Computer Science & Education (ICCSE) pp. 729 - 734
• Main Authors: Ye Jihua, Shi Shuxia, Chen Yahui
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2015
• Subjects: Accuracy; Face; Face recognition; feature descriptors; Feature extraction; Lighting; LLE; LLE-SIFT; Principal component analysis
• ISBN: 9781479965984; 1479965987
• DOI: 10.1109/ICCSE.2015.7250341

Scale Invariant Feature Transform(SIFT) could influence the real time due to a higher dimension calculation and a longer computation time in large-scale data storage and computing. This paper presents a concept about LLE-SIFT feature descriptors with LLE algorithm. The first step is to calculate feature points of all train images by the standard SIFT algorithm, and searching the neighbor area of these points in the original image, then calculating the gradient of the horizontal direction and the vertical direction to form a vector matrix, whose dimension is reduced by LLE algorithm to obtain a projection matrix. The second step is to obtain the neighbor area using the critical information of points in the scale image, and calculating the horizontal direction and the vertical direction of the neighbors area to form a vector matrix, then the LLE-SIFT feature descriptor is the multiplied of the vector matrix and the projection matrix. Experiments shows that LLE-SIFT is effective.