A robust face recognition through statistical learning of local features
ICONIP'11 Proceedings of the 18th international conference on Neural Information Processing - Volume Part II
Combining SIFT and global features for web image classification
PCM'12 Proceedings of the 13th Pacific-Rim conference on Advances in Multimedia Information Processing
Hi-index | 0.01 |
Recently, the Scale Invariant Feature Transform (SIFT) proposed by Lowe has emerged as a cut edge methodology in general object recognition as well as for other machine vision applications. However, SIFT method has not shown successful results in face recognition problem because of its original matching strategy which does not consider the location of local keypoints. This paper proposes a novel keypoints matching strategy for face recognition. The proposed matching strategy can avoid mis-matching of local keypoints by using regular grid of face image and can give robustness to various transformations by using keypoint voting strategy. By performing computational experiment on the AR face data set, we confirmed the proposed matching strategy gives better performance than the conventional methods. Especially, the proposed method can give robust and best performance for facial images with occlusions.