Adaptively weighted sub-pattern PCA for face recognition

Local ridge regression for face recognition

“…For using as much local information hidden in face images as possible to relax the influence of local variation for recognition, recently various local region matching techniques have been developed [4][5][6][7][8][9][10]. The general idea of local region matching techniques is to first locate several facial features (components), and then classify the faces by comparing and combining the corresponding local statistics [11].…”

“…Zou et al [11] have verified that comparison of corresponding local regions is better than comparing corresponding facial components. So, in this paper, we adopt the simplest rectangular regions to partition images, which not only are conveniently used but also can better preserve the spatial and geometrical information in the original images [4][5][6].…”

“…Then we collect these vectors at the same position of all face images to form a specific local region training set, in this way, L separate local region sets are formed [5]. This process is illustrated in Fig.…”

Warp that smile on your face: Optimal and smooth deformations for face recognition

“…[30] * 4.67 -Stringface [4] 13.00 -PWCMr [14] -16.00 LGBPHS [36] 16.00 -SOM [29] 25 - * used only a subset of occlusions Partial Dist. [30] 0.00 3.00 7.00 12.00 14.00 37.00 12.00 Aw-SpPCA [28] 0.00 2.00 12.00 12.00 10.00 36.00 12.00 SOM [29] 0.00 2.00 12.00 12.00 10.00 36.00 12.00 SubsetModel [22] 3.00 10.00 17.00 26.00 23.00 25.00 17.00…”

“…In comparison to state-of-the-art methods the proposed extensions achieve significantly better recognition results. This is interesting since all of competitive approches except for already mentioned Partial Distance [30] method use a significantly larger amount of training data in order to learn a representative subspace via Gaussian mixtures [22] and self-organising maps [28,30] which can not compete with the performance of the presented warping algorithms, but are probably more efficient.…”