On Effectiveness of Histogram of Oriented Gradient Features for Visible to Near Infrared Face Matching
Citations
231 citations
150 citations
Cites background or result from "On Effectiveness of Histogram of Or..."
...[4] study how the histogram of oriented gradients (HOG) feature and its variants can help crossspectral face recognition tasks....
[...]
...We are able to significantly outperform the baseline [24] as well as some good results reported in [4] (73....
[...]
146 citations
143 citations
Cites background from "On Effectiveness of Histogram of Or..."
...…2013; Jin, Lu, and Ruan 2015), coupled discriminant face descriptor (C-DFD) (Lei, Pietikainen, and Li 2014; Jin, Lu, and Ruan 2015), DSIFT+PCA+LDA (Dhamecha et al. 2014), coupled discriminant feature learning (CDFL) (Jin, Lu, and Ruan 2015), Gabor+RBM+Remove 11PCs (Yi et al. 2015), VIS+NIR…...
[...]
...To verify the performance of IDR, we compare our method with state-of-the-art NIR-VIS recognition methods, including PCA+Sym+HCA (Li et al. 2013), learning coupled feature spaces (LCFS) method (Wang et al. 2013; Jin, Lu, and Ruan 2015), coupled discriminant face descriptor (C-DFD) (Lei, Pietikainen, and Li 2014; Jin, Lu, and Ruan 2015), DSIFT+PCA+LDA (Dhamecha et al. 2014), coupled discriminant feature learning (CDFL) (Jin, Lu, and Ruan 2015), Gabor+RBM+Remove 11PCs (Yi et al. 2015), VIS+NIR reconstruction+UDP (Juefei-Xu, Pal, and Savvides 2015)....
[...]
...2013; Jin, Lu, and Ruan 2015), coupled discriminant face descriptor (C-DFD) (Lei, Pietikainen, and Li 2014; Jin, Lu, and Ruan 2015), DSIFT+PCA+LDA (Dhamecha et al. 2014), coupled discriminant feature learning (CDFL) (Jin, Lu, and Ruan 2015), Gabor+RBM+Remove 11PCs (Yi et al....
[...]
...The methods can be ordered in ascending rank-1 accuracy as PCA+Sym+HCA, LCFS, C-DFD, CDFL, DSIFT+PCA+LDA, VIS+NIR reconstruction+UDP, Gabor+RBM+Remove 11PCs, IDR....
[...]
128 citations
References
46,906 citations
31,952 citations
"On Effectiveness of Histogram of Or..." refers background or methods in this paper
...Note that, DSIFT and HOG-UoCTTI utilize directed gradient whereas HOG-DT utilizes undirected gradient....
[...]
...The key contributions of this work are: • Evaluation and performance analysis of three HOG variants, namely ◦ Dense Scale Invariant Feature Transform (DSIFT) [16], ◦ Dalal-Triggs HOG (HOG-DT) [17], and ◦ HOG-UoCTTI [18], along with classification by LDA and direct feature matching....
[...]
...• The performances of DSIFT, HOG-UoCTTI, and HOG-DT reveal that the later lags in the comparison of formers....
[...]
...For classical HOG d = 128, for HOGDT d = 36, and for HOG-UoCTTI d = 31 (no is set as 8, 9, and 9, respectively and m = 8 leads to a total of 64 key points located on a uniform grid)....
[...]
...There are many variants of HOG descriptors such as DSIFT [16], HOG-DT [17], and HOG-UoCTTI [18]....
[...]
14,773 citations
10,501 citations
"On Effectiveness of Histogram of Or..." refers background or methods in this paper
...Note that, DSIFT and HOG-UoCTTI utilize directed gradient whereas HOG-DT utilizes undirected gradient....
[...]
...The key contributions of this work are: • Evaluation and performance analysis of three HOG variants, namely ◦ Dense Scale Invariant Feature Transform (DSIFT) [16], ◦ Dalal-Triggs HOG (HOG-DT) [17], and ◦ HOG-UoCTTI [18], along with classification by LDA and direct feature matching....
[...]
...• The performances of DSIFT, HOG-UoCTTI, and HOG-DT reveal that the later lags in the comparison of formers....
[...]
...For classical HOG d = 128, for HOGDT d = 36, and for HOG-UoCTTI d = 31 (no is set as 8, 9, and 9, respectively and m = 8 leads to a total of 64 key points located on a uniform grid)....
[...]
...Dimensionality of the feature descriptor f is (m×m)×d. Generally, the size of this descriptor is 2,000-10,000 which is a significantly large feature vector size with the training set of approximately 2000 samples....
[...]
6,384 citations