Memetically Optimized MCWLD for Matching Sketches With Digital Face Images
Citations
63 citations
61 citations
Cites methods from "Memetically Optimized MCWLD for Mat..."
...[24] used modified Weber’s local descriptor and memetic optimization and achieved an accuracy of 84....
[...]
...We demonstrate the effectiveness of the proposed method on three face sketch databases: 1) the Chinese University of Hong Kong (CUHK) face sketch (CUFS) database [5]; 2) the CUHK face sketch FERET (CUFSF) database [23]; and 3) the IIIT-D viewed sketch database [24], and show that the proposed S-FSPS method achieves superior performance compared with the state-of-the-art methods....
[...]
...Then, we validate the superior performance of our method compared with the state-of-the-art synthesis methods in terms of both qualitative and quantitative experiments on three public face sketch databases: 1) the CUFS database [5]; 2) the CUFSF database [23]; and 3) the IIIT-D viewed sketch database [24]....
[...]
58 citations
Cites background from "Memetically Optimized MCWLD for Mat..."
...0 [32], CUHK VIS-NIR, IIIT-D Sketch [40], and CUHK Face Sketch (CUFS) [41]....
[...]
...We then conduct extensive experiments on CASIA NIR-VIS 2.0 [32], CUHK VIS-NIR, IIIT-D Sketch [40], and CUHK Face Sketch (CUFS) [41]....
[...]
57 citations
Cites background or methods from "Memetically Optimized MCWLD for Mat..."
...The main sketch/photo databases are 159 pairs identified by [12], and 190 pairs in the IIITD database [2]....
[...]
...Motivated by this, the computer vision [12] and biometrics [2] fields have extensively studied sketch to photo face matching....
[...]
...In computer vision, facial sketch-photo matching has been studied extensively using a variety of approaches including invariant feature engineering [1, 2, 4, 12], crossmodal regression/synthesis [22, 23] and shared subspace learning [20]....
[...]
...The cross-modal sketch-photo gap is thus small, and viewed sketches are relatively easy to match – resulting in benchmark performance saturated at near-perfect [1, 2, 4, 12]....
[...]
...Later studies such as [2] improved these results, again combining feature engineering (Weber and Wavelet descriptors) plus the discriminative learning (genetic algorithms) strategy to maximise matching accuracy....
[...]
53 citations
Cites background from "Memetically Optimized MCWLD for Mat..."
...Later studies such as [13] improved these results, again combining feature engineering (Weber and Wavelet descriptors) plus discriminative learning (genetic algorithms) strategy to maximize matching accuracy; while [16] followed up also with feature engineering (LBP) and discriminative learning (RS-LDA)....
[...]
...Alternatively, matrices W may also be learned by discriminative models [1, 8, 13] to maximize matching rate....
[...]
...In each case, strategies to bridge the cross-modal gap broadly break down into four categories: (i) those that learn a cross-modal mapping to synthesise one modality from the other, and then perform within-modality matching [10, 11], (ii) those that learn a common subspace where the two modalities are more comparable [12], (iii) those that learn discriminative models to maximise matching accuracy [1, 13], and (iv) those that engineer features which are simultaneously invariant to the details of each modality, while being variant to person identity [14, 4]....
[...]
...(1), where |·| indicates some distance metric such as L1, L2 [1] or X 2 [13, 14]....
[...]
References
46,906 citations
14,708 citations
Additional excerpts
...On the other hand, sparse descriptor such as Scale Invariant Feature Transform (SIFT ) [23] is based on interest point detection and computing the descriptor in the vicinity of detected interest points....
[...]
14,157 citations
5,563 citations