Pith. sign in

REVIEW 4 major objections 5 minor 95 references

Scoot: A Perceptual Metric for Facial Sketches

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read The paper claims a co-occurrence-texture metric, Scoot, tracks human judgments of sketch similarity better than standard image-quality metrics.

desk verdict Nice dataset and meta-measure idea, but Scoot's reported advantage is a fit to the same human judgments used to pick its hyperparameters—and even then Gabor beats it on the larger set. read the letter →

arxiv 1908.08433 v2 pith:WP5DY52W submitted 2019-08-21 cs.CV cs.DB

classification cs.CVcs.DB
keywords perceptualmetricfacesketchsynthesisco-occurrencetexturespatialstructureimagequalityassessmenthumanperception2AFCjudgmentsmeta-measure
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Facial-sketch comparison is a human perceptual task, yet the metrics commonly used to evaluate synthesized sketches such as SSIM and FSIM were built for photographic distortions and often rank sketches against human intuition. This paper tries to establish that a metric combining block-level spatial structure with co-occurrence texture statistics, named Scoot, tracks human perceptual choices far more closely, reaching about 76% agreement with human judgments on one benchmark and about 79% on another, versus roughly 25–59% for standard metrics. To support that claim, the paper proposes three meta-measures for resizing stability, rotation sensitivity, and content capture, and releases a 152k-judgment human-ranked sketch database. If the claim is right, automatic evaluation of face sketch synthesis can become substantially more perceptually aligned without a learned model.

What carries the argument

The load-bearing device is the gray-level co-occurrence matrix $\mathbf M$ computed on a quantized sketch $I'$, where $M(i,j)|_d$ counts how often gray value $i$ appears at a displacement $d$ from gray value $j$. The paper's contribution is to apply this classical texture statistic at block level: the image is split into a $k\times k$ grid, the matrix is normalized within each block, and the Contrast and Energy statistics are concatenated across blocks and averaged over four displacement orientations to form a feature vector; similarity is then $E_s = 1/(1+\|\vec\Psi(X'_s)-\vec\Psi(Y'_s)\|_2)$. This machinery carries the argument because it makes the metric sensitive to stroke direction and regional content while staying insensitive to small resizing and rotation, exactly where pixel-level metrics fail.

What would settle it

A held-out 2AFC sketch-similarity dataset built without the paper's selection rules, where the complete-versus-light-stroke ordering is not assumed in advance, would settle the claim: if a standard metric such as SSIM or FSIM, or a deep-feature metric, agreed with human choices at least as often as Scoot on such a benchmark, the claimed perceptual superiority would fail.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the perceptual gap in sketch evaluation comes from a mismatch of scale: pixel-level image-quality metrics miss the block-level spatial arrangement and stroke-direction statistics that human viewers use. Scoot measures similarity by quantizing each sketch to six gray levels, computing gray-level co-occurrence matrices inside a 4-by-4 grid of blocks, extracting Contrast and Energy statistics per block, averaging over four orientations, and comparing the resulting feature vectors with a Euclidean-distance-to-similarity mapping $E_s = 1/(1+\|\vec\Psi(X'_s)-\vec\Psi(Y'_s)\|_2)$. On the paper's three meta-measures and on its human two-alternative forced-choice databases, this construction outperforms IFC, SSIM, FSIM, VIF, and GMSD by wide margins. The paper reads that evidence as showing that spatial structure and co-occurrence texture are generally applicable perceptual features in face sketch synthesis.

Load-bearing premise

The claim rests on treating the three meta-measures and the newly collected human rankings as valid, unbiased yardsticks for perceptual similarity — in particular, on assuming a complete synthetic sketch is always perceptually better than a light-stroke thresholded version — and on assuming the settings tuned on those same benchmarks will keep their advantage on other sketches.

Editorial extensions

If this is right

  • Face sketch synthesis papers could evaluate quality with a metric that, on the paper's human-judgment data, agrees with people roughly 26 percentage points more often than the best standard metric tested.
  • Slight resizing and rotation of reference sketches, common when artist sketches do not align with photos, would no longer flip the ranking of synthesis algorithms the way they do for SSIM, VIF, and GMSD.
  • The block-level co-occurrence construction offers a simple, non-learned alternative to learned perceptual metrics for tasks where texture and stroke direction matter.
  • Complete sketches that preserve hair, eyes, and other facial texture are ranked above incomplete light-stroke versions, matching the expectation that content capture is part of perceptual quality.
  • The released human-judgment databases give future sketch-synthesis methods a way to test their evaluation choices against actual human perception.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • As an editorial extension, the same construction could be tested on non-face line drawings and stylized illustrations; the paper's argument only requires texture and spatial structure to be the relevant perceptual cues, which is plausible but unverified there.
  • Because the grid size, quantization level, and feature pair were selected on the same benchmarks used for evaluation, an independent evaluation on a separate dataset is needed before treating the reported margin as the metric's true advantage.
  • The meta-measures themselves could serve as a reusable protocol: any future sketch-similarity metric, learned or hand-crafted, could be reported against the same three properties, making cross-paper comparisons direct.
  • A natural extension is to let a learned deep network consume the block-level co-occurrence features as input; the paper shows hand-crafted features already carry a large share of the perceptual signal, so a hybrid might combine their stability with learned flexibility.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes Scoot, a perceptual similarity metric for facial sketches that combines gray-level co-occurrence matrix statistics (Contrast and Energy) computed on a k x k block grid, with features averaged over four orientations after quantizing images to N_l gray levels. The authors introduce three meta-measures: stability to slight resizing, rotation sensitivity, and content capture capability. They also collect a two-alternative forced-choice (2AFC) human judgment dataset (RCUFS/RCUFSF) with about 152k judgments. Experiments in Table 2 compare Scoot with classical IQA metrics and with texture/edge features, reporting that Scoot achieves the best performance on the meta-measures and high agreement with human judgments.

Significance. If the central claim were established, the paper would provide a simple and potentially useful perceptual metric for face sketch evaluation, and the human judgment dataset plus the meta-measure suite would be valuable community resources. The manuscript deserves credit for collecting a large-scale human judgment dataset, evaluating a broad set of texture and edge features, and including sensitivity analyses for grid size and quantization. The main obstacle is that the reported advantage of Scoot is partly a selection artifact: its hyperparameters and feature combination were chosen using the same evaluation criteria on which it is subsequently judged. The claim is also contradicted by the paper's own Table 2 for one of the two human-judgment benchmarks. The contribution is therefore promising but not yet supported by an independent test.

major comments (4)
  1. [Sec. 6, Table 2, Fig. 7] The hyperparameters of Scoot are selected using the same benchmarks that later serve as evidence. Section 6 states that the three meta-measures and the human judgments were applied to test every single feature, every feature pair, and the triple, and Fig. 7 shows k=4 and N_l=6 chosen after inspecting MM1-MM3 and Jud curves. Consequently, the favorable results for Scoot/CE in Table 2 are a fit to the evaluation criteria, not an out-of-sample prediction. The paper needs a validation protocol in which k, N_l, and the feature set are frozen before evaluating on held-out datasets, or a nested cross-validation.
  2. [Table 2] The claim that Scoot exceeds prior work is not supported by Table 2 on the larger human-judgment benchmark. On RCUFSF, the Gabor feature configuration reaches 80.9% human-judgment agreement while Scoot/CE reaches 78.8%, and HE reaches 80.3%. The abstract and conclusion state unqualified superiority. This discrepancy must be addressed, either by restricting the claims to the specific meta-measures or by reporting a formal comparison that justifies the ordering across all criteria.
  3. [Sec. 4.2, Fig. 5] The MM3 content-capture meta-measure assumes as ground truth that a complete SOTA synthetic sketch is always perceptually better than a 'light stroke' version produced by thresholding at gray level 170. This assumption is not validated with human judgments and may not hold for all reference sketches or synthesis algorithms. Since MM3 is one of the three pillars of the evaluation, the assumption needs justification, or the incomplete versions should be constructed and validated using human judgments.
  4. [Table 2 caption] The caption claims that all differences are statistically significant at the alpha<0.05 level, but no statistical test, sample size, variance estimate, or multiple-comparison correction is described anywhere in the manuscript. Without this information the reader cannot verify the significance claim, and the test would in any case be invalidated by the parameter selection on the same data. The authors should report a concrete test procedure and confidence intervals.
minor comments (5)
  1. [Abstract] The phrase 'quick assess' should be 'quickly assess', and there is a spacing error in 'metric,called'.
  2. [Sec. 3.3, Eq. (6)] The norm notation in Eq. (6) uses triple vertical bars with a subscript 2; this should be typeset consistently as a standard Euclidean norm.
  3. [Fig. 7] The caption refers to MM4, but only MM1-MM3 are defined in Sec. 4.2; clarify what MM4 denotes, likely the human-judgment measure.
  4. [Sec. 5.3] The sentence 'To increase an inherently noisy process' should be revised to something like 'To reduce the effect of noise in human judgments'.
  5. [Algorithm 1] Step 4 refers to 'CE features' before the choice of C and E is explained; the algorithm should state that p=2 and that the selected statistics are Contrast and Energy, with the selection procedure described in Sec. 6.

Circularity Check

1 steps flagged · score 6.0 of 10

The benchmark advantage is partly a fit: the CE feature set and hyperparameters were selected on the same human-judgment and meta-measure columns that Table 2 uses as evidence, with no held-out validation.

  1. fitted input called prediction [Sec. 4.1 Implementation Details; Sec. 6 Discussion; Table 2]
    "The size of spatial structure k in Sec. 3.2 is set to 4 to achieve the best performance. The quantization parameter Nl in Eq. (6) is set to 6 grades. We have demonstrated that p = 2 (e.g., C in Eq. 3 combined with E in Eq. 4) achieve the best performance (see Sec. 6). ... We have applied our three meta-measures as well as human judgments to test the performance of the Scoot metric using each single feature, each feature pair and the combination of all three features. ... Thus, we choose 'CE' feature as our final combination."

    The final comparison in Table 2 reports Scoot's agreement with the RCUFS/RCUFSF human judgments and its MM1/MM2/MM3 scores. Those are the same columns used in Sec. 6 to select the CE feature combination and in Sec. 4.1 and Fig. 7 to select k=4 and Nl=6. The paper provides no frozen-configuration held-out evaluation, so the headline claim that Scoot exceeds prior work is not an out-of-sample prediction; it is a report of the selection criterion. This is the fitted-input-called-prediction pattern: the configuration that performed best on the test criteria is then presented as the configuration that beats the baselines on those same criteria.

full rationale

The central issue is not that Scoot's formula is defined in terms of its evaluation data; the co-occurrence statistics, block grid, and Euclidean similarity are ordinary, independently motivated components. Neither is the paper's reliance on self-citations load-bearing: [9,10] are cited only as methodological inspiration alongside external references [33,37], and no uniqueness theorem or exclusive justification is imported from the authors' prior work. The real circularity is in the evaluation loop. Section 6 explicitly states that the three meta-measures and human judgments were used to test each single feature, each pair, and the triple, and that CE was chosen as the final combination on that basis; Section 4.1 similarly states that k=4 and Nl=6 were set to achieve the best performance. Table 2 then reports Scoot's performance on exactly those same meta-measure and human-judgment columns as evidence of superiority. Because no separate validation set or frozen-configuration test is described, the reported advantage is partly a selection artifact rather than an independent verification. The table itself also weakens the headline claim: on RCUFSF, Gabor reaches 80.9% human-judgment agreement and HE reaches 80.3%, while Scoot/CE reaches 78.8%, so the selected configuration is not uniformly best even on the selection criterion. This is partial circularity of the fitted-input-called-prediction kind, not a fully self-definitional derivation.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim rests on a small set of hand-chosen parameters (quantization levels, grid size, feature pair, orientations) and on domain assumptions borrowed from texture analysis and scene recognition. No new physical or conceptual entities are introduced. The main burden is that these parameters and assumptions are validated only on the authors' own benchmarks.

free parameters (5)
  • N_l = 6 (number of gray levels after quantization) = 6
    Selected via sensitivity experiments on MM1-MM4 (Fig. 7). Not derived from theory.
  • k = 4 (grid size for block-level spatial structure) = 4
    Selected as the most robust across MM1 and MM3 (Fig. 7).
  • p = 2 (feature combination Contrast and Energy) = CE
    Chosen because H+E+C gave no improvement and CE performed best on the authors' meta-measures and human judgments (Sec. 6, Tab. 2).
  • T = 4 (number of orientations) = 4
    Set empirically to four directions; the paper states 8 neighbor directions reduce to 4 due to symmetry (Sec. 4.1).
  • Threshold 170 for light-stroke meta-measure = 170
    Used to construct MM3 stimuli (Sec. 4.2).
assumptions (3)
  • domain assumption Co-occurrence statistics capture perceptual texture similarity.
    Borrowed from texture analysis literature (Haralick et al.); the paper applies it to face sketches without independent validation in this domain.
  • domain assumption Block-level spatial division preserves holistic content.
    Based on the spatial envelope strategy of Oliva and Torralba, assumed to transfer to sketch similarity.
  • domain assumption 2AFC human preferences are ground truth for perceptual similarity.
    The paper treats viewer rankings as the reference standard for all metric comparisons.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scoot: A Perceptual Metric for Facial Sketches." pith.science (2026). https://pith.science/paper/WP5DY52W

@misc{pith2026190808433,
  author       = {Pith},
  title        = {Pith review of: Scoot: A Perceptual Metric for Facial Sketches},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WP5DY52W}},
  note         = {Machine review of arXiv:1908.08433}
}
read the original abstract

Human visual system has the strong ability to quick assess the perceptual similarity between two facial sketches. However, existing two widely-used facial sketch metrics, e.g., FSIM and SSIM fail to address this perceptual similarity in this field. Recent study in facial modeling area has verified that the inclusion of both structure and texture has a significant positive benefit for face sketch synthesis (FSS). But which statistics are more important, and are helpful for their success? In this paper, we design a perceptual metric,called Structure Co-Occurrence Texture (Scoot), which simultaneously considers the block-level spatial structure and co-occurrence texture statistics. To test the quality of metrics, we propose three novel meta-measures based on various reliable properties. Extensive experiments demonstrate that our Scoot metric exceeds the performance of prior work. Besides, we built the first large scale (152k judgments) human-perception-based sketch database that can evaluate how well a metric is consistent with human perception. Our results suggest that "spatial structure" and "co-occurrence texture" are two generally applicable perceptual features in face sketch synthesis.

Figures

Figures reproduced from arXiv: 1908.08433 by the authors.

Figure 1
Figure 1. Which synthesized sketch is more similar to the mid￾dle sketch? For the right case, sketch 0 (S0) is more similar than sketch 1 (S1) w.r.t. reference (R) in terms of structure and texture. Sketch 1 almost completely destroys the texture of the hair. The widely-used (SSIM [66], FSIM [76]), classic (IFC [40], VIF [39]) and recently released (GMSD [72]) metrics disagree with humans. Only our Scoot metric agrees well wi… view at source ↗
Figure 2
Figure 2. Motivation of the proposed Scoot metric. (a) Pencil grades and their strokes. (b) Using stroke tones to present texture. The stroke textures used, from top to bottom, are: “cross-hatching”, “stippling”. The stroke attributes, from left to right, are: spare to dense. Images are from [67]. (c) The artist draws the sketch from guideline to details. (d) The original sketches. (e) The quantized sketches. (f) Creating var… view at source ↗
Figure 3
Figure 3. Meta-measure 1: Stability to Slight Re-sizing. (a) Reference (b) R-Reference (c) Pix2pix (d) MWF [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: Meta-measure 3: Content Capture Capability. For this meta-measure, we applied the θ = 1 − ρ [2] measure to test the metric ranking stability before and after the reference downsizing was performed. The value of θ falls in the range [0, 2]. Tab. 2 shows the results: the…
Figure 6
Figure 6. Figure 6: Our distortions. These distortions are generated by various real synthesis algorithms [27, 31, 45, 53, 55, 59, 64, 75, 92, 93]. ment (∼152K judgments) through the following stages: • We let the first group of viewers (7 subjects) select four out of ten sketches for eac…
Figure 7
Figure 7. Figure 7: Sensitivity experiments of the spatial structure (top) and quantization (bottom). For MM1 & MM2, the lower the better. For MM3 & MM4, the higher the better. MM1 & 2 & 3. Gabor is reasonable in terms of MM1 & 2, but not good at MM3. For edge-based features, Canny fails …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

95 extracted references · 79 canonical work pages

  1. [1]

    An investigation of the textural characteristics associated with gray level cooc- currence matrix statistical parameters

    Andrea Baraldi and Flavio Parmiggiani. An investigation of the textural characteristics associated with gray level cooc- currence matrix statistical parameters. IEEE T Geosci. Re- mote., 33(2):293–304, 1995

  2. [2]

    Algorithm AS 89: the upper tail probabilities of Spearman’s rho

    DJ Best and DE Roberts. Algorithm AS 89: the upper tail probabilities of Spearman’s rho. J R STAT SOC C-APPL , 24(3):377–379, 1975

  3. [3]

    A computational approach to edge detection

    John Canny. A computational approach to edge detection. IEEE TPAMI, 8:679–698, 1986

  4. [4]

    A new LDA-based face recognition system which can solve the small sample size problem

    Li-Fen Chen, Hong-Yuan Mark Liao, Ming-Tat Ko, Ja-Chen Lin, and Gwo-Jong Yu. A new LDA-based face recognition system which can solve the small sample size problem. Pat- tern Recognition, 33(10):1713–1726, 2000

  5. [5]

    BING: Binarized normed gradients for objectness estimation at 300fps.Computational Visual Media, 5(1):3–20, 2019

    Ming-Ming Cheng, Yun Liu, Wen-Yan Lin, Ziming Zhang, Paul L Rosin, and Philip HS Torr. BING: Binarized normed gradients for objectness estimation at 300fps.Computational Visual Media, 5(1):3–20, 2019

  6. [6]

    An analysis of co-occurrence texture statis- tics as a function of grey level quantization

    David A Clausi. An analysis of co-occurrence texture statis- tics as a function of grey level quantization. Can. J Remote. Sens., 28(1):45–62, 2002

  7. [7]

    Generating images with perceptual similarity metrics based on deep networks

    Alexey Dosovitskiy and Thomas Brox. Generating images with perceptual similarity metrics based on deep networks. In NIPS, pages 658–666, 2016

  8. [8]

    Link the head to the beak: Zero shot learning from noisy text description at part precision

    Mohamed Elhoseiny, Yizhe Zhu, Han Zhang, and Ahmed Elgammal. Link the head to the beak: Zero shot learning from noisy text description at part precision. In IEEE CVPR, 2017

Show all 95 references
  1. [9]

    Structure-measure: A New Way to Evaluate Fore- ground Maps

    Deng-Ping Fan, Ming-Ming Cheng, Yun Liu, Tao Li, and Ali Borji. Structure-measure: A New Way to Evaluate Fore- ground Maps. In IEEE ICCV, pages 4548–4557, 2017

  2. [10]

    Enhanced-alignment Measure for Binary Foreground Map Evaluation

    Deng-Ping Fan, Cheng Gong, Yang Cao, Bo Ren, Ming- Ming Cheng, and Ali Borji. Enhanced-alignment Measure for Binary Foreground Map Evaluation. In IJCAI, pages 698–704, 2018

  3. [11]

    Rethinking RGB-D salient object detection: Mod- els, datasets, and large-scale benchmarks

    Deng-Ping Fan, Zheng Lin, Jia-Xing Zhao, Yun Liu, Zhao Zhang, Qibin Hou, Menglong Zhu, and Ming-Ming Cheng. Rethinking RGB-D salient object detection: Mod- els, datasets, and large-scale benchmarks. arXiv preprint arXiv:1907.06781, 2019

  4. [12]

    Salient objects in clutter: Bringing salient object detection to the foreground

    Deng-Ping Fan, Jiang-Jiang Liu, Shang-Hua Gao, Qibin Hou, Ali Borji, and Ming-Ming Cheng. Salient objects in clutter: Bringing salient object detection to the foreground. In ECCV, pages 1597–1604. Springer, 2018

  5. [13]

    Shifting more attention to video salient object detection

    Deng-Ping Fan, Wenguan Wang, Ming-Ming Cheng, and Jianbing Shen. Shifting more attention to video salient object detection. In IEEE CVPR, pages 8554–8564, 2019

  6. [14]

    Example-based synthesis of stylized facial animations

    Jakub Fi ˇser, Ondˇrej Jamriˇska, David Simons, Eli Shechtman, Jingwan Lu, Paul Asente, Michal Luk´aˇc, and Daniel S`ykora. Example-based synthesis of stylized facial animations. ACM TOG, 36(4):155, 2017

  7. [15]

    Learning style translation for the lines of a drawing

    William T Freeman, Joshua B Tenenbaum, and Egon C Pasz- tor. Learning style translation for the lines of a drawing. ACM TOG, 22(1):33–46, 2003

  8. [16]

    Theory of communication

    Dennis Gabor. Theory of communication. part 1: The anal- ysis of information. Journal of the Institution of Electrical Engineers-Part III: Radio and Communication Engineering, 93(26):429–441, 1946

  9. [17]

    Texture analysis using grey level run lengths

    Mary M Galloway. Texture analysis using grey level run lengths. NASA STI/Recon Technical Report N, 75, 1974

  10. [18]

    Composition-aided sketch-realistic portrait generation

    Fei Gao, Shengjie Shi, Jun Yu, and Qingming Huang. Composition-aided sketch-realistic portrait generation. arXiv preprint arXiv:1712.00899, 2017

  11. [19]

    Deepsim: Deep similarity for image quality assess- ment

    Fei Gao, Yi Wang, Panpeng Li, Min Tan, Jun Yu, and Yani Zhu. Deepsim: Deep similarity for image quality assess- ment. Neurocomputing, 257:104–114, 2017

  12. [20]

    Face sketch–photo synthesis and retrieval using sparse rep- resentation

    Xinbo Gao, Nannan Wang, Dacheng Tao, and Xuelong Li. Face sketch–photo synthesis and retrieval using sparse rep- resentation. IEEE TCSVT, 22(8):1213–1226, 2012

  13. [21]

    Face sketch synthesis algorithm based on E-HMM and selective ensemble

    Xinbo Gao, Juanjuan Zhong, Jie Li, and Chunna Tian. Face sketch synthesis algorithm based on E-HMM and selective ensemble. IEEE TCSVT, 18(4):487–496, 2008

  14. [22]

    Local face sketch synthesis learning

    Xinbo Gao, Juanjuan Zhong, Dacheng Tao, and Xuelong Li. Local face sketch synthesis learning. Neurocomputing, 71(10-12):1921–1930, 2008

  15. [23]

    Programmable style for NPR line draw- ing

    St ´ephane Grabli, Emmanuel Turquin, Fr ´edo Durand, and Franc ¸ois X Sillion. Programmable style for NPR line draw- ing. Rendering Techniques (Eurographics Symposium on Rendering), 2004

  16. [24]

    Statistical and structural approaches to texture

    Robert M Haralick et al. Statistical and structural approaches to texture. Proceedings of the IEEE, 67(5):786–804, 1979

  17. [25]

    Textu- ral features for image classification

    Robert M Haralick, Karthikeyan Shanmugam, et al. Textu- ral features for image classification. IEEE Transactions on Systems, Man, and Cybernetics, pages 610–621, 1973

  18. [26]

    Locality preserving projec- tions

    Xiaofei He and Partha Niyogi. Locality preserving projec- tions. In NIPS, pages 153–160, 2004

  19. [27]

    Image-to-image translation with conditional adver- sarial networks

    Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A Efros. Image-to-image translation with conditional adver- sarial networks. In IEEE CVPR, pages 1125–1134, 2017

  20. [28]

    A modified convolutional neural network for face sketch synthesis

    Licheng Jiao, Sibo Zhang, Lingling Li, Fang Liu, and Wen- ping Ma. A modified convolutional neural network for face sketch synthesis. PR, 76:125–136, 2018

  21. [29]

    Adap- tive representation-based face sketch-photo synthesis

    Jie Li, Xinye Yu, Chunlei Peng, and Nannan Wang. Adap- tive representation-based face sketch-photo synthesis. Neu- rocomputing, 269:152–159, 2017

  22. [30]

    Free-hand sketch synthesis with deformable stroke models

    Yi Li, Yi-Zhe Song, Timothy M Hospedales, and Shaogang Gong. Free-hand sketch synthesis with deformable stroke models. IJCV, 122(1):169–190, 2017

  23. [31]

    A nonlinear approach for face sketch synthesis and recognition

    Qingshan Liu, Xiaoou Tang, Hongliang Jin, Hanqing Lu, and Songde Ma. A nonlinear approach for face sketch synthesis and recognition. In IEEE CVPR , volume 1, pages 1005– 1010, 2005

  24. [32]

    Bayesian Tensor Inference for Sketch-Based Facial Photo Hallucination

    Wei Liu, Xiaoou Tang, and Jianzhuang Liu. Bayesian Tensor Inference for Sketch-Based Facial Photo Hallucination. In IJCAI, pages 2141–2146, 2007

  25. [33]

    How to evaluate foreground maps? In IEEE CVPR, pages 248–255, 2014

    Ran Margolin, Lihi Zelnik-Manor, and Ayellet Tal. How to evaluate foreground maps? In IEEE CVPR, pages 248–255, 2014

  26. [34]

    Modeling the shape of the scene: A holistic representation of the spatial envelope

    Aude Oliva and Antonio Torralba. Modeling the shape of the scene: A holistic representation of the spatial envelope. IJCV, 42(3):145–175, 2001

  27. [35]

    Superpixel-based face sketch–photo synthesis.IEEE TCSVT, 27(2):288–299, 2017

    Chunlei Peng, Xinbo Gao, Nannan Wang, and Jie Li. Superpixel-based face sketch–photo synthesis.IEEE TCSVT, 27(2):288–299, 2017

  28. [36]

    Multiple representations-based face sketch–photo synthesis

    Chunlei Peng, Xinbo Gao, Nannan Wang, Dacheng Tao, Xuelong Li, and Jie Li. Multiple representations-based face sketch–photo synthesis. IEEE TNNLS, 27(11):2201–2215, 2016

  29. [37]

    Measures and meta- measures for the supervised evaluation of image segmenta- tion

    Jordi Pont-Tuset and Ferran Marques. Measures and meta- measures for the supervised evaluation of image segmenta- tion. In IEEE CVPR, pages 2131–2138, 2013

  30. [38]

    Neural style transfer: a paradigm shift for image-based artistic ren- dering? In ACM NPAR, page 5, 2017

    Amir Semmo, Tobias Isenberg, and J ¨urgen D¨ollner. Neural style transfer: a paradigm shift for image-based artistic ren- dering? In ACM NPAR, page 5, 2017

  31. [39]

    Image information and visual quality

    Hamid R Sheikh and Alan C Bovik. Image information and visual quality. IEEE TIP, 15(2):430–444, 2006

  32. [40]

    An information fidelity criterion for image quality assess- ment using natural scene statistics

    Hamid R Sheikh, Alan C Bovik, and Gustavo De Veciana. An information fidelity criterion for image quality assess- ment using natural scene statistics. IEEE TIP, 14(12):2117– 2128, 2005

  33. [41]

    Lazy random walks for superpixel segmentation

    Jianbing Shen, Yunfan Du, Wenguan Wang, and Xuelong Li. Lazy random walks for superpixel segmentation. IEEE TIP, 23(4):1451–1462, 2014

  34. [42]

    Real-time superpixel seg- mentation by DBSCAN clustering algorithm

    Jianbing Shen, Xiaopeng Hao, Zhiyuan Liang, Yu Liu, Wen- guan Wang, and Ling Shao. Real-time superpixel seg- mentation by DBSCAN clustering algorithm. IEEE TIP , 25(12):5933–5942, 2016

  35. [43]

    Submodular trajectories for better motion segmentation in videos

    Jianbing Shen, Jianteng Peng, and Ling Shao. Submodular trajectories for better motion segmentation in videos. IEEE TIP, 27(6):2688–2700, 2018

  36. [44]

    An isotropic 3×3 image gradient operator

    Irvin Sobel. An isotropic 3×3 image gradient operator. Ma- chine Vision for Three-dimensional Scenes, pages 376–379, 1990

  37. [45]

    Real-time exemplar-based face sketch synthe- sis

    Yibing Song, Linchao Bao, Qingxiong Yang, and Ming- Hsuan Yang. Real-time exemplar-based face sketch synthe- sis. In ECCV, pages 800–813. Springer, 2014

  38. [46]

    A benchmark for automatic visual classification of clinical skin disease images

    Xiaoxiao Sun, Jufeng Yang, Ming Sun, and Kai Wang. A benchmark for automatic visual classification of clinical skin disease images. In ECCV, pages 206–222. Springer, 2016

  39. [47]

    Beginner’s Guide to Sketching

    Bomba Sylwia, Cai Rovina, Croes Brun, Gerard Justin, and Lewis Marisa. Beginner’s Guide to Sketching. 3dtotal Pub- lishing, 2015

  40. [48]

    NIMA: Neural image assessment

    Hossein Talebi and Peyman Milanfar. NIMA: Neural image assessment. IEEE TIP, 27(8):3998–4011, 2018

  41. [49]

    Face sketch synthesis and recognition

    Xiaoou Tang and Xiaogang Wang. Face sketch synthesis and recognition. In IEEE CVPR, pages 687–694, 2003

  42. [50]

    Face sketch recognition

    Xiaoou Tang and Xiaogang Wang. Face sketch recognition. IEEE TCSVT, 14(1):50–57, 2004

  43. [51]

    Facial Sketch Synthesis Using 2D Direct Combined Model-Based Face-Specific Markov Network

    Ching-Ting Tu, Yu-Hsien Chan, and Yi-Chung Chen. Facial Sketch Synthesis Using 2D Direct Combined Model-Based Face-Specific Markov Network. IEEE TIP , 25(8):3546– 3561, 2016

  44. [52]

    High- quality facial photo-sketch synthesis using multi-adversarial networks

    Lidan Wang, Vishwanath Sindagi, and Vishal Patel. High- quality facial photo-sketch synthesis using multi-adversarial networks. In IEEE FG, pages 83–90, 2018

  45. [53]

    Random sampling for fast face sketch synthesis

    Nannan Wang, Xinbo Gao, and Jie Li. Random sampling for fast face sketch synthesis. Pattern Recognition, 76:215–227, 2018

  46. [54]

    Evaluation on synthesized face sketches

    Nannan Wang, Xinbo Gao, Jie Li, Bin Song, and Zan Li. Evaluation on synthesized face sketches. Neurocomputing, 214:991–1000, 2016

  47. [55]

    Bayesian face sketch synthesis

    Nannan Wang, Xinbo Gao, Leiyu Sun, and Jie Li. Bayesian face sketch synthesis. IEEE TIP, 26(3):1264–1274, 2017

  48. [56]

    Heterogeneous image transformation

    Nannan Wang, Jie Li, Dacheng Tao, Xuelong Li, and Xinbo Gao. Heterogeneous image transformation. PRL, 34(1):77– 84, 2013

  49. [57]

    Transductive face sketch-photo synthesis

    Nannan Wang, Dacheng Tao, Xinbo Gao, Xuelong Li, and Jie Li. Transductive face sketch-photo synthesis. IEEE TNNLS, 24(9):1364–1376, 2013

  50. [58]

    A comprehensive survey to face hallucination

    Nannan Wang, Dacheng Tao, Xinbo Gao, Xuelong Li, and Jie Li. A comprehensive survey to face hallucination. IJCV, 106(1):9–30, 2014

  51. [59]

    Data-driven vs

    Nannan Wang, Mingrui Zhu, Jie Li, Bin Song, and Zan Li. Data-driven vs. model-driven: Fast face sketch synthesis. Neurocomputing, 2017

  52. [60]

    Semi-coupled dictionary learning with applications to image super-resolution and photo-sketch synthesis

    Shenlong Wang, Lei Zhang, Yan Liang, and Quan Pan. Semi-coupled dictionary learning with applications to image super-resolution and photo-sketch synthesis. In IEEE CVPR, pages 2216–2223, 2012

  53. [61]

    Dual-space linear dis- criminant analysis for face recognition

    Xiaogang Wang and Xiaoou Tang. Dual-space linear dis- criminant analysis for face recognition. In IEEE CVPR, vol- ume 2, pages II–II, 2004

  54. [62]

    Random sampling lda for face recognition

    Xiaogang Wang and Xiaoou Tang. Random sampling lda for face recognition. In IEEE CVPR, pages 259–265, 2004

  55. [63]

    Random sampling for subspace face recognition

    Xiaogang Wang and Xiaoou Tang. Random sampling for subspace face recognition. IJCV, 70(1):91–104, 2006

  56. [64]

    Face photo-sketch syn- thesis and recognition

    Xiaogang Wang and Xiaoou Tang. Face photo-sketch syn- thesis and recognition. IEEE TPAMI, 31(11):1955–1967, 2009

  57. [65]

    A universal image quality index

    Zhou Wang and Alan C Bovik. A universal image quality index. IEEE SPL, 9(3):81–84, 2002

  58. [66]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visibility to structural similarity. IEEE TIP, 13(4):600–612, 2004

  59. [67]

    Computer- generated pen-and-ink illustration

    Georges Winkenbach and David H Salesin. Computer- generated pen-and-ink illustration. In ACM SIGGRAPH , pages 91–100, 1994

  60. [68]

    Robust face recognition via sparse represen- tation

    John Wright, Allen Y Yang, Arvind Ganesh, S Shankar Sas- try, and Yi Ma. Robust face recognition via sparse represen- tation. IEEE TPAMI, 31(2):210–227, 2009

  61. [69]

    Ip102: A large-scale benchmark dataset for insect pest recognition

    Xiaoping Wu, Chi Zhan, Yu-Kun Lai, Ming-Ming Cheng, and Jufeng Yang. Ip102: A large-scale benchmark dataset for insect pest recognition. In IEEE CVPR, pages 8787–8796, 2019

  62. [70]

    Photo-sketch synthesis and recognition based on subspace learning

    Bing Xiao, Xinbo Gao, Dacheng Tao, Yuan Yuan, and Jie Li. Photo-sketch synthesis and recognition based on subspace learning. Neurocomputing, 73(4-6):840–852, 2010

  63. [71]

    Multi-channel weighted nuclear norm minimization for real color image denoising

    Jun Xu, Lei Zhang, David Zhang, and Xiangchu Feng. Multi-channel weighted nuclear norm minimization for real color image denoising. In IEEE ICCV, 2017

  64. [72]

    Gradient magnitude similarity deviation: A highly efficient perceptual image quality index

    Wufeng Xue, Lei Zhang, Xuanqin Mou, and Alan C Bovik. Gradient magnitude similarity deviation: A highly efficient perceptual image quality index. IEEE TIP, 23(2):684–695, 2014

  65. [73]

    Clinical skin lesion diagnosis using representations inspired by dermatologist criteria

    Jufeng Yang, Xiaoxiao Sun, Jie Liang, and Paul L Rosin. Clinical skin lesion diagnosis using representations inspired by dermatologist criteria. In IEEE CVPR, pages 1258–1266, 2018

  66. [74]

    Content-adaptive sketch por- trait generation by decompositional representation learning

    Dongyu Zhang, Liang Lin, Tianshui Chen, Xian Wu, Wen- wei Tan, and Ebroul Izquierdo. Content-adaptive sketch por- trait generation by decompositional representation learning. IEEE TIP, 26(1):328–339, 2017

  67. [75]

    End-to-end photo-sketch generation via fully convolutional representation learning

    Liliang Zhang, Liang Lin, Xian Wu, Shengyong Ding, and Lei Zhang. End-to-end photo-sketch generation via fully convolutional representation learning. In ACM ICMR, pages 627–634, 2015

  68. [76]

    FSIM: A feature similarity index for image quality assess- ment

    Lin Zhang, Lei Zhang, Xuanqin Mou, and David Zhang. FSIM: A feature similarity index for image quality assess- ment. IEEE TIP, 20(8):2378–2386, 2011

  69. [77]

    Com- positional model-based sketch generator in facial entertain- ment

    Mingjin Zhang, Jie Li, Nannan Wang, and Xinbo Gao. Com- positional model-based sketch generator in facial entertain- ment. IEEE TOC, 48(3):904–915, 2018

  70. [78]

    Markov random neural fields for face sketch synthesis

    Mingjin Zhang, Nannan Wang, Xinbo Gao, and Yunsong Li. Markov random neural fields for face sketch synthesis. In IJCAI, pages 1142–1148, 2018

  71. [79]

    Face sketch synthesis from coarse to fine

    Mingjin Zhang, Nannan Wang, Yunsong Li, Ruxin Wang, and Xinbo Gao. Face sketch synthesis from coarse to fine. In AAAI, 2018

  72. [80]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In IEEE CVPR, pages 586– 595, 2018

  73. [81]

    Robust face sketch style synthesis

    Shengchuan Zhang, Xinbo Gao, Nannan Wang, and Jie Li. Robust face sketch style synthesis. IEEE TIP, 25(1):220– 232, 2016

  74. [82]

    Face sketch synthesis from a single photo–sketch pair

    Shengchuan Zhang, Xinbo Gao, Nannan Wang, and Jie Li. Face sketch synthesis from a single photo–sketch pair. IEEE TCSVT, 27(2):275–287, 2017

  75. [83]

    Face sketch synthesis via sparse representation-based greedy search

    Shengchuan Zhang, Xinbo Gao, Nannan Wang, Jie Li, and Mingjin Zhang. Face sketch synthesis via sparse representation-based greedy search. IEEE TIP, 24(8):2466– 2477, 2015

  76. [84]

    Robust face sketch synthesis via generative adver- sarial fusion of priors and parametric sigmoid

    Shengchuan Zhang, Rongrong Ji, Jie Hu, Yue Gao, and Chia- Wen Lin. Robust face sketch synthesis via generative adver- sarial fusion of priors and parametric sigmoid. In IJCAI, pages 1163–1169, 2018

  77. [85]

    Face sketch synthesis by multidomain adversar- ial learning

    Shengchuan Zhang, Rongrong Ji, Jie Hu, Xiaoqiang Lu, and Xuelong Li. Face sketch synthesis by multidomain adversar- ial learning. IEEE TNNLS, 2018

  78. [86]

    Lighting and pose robust face sketch synthesis

    Wei Zhang, Xiaogang Wang, and Xiaoou Tang. Lighting and pose robust face sketch synthesis. In ECCV, pages 420–433. Springer, 2010

  79. [87]

    Coupled information-theoretic encoding for face photo-sketch recog- nition

    Wei Zhang, Xiaogang Wang, and Xiaoou Tang. Coupled information-theoretic encoding for face photo-sketch recog- nition. In IEEE CVPR, pages 513–520, 2011

  80. [88]

    Fast face sketch synthesis via kd-tree search

    Yuqian Zhang, Nannan Wang, Shengchuan Zhang, Jie Li, and Xinbo Gao. Fast face sketch synthesis via kd-tree search. In ECCV, pages 64–77. Springer, 2016

  81. [89]

    Jiaxing Zhao, Ren Bo, Qibin Hou, Ming-Ming Cheng, and Paul L. Rosin. FLIC: fast linear iterative clustering with active search. Computational Visual Media, 4(4):333–348, 2018

  82. [90]

    Contrast prior and fluid pyramid integration for RGBD salient object detection

    Jia-Xing Zhao, Yang Cao, Deng-Ping Fan, Ming-Ming Cheng, Xuan-Yi Li, and Le Zhang. Contrast prior and fluid pyramid integration for RGBD salient object detection. In IEEE CVPR, 2019

  83. [91]

    Edge-based network for salient object detection

    Jia-Xing Zhao, Jiang-Jiang Liu, Deng-Ping Fan, Jufeng Yang, and Ming-Ming Cheng. Edge-based network for salient object detection. In IEEE ICCV, 2019

  84. [92]

    Markov weight fields for face sketch synthesis

    Hao Zhou, Zhanghui Kuang, and Kwan-Yee K Wong. Markov weight fields for face sketch synthesis. In IEEE CVPR, pages 1091–1097, 2012

  85. [93]

    Deep graphical feature learning for face sketch synthesis

    Mingrui Zhu, Nannan Wang, Xinbo Gao, and Jie Li. Deep graphical feature learning for face sketch synthesis. InIJCAI, pages 3574–3580, 2017

  86. [94]

    A multilayer-based framework for online background subtraction with freely moving cameras

    Yizhe Zhu and Ahmed Elgammal. A multilayer-based framework for online background subtraction with freely moving cameras. In IEEE ICCV, 2017

  87. [95]

    Two stages of curve detection suggest two styles of visual com- putation

    Steven W Zucker, Allan Dobbins, and Lee Iverson. Two stages of curve detection suggest two styles of visual com- putation. Neural computation, 1(1):68–81, 1989

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.