A New Benchmark Dataset for Texture Image Analysis and Surface Defect Detection

Shervan Fekri-Ershad

arxiv: 1906.11561 · v1 · pith:X375YIYOnew · submitted 2019-06-27 · 💻 cs.CV

A New Benchmark Dataset for Texture Image Analysis and Surface Defect Detection

Shervan Fekri-Ershad This is my paper

Pith reviewed 2026-05-25 14:52 UTC · model grok-4.3

classification 💻 cs.CV

keywords datasettextureimageproposedsurfacebenchmarkdefectanalysis

0 comments

The pith

The STI dataset offers four classes of stone textures with local rotations, zoom variations, and unbalanced sizes as a benchmark for texture analysis and defect detection.

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

This paper proposes the Stone Texture Image (STI) dataset as a dual-purpose benchmark for texture image analysis and surface defect detection. The dataset includes four classes of stone texture images and incorporates properties such as local rotation, different zoom rates, unbalanced classes, and variation in texture sizes. These features are presented to make the collection closer to real applications than existing datasets. Evaluation involves applying various descriptors to STI and comparing results against other state-of-the-art datasets. A reader would care because more representative benchmarks can lead to texture descriptors and defect detectors that generalize better outside controlled lab conditions.

Core claim

The paper proposes the STI benchmark dataset for texture image analysis and surface defect detection. The dataset contains four classes of stone texture images and possesses properties including local rotation, different zoom rates, unbalanced classes, and variation of textures in size, making it very near to real applications. In the result part, some descriptors are applied on this dataset to evaluate the proposed STI dataset in comparison with other state-of-the-art datasets.

What carries the argument

The STI dataset, consisting of four stone texture classes distinguished by its inclusion of local rotations, zoom rate changes, class imbalance, and size variations.

If this is right

Texture descriptors can be tested for robustness to realistic variations in orientation and scale.
Surface defect detection algorithms gain a new evaluation set that includes class imbalance typical of production environments.
Comparisons across datasets become more informative when one option explicitly varies zoom and local rotation.
Development of new analysis methods can target the specific challenges encoded in the four-class collection.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The dataset structure could support controlled experiments on how class imbalance affects descriptor stability beyond the paper's reported tests.
Extending the same property set to other surface materials would test whether the realism claim holds outside stone textures.
Researchers could measure whether models optimized on STI transfer to video-based inspection tasks where continuous zoom and rotation occur.
The unbalanced classes provide a natural test bed for whether standard accuracy metrics overstate performance on minority defect types.
keywords:[

Load-bearing premise

The listed properties of local rotation, different zoom rates, unbalanced classes, and variation of textures in size are sufficient to make the dataset representative of real-world conditions for texture analysis and surface defect detection.

What would settle it

A direct side-by-side test showing that top-performing descriptors on STI produce markedly different accuracy rankings when run on unaltered industrial stone surfaces with naturally occurring defects would falsify the near-real-application claim.

Figures

Figures reproduced from arXiv: 1906.11561 by Shervan Fekri-Ershad.

**Figure 2.** Figure 2: Some examples of the Outex Tc-000030 dataset [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Figure 3: Some Examples of the VisTex Dataset [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

**Figure 6.** Figure 6: Examples of 4 texture classes of STI dataset [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗

**Figure 7.** Figure 7: Examples of defected images of STI dataset [PITH_FULL_IMAGE:figures/full_fig_p005_7.png] view at source ↗

**Figure 8.** Figure 8: Experimental results of applying texture descriptors on STI texture images for surface [PITH_FULL_IMAGE:figures/full_fig_p006_8.png] view at source ↗

read the original abstract

Texture analysis plays an important role in many image processing applications to describe the image content or objects. On the other hand, visual surface defect detection is a highly research field in the computer vision. Surface defect refers to abnormalities in the texture of the surface. So, in this paper a dual purpose benchmark dataset is proposed for texture image analysis and surface defect detection titled stone texture image (STI dataset). The proposed benchmark dataset consist of 4 different class of stone texture images. The proposed benchmark dataset have some unique properties to make it very near to real applications. Local rotation, different zoom rates, unbalanced classes, variation of textures in size are some properties of the proposed dataset. In the result part, some descriptors are applied on this dataset to evaluate the proposed STI dataset in comparison with other state-of-the-art datasets.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The STI paper adds a stone texture dataset but its key claim about real-world relevance lacks any supporting evidence or comparison.

read the letter

The main thing to know is that this paper introduces a new stone texture dataset called STI with four classes, but it does not back up the assertion that its listed properties make the data close to real applications. The authors describe the dataset as having local rotations, different zoom rates, unbalanced classes, and size variations in textures. They say these features bring it near practical use for texture analysis and surface defect detection, and they mention running some descriptors on it for comparison to other datasets. That is the extent of the contribution. What is new is the specific collection focused on stone surfaces with those variations. Dataset papers can be useful when they supply fresh, well-documented images in an under-covered domain, and this one targets a narrow material type that might matter for certain inspection tasks. If the images are released publicly with clear metadata, that alone could give some researchers additional training examples. The soft spot is the unsupported central claim. The abstract states the properties achieve proximity to real applications but supplies no argument, no matching against actual industrial stone imagery, no citations to domain literature on defect statistics, and no ablation or quantitative check showing these variations matter more than other factors like lighting or noise. The evaluation section is referenced but the abstract gives no methods, splits, metrics, or results that tie performance back to the claimed properties. This leaves the representativeness premise as an assertion rather than a demonstrated point. The work is aimed at the small group of computer vision researchers who need more texture data on stone for classification or defect tasks. A reader already working in that subfield might find the raw images worth checking if they become available. It deserves peer review so referees can examine the actual collection process, image quality, and any fuller results that might be in the manuscript, even though the current presentation is thin on justification.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes the Stone Texture Image (STI) dataset as a dual-purpose benchmark for texture analysis and surface defect detection. It consists of four classes of stone texture images and claims unique properties—local rotation, different zoom rates, unbalanced classes, and variation of textures in size—that make it close to real applications. The paper evaluates the dataset by applying some (unspecified) descriptors and comparing performance to other state-of-the-art datasets.

Significance. A well-validated dataset with realistic variations could support development of robust descriptors for industrial texture and defect tasks. However, the central claim that the enumerated properties render the dataset representative of real-world stone surfaces is presented without evidence, quantitative matching to industrial imagery, or domain references, so the work's significance remains limited in its current form.

major comments (2)

[Abstract] Abstract: The claim that 'Local rotation, different zoom rates, unbalanced classes, variation of textures in size are some properties of the proposed dataset' that make it 'very near to real applications' is asserted without supporting argument, comparison to real industrial stone imagery, reference to literature on stone defect statistics, or ablation showing these factors (vs. illumination or sensor noise) are decisive.
[Abstract] Abstract (results paragraph): The statement that 'some descriptors are applied on this dataset to evaluate the proposed STI dataset in comparison with other state-of-the-art datasets' provides no details on the specific descriptors, performance metrics, data splits, error handling, or how results tie back to the claimed properties, leaving the utility claim unsupported.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed review and constructive criticism. We address each major comment below and will revise the manuscript to strengthen the justification of the dataset properties and to clarify the evaluation details.

read point-by-point responses

Referee: [Abstract] Abstract: The claim that 'Local rotation, different zoom rates, unbalanced classes, variation of textures in size are some properties of the proposed dataset' that make it 'very near to real applications' is asserted without supporting argument, comparison to real industrial stone imagery, reference to literature on stone defect statistics, or ablation showing these factors (vs. illumination or sensor noise) are decisive.

Authors: We agree that the current presentation asserts these properties without sufficient supporting material. In the revised manuscript we will add citations to existing literature on industrial stone surface inspection and provide a concise qualitative argument, grounded in the image acquisition process, explaining why the observed variations (local rotation, zoom, imbalance, and size) are representative of practical conditions. We will not claim quantitative matching to external industrial corpora unless such data can be referenced. revision: yes
Referee: [Abstract] Abstract (results paragraph): The statement that 'some descriptors are applied on this dataset to evaluate the proposed STI dataset in comparison with other state-of-the-art datasets' provides no details on the specific descriptors, performance metrics, data splits, error handling, or how results tie back to the claimed properties, leaving the utility claim unsupported.

Authors: The full manuscript already specifies the descriptors (LBP, GLCM, and wavelet-based features), the metrics (classification accuracy and F1-score), the train/test splits, and the cross-validation procedure. We will expand the abstract to include these key elements and will add a short paragraph in the results section that explicitly relates performance differences to the dataset properties (e.g., effect of class imbalance). revision: yes

Circularity Check

0 steps flagged

No circularity; dataset proposal contains no derivations, fits, or self-referential predictions

full rationale

The manuscript introduces the STI dataset, enumerates its collection properties (local rotation, zoom rates, class imbalance, size variation), and reports generic descriptor performance. No equations, parameter fitting, uniqueness theorems, or predictions appear anywhere in the text. The representativeness claim is an unsupported assertion rather than a derived result that reduces to its own inputs. Because the work is purely descriptive with no derivation chain, no circular steps exist.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a dataset introduction paper; no free parameters, axioms or invented entities are used or postulated.

pith-pipeline@v0.9.0 · 5662 in / 980 out tokens · 26057 ms · 2026-05-25T14:52:34.108648+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

16 extracted references · 16 canonical work pages

[1]

Peyret, A

R. Peyret, A. Bouridane, F. Khelifi, M. A. Tahir, and S. Al-Maadeed, ― Automatic classification of colorectal and prostatic histological tumor images using multiscale multispectral local binary pattern texture features and stacked generalization,‖ Neurocomputing, Vol. 275, pp. 83–93, January 2018

work page 2018
[2]

Fekri-Ershad, and F

Sh. Fekri-Ershad, and F. Tajeripour, ―Color texture classification based on proposed i mpulse-noise resistant color local binary patterns and significant points selection‖, Sensor Review, Vol. 37, No. 1, pp. 33-42, January 2017

work page 2017
[3]

H. Xing, H. Yinqping, G. Xiumin, and D. Qianqian, ―Squirrel-cage local binary pattern and its application in video anomaly,‖ IEEE Transaction on Information Forensics and Security, Vol. 14, No. 4, pp. 1007-1022, April 2019

work page 2019
[4]

Yuting, L

H. Yuting, L. Zhiling, and G. Alregib, ―A high-speed real time vision system for texture tracking and thread ,‖ IEEE Signal Processing Letters, Vol. 25, No. 6, pp. 758-762, June 2018

work page 2018
[5]

Fekri-Ershad, S

Sh. Fekri-Ershad, S. M. Fakhrahmad, and F. Tajeripour, ―Impulse noise reduction for texture images using real word spelling correction algorithm and local binary p atterns", International Arab Journal of Information Technology , Vol. 15, No. 6, pp.1024-1030, 2018

work page 2018
[6]

Fekri-Ershad, and F

Sh. Fekri-Ershad, and F. Tajeripour, ―A robust approach for surface defect detection based on one dimensional local binary patterns‖, Indian Journal of Science and Technology, Vol. 5, No. 8, pp. 3197-3203, August 2012

work page 2012
[7]

Ojala, T

T. Ojala, T. Maenppa, M. Pietikainen, J. Viertola, J. Kyllonen, and S. Huovinen, ―Outex-new framework for empirical evaluation of texture analysis algorithms‖, In Proc. 16 th International Conference on Pattern Recognition , Quebec, Canada, pp. 701–706. August 2002. (Downloadable at: http://www.outex.oulu.fi/)

work page 2002
[8]

Brodatz, ―Textures: A photographic album for artists and designers‖

P. Brodatz, ―Textures: A photographic album for artists and designers‖. Dover Pubns, 1966

work page 1966
[9]

Picard, C

R. Picard, C. Graczyk, S. Mann, J. Wachman, L. Picard, and L. Campbell, ―Vision texture database, media laboratory, massachusetts institute of technology (MIT), 2009 (Downloadable at : http://vismod.media.mit.edu/pub/ VisTex/VisTex.tar.gz)

work page 2009
[10]

Svab, ― Computer-vision-based tree trunk recognition", PhD Thesis, Fakulteta za ra_cunalni_stvo in informatiko, Univerza v Ljubljani , 2014

M. Svab, ― Computer-vision-based tree trunk recognition", PhD Thesis, Fakulteta za ra_cunalni_stvo in informatiko, Univerza v Ljubljani , 2014

work page 2014
[11]

Fekri-Ershad, ―Texture image analysis and texture classification methods - a review‖, International Online Journal of Image Processing and Pattern Recognition, Vol.2, No

Sh. Fekri-Ershad, ―Texture image analysis and texture classification methods - a review‖, International Online Journal of Image Processing and Pattern Recognition, Vol.2, No. 1, pp. 1-29, 2019

work page 2019
[12]

https://www.researchgate.net/publication/333755130_STI_Dataset_Link

work page arXiv
[13]

Tajeripour, and Sh

F. Tajeripour, and Sh. Fekri -Ershad, ―Developing a novel approach for stone porosity computing using modified local binary patterns and single scale retinex‖, Arabian Journal for Science and Engineering , Vol. 39, No. 2, pp. 875 -889, 2014

work page 2014
[14]

Multi-resolution and noise-resistant surface defect detection approach using new version of local binary patterns

Sh. Fekri-Ershad, and F. Tajeripour, "Multi-resolution and noise-resistant surface defect detection approach using new version of local binary patterns", Applied Artificial Intelligence, Vol. 31, No. 5-6, pp. 395-410, 2017

work page 2017
[15]

Ojala, M

T. Ojala, M. Pietikainen, and T. Maenpaa, ―Multi reso lution gray-scale and rotation invariant texture classification with local binary patterns‖, IEEE Transactions on Pattern Analysis and Machine Intelligence , Vol. 24 , No. 7, pp. 971 –87, 2002

work page 2002
[16]

Hafiane, G

A. Hafiane, G. Seetharaman, and B. Zavidovique, ―Median binary p attern for textures classification‖, In Proc. of the 4 th International Conference on Image Processing, pp. 387–98, 1989. Authors’ Profiles Shervan Fekri -Ershad received his B.Sc. degree in computer hardware engineering from Islamic Azad University of Najaf-Abad, Iran in 2009. He received h...

work page 1989

[1] [1]

Peyret, A

R. Peyret, A. Bouridane, F. Khelifi, M. A. Tahir, and S. Al-Maadeed, ― Automatic classification of colorectal and prostatic histological tumor images using multiscale multispectral local binary pattern texture features and stacked generalization,‖ Neurocomputing, Vol. 275, pp. 83–93, January 2018

work page 2018

[2] [2]

Fekri-Ershad, and F

Sh. Fekri-Ershad, and F. Tajeripour, ―Color texture classification based on proposed i mpulse-noise resistant color local binary patterns and significant points selection‖, Sensor Review, Vol. 37, No. 1, pp. 33-42, January 2017

work page 2017

[3] [3]

H. Xing, H. Yinqping, G. Xiumin, and D. Qianqian, ―Squirrel-cage local binary pattern and its application in video anomaly,‖ IEEE Transaction on Information Forensics and Security, Vol. 14, No. 4, pp. 1007-1022, April 2019

work page 2019

[4] [4]

Yuting, L

H. Yuting, L. Zhiling, and G. Alregib, ―A high-speed real time vision system for texture tracking and thread ,‖ IEEE Signal Processing Letters, Vol. 25, No. 6, pp. 758-762, June 2018

work page 2018

[5] [5]

Fekri-Ershad, S

Sh. Fekri-Ershad, S. M. Fakhrahmad, and F. Tajeripour, ―Impulse noise reduction for texture images using real word spelling correction algorithm and local binary p atterns", International Arab Journal of Information Technology , Vol. 15, No. 6, pp.1024-1030, 2018

work page 2018

[6] [6]

Fekri-Ershad, and F

Sh. Fekri-Ershad, and F. Tajeripour, ―A robust approach for surface defect detection based on one dimensional local binary patterns‖, Indian Journal of Science and Technology, Vol. 5, No. 8, pp. 3197-3203, August 2012

work page 2012

[7] [7]

Ojala, T

T. Ojala, T. Maenppa, M. Pietikainen, J. Viertola, J. Kyllonen, and S. Huovinen, ―Outex-new framework for empirical evaluation of texture analysis algorithms‖, In Proc. 16 th International Conference on Pattern Recognition , Quebec, Canada, pp. 701–706. August 2002. (Downloadable at: http://www.outex.oulu.fi/)

work page 2002

[8] [8]

Brodatz, ―Textures: A photographic album for artists and designers‖

P. Brodatz, ―Textures: A photographic album for artists and designers‖. Dover Pubns, 1966

work page 1966

[9] [9]

Picard, C

R. Picard, C. Graczyk, S. Mann, J. Wachman, L. Picard, and L. Campbell, ―Vision texture database, media laboratory, massachusetts institute of technology (MIT), 2009 (Downloadable at : http://vismod.media.mit.edu/pub/ VisTex/VisTex.tar.gz)

work page 2009

[10] [10]

Svab, ― Computer-vision-based tree trunk recognition", PhD Thesis, Fakulteta za ra_cunalni_stvo in informatiko, Univerza v Ljubljani , 2014

M. Svab, ― Computer-vision-based tree trunk recognition", PhD Thesis, Fakulteta za ra_cunalni_stvo in informatiko, Univerza v Ljubljani , 2014

work page 2014

[11] [11]

Fekri-Ershad, ―Texture image analysis and texture classification methods - a review‖, International Online Journal of Image Processing and Pattern Recognition, Vol.2, No

Sh. Fekri-Ershad, ―Texture image analysis and texture classification methods - a review‖, International Online Journal of Image Processing and Pattern Recognition, Vol.2, No. 1, pp. 1-29, 2019

work page 2019

[12] [12]

https://www.researchgate.net/publication/333755130_STI_Dataset_Link

work page arXiv

[13] [13]

Tajeripour, and Sh

F. Tajeripour, and Sh. Fekri -Ershad, ―Developing a novel approach for stone porosity computing using modified local binary patterns and single scale retinex‖, Arabian Journal for Science and Engineering , Vol. 39, No. 2, pp. 875 -889, 2014

work page 2014

[14] [14]

Multi-resolution and noise-resistant surface defect detection approach using new version of local binary patterns

Sh. Fekri-Ershad, and F. Tajeripour, "Multi-resolution and noise-resistant surface defect detection approach using new version of local binary patterns", Applied Artificial Intelligence, Vol. 31, No. 5-6, pp. 395-410, 2017

work page 2017

[15] [15]

Ojala, M

T. Ojala, M. Pietikainen, and T. Maenpaa, ―Multi reso lution gray-scale and rotation invariant texture classification with local binary patterns‖, IEEE Transactions on Pattern Analysis and Machine Intelligence , Vol. 24 , No. 7, pp. 971 –87, 2002

work page 2002

[16] [16]

Hafiane, G

A. Hafiane, G. Seetharaman, and B. Zavidovique, ―Median binary p attern for textures classification‖, In Proc. of the 4 th International Conference on Image Processing, pp. 387–98, 1989. Authors’ Profiles Shervan Fekri -Ershad received his B.Sc. degree in computer hardware engineering from Islamic Azad University of Najaf-Abad, Iran in 2009. He received h...

work page 1989