Integrity Review Report: Robust and fast QR code images deblurring via local maximum and minimum intensity prior

Academic fraud report · Geng Detector

Summary

This report assesses the 2024 paper by Bin Xu, Rushi Jin, Jinhua Li, Bo Zhang, and Kai Liu, published in The Visual Computer (DOI: 10.1007/s00371-024-03272-y). The paper proposes an algorithm for deblurring QR code images using local maximum and minimum intensity priors. The overall verdict is that no clear indicators of academic fraud were identified within the scope of text-based analysis. Specifically, the publication timeline is internally consistent, with 2023 references aligning with the January 2024 acceptance date. The experimental data in Table 1 (runtime comparisons of 74.31s, 437.50s, 408.24s, 108.04s, and 36.59s) shows a natural decimal distribution without suspicious rounding patterns. No fabricated citations or anachronistic tool usage (Zxing 3.5.1, WeChat 8.0.32, Matlab) were detected. The primary limitation of this review is the absence of raw high-resolution image files, preventing pixel-level inspection of reused or manipulated figures (Figures 2, 8, 9, 11, 12). Confidence is moderate; image-related concerns cannot be excluded without original assets, but no text-level red flags emerged.

Verdict

✅ Cleared (within the limits of text-based analysis). No systematic indicators of academic misconduct were detected in the submitted material. Figure-level image reuse or photo-manipulation checks could not be completed due to the lack of original high-resolution image files.

Key findings

Timeline consistency: References [32], [33], [34] are all 2023 publications, consistent with the 8 January 2024 acceptance date and 27 February 2024 online publication. Tools referenced (Zxing 3.5.1, WeChat 8.0.32, Matlab functions) are temporally plausible.
Data plausibility in Table 1: Runtime values (Pan [15]: 74.31s, Pan [16]: 437.50s, Yan [12]: 408.24s, Chen [17]: 108.04s, Ours: 36.59s) show natural trailing-digit distribution (1, 0, 4, 4, 9) without signs of artificial fabrication or excessive rounding.
Experiment scale is standard for the field: 880 synthetic + 448 nonlinear-movement blurred images, with blur angles spanning 0°–90° and lengths 20–190, representing a typical benchmark setup.
Formula rendering anomaly (non-fraud): Page 8812, equations (6) and (8) display garbled text (e.g., λ z∈Nk I η y+ ◦ s 2 ≤ −z ≥ k(z)), attributable to PDF text-extraction / OCR errors rather than authorial misconduct; surrounding equations (1)–(5) are logically coherent.
No textual indicators of figure reuse, cropping artifacts, or image splicing could be evaluated because only extracted text was available.

Evidence highlights

DOI: 10.1007/s00371-024-03272-y
Journal: The Visual Computer (2024) 40:8809–8823
Table 1 runtime figures: 74.31 / 437.50 / 408.24 / 108.04 / 36.59 (seconds)
Dataset sizes cited: 880 synthetic blurred images + 448 nonlinear-movement images
Reference recency cutoff: predominantly ≤2023, matching acceptance date
Page 8812 equations (6) and (8) flagged as PDF extraction artifacts, not author error

Notes

Detection scope is limited: The reviewer had access only to PDF-extracted text, not the original image files. Conclusions about image manipulation (rotation/reuse, Photoshop splicing) cannot be drawn from this analysis alone.
Recommended follow-up: For full validation, request the Matlab source code and the 880-image synthetic dataset from the authors to enable reproducibility testing of the claimed 36.59-second runtime.
Disclaimer: This report was AI-assisted and is intended for academic discussion only. Final determinations of misconduct require formal institutional investigation. False positives and false negatives are possible.