EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Deep-Petri: A Decision-Intelligent Petri Net and Deep Learning Framework for Adaptive Image Enhancement

Hani M. Ibrahem, Ameer El-Sayed

Journal of Intelligent Decision Making and Information Science · 17.09.2026 · 3 (12s) · 2777-2801

DOI: 10.17762/jidmis.v3.3705 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 17.09.2026. Verfasst von Hani M. Ibrahem und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Image contrast enhancement requires intelligent decisions regarding which image information should be enhanced, preserved, fused, or suppressed while maintaining structural fidelity and avoiding noise amplification and over-enhancement. This paper proposes Deep-Petri, a decision-intelligent computational framework that integrates Petri-net workflow modeling with deep learning for adaptive and robust image enhancement. The framework organizes enhancement as a structured decision process across three interconnected phases: preprocessing, attention-guided fusion, and post-processing. During preprocessing, multi-scale convolutional feature extraction, residual learning, and Modified Fast Gray Level Grouping (MFGLG)-inspired processing identify and enhance informative contrast patterns. In the fusion phase, an attention mechanism performs adaptive feature-weighting decisions between the original and preprocessed images, assigning greater importance to regions with superior contrast and texture characteristics. The Petri-net model provides an interpretable decision-control layer in which image states are represented as places and processing operations as transitions, enabling dependency-aware execution, synchronization, parallelism, and controlled progression through the enhancement workflow. Post-processing subsequently applies denoising, edge-preserving filtering, and refinement to suppress artifacts while retaining structural information. The model is trained using a multi-objective loss combining L1, Structural Similarity Index Measure (SSIM), and perceptual losses. Experimental evaluation demonstrates that Deep-Petri achieves PSNR values of 21.98–24.45 dB and SSIM values of 0.704–0.878 across the evaluated test images, consistently providing competitive or superior performance compared with Histogram Equalization (HE), Contrast Limited Adaptive Histogram Equalization (CLAHE), Bi-Histogram Equalization (BHE), and Dynamic Histogram Equalization (DHE). For example, on the Cartoon image, Deep-Petri achieves 24.31 dB PSNR and 0.851 SSIM, compared with 22.92 dB and 0.838 for CLAHE, while on the Medical image it reaches 22.03 dB and 0.878, respectively. These results demonstrate that combining formal Petri-net decision control with adaptive deep feature learning provides an interpretable and effective decision-intelligent framework for high-quality image enhancement.

Zitieren

APA

Ibrahem, H. M., & El-Sayed, A. (2026). Deep-Petri: A Decision-Intelligent Petri Net and Deep Learning Framework for Adaptive Image Enhancement. Journal of Intelligent Decision Making and Information Science, 3(12s), 2777-2801. https://doi.org/10.17762/jidmis.v3.3705

Vancouver

Ibrahem HM, El-Sayed A. Deep-Petri: A Decision-Intelligent Petri Net and Deep Learning Framework for Adaptive Image Enhancement. Journal of Intelligent Decision Making and Information Science. 2026;3(12s):2777-2801. doi:10.17762/jidmis.v3.3705

BibTeX
@article{Ibrahem2026,
  author = {Ibrahem, Hani M. and El-Sayed, Ameer},
  title = {Deep-Petri: A Decision-Intelligent Petri Net and Deep Learning Framework for Adaptive Image Enhancement},
  journal = {Journal of Intelligent Decision Making and Information Science},
  year = {2026},
  volume = {3},
  number = {12s},
  pages = {2777--2801},
  doi = {10.17762/jidmis.v3.3705}
}
Kostenlos starten →