Advanced Digital Image Processing & Computer Vision

Unveiling the Unseen: Pioneering Intelligent Vision Systems for Precision Analysis, From Industrial Quality Control to Complex Image Interpretation.
Real-time AI powered by GPU architectures for scalable processing, edge inference, and visual AI acceleration

Subtopics:
• Unsupervised image clustering
• Real-time pattern recognition
• Faster R-CNN for high-res defect analysis
• Edge + cloud GPU pipelines
• SEM image analysis (as sub-case)

My Contribution: My research pushes the boundaries of digital image processing and computer vision by developing sophisticated algorithms for tasks such as robust image denoising, precise contour detection, and advanced defect identification. I integrate machine learning, deep learning, and unsupervised clustering to enable highly accurate and automated visual analysis across diverse applications, including high-variability images in manufacturing metrology and the nuanced interpretation of visual data for broad scientific and industrial uses.

Publications & Patents:

Unsupervised machine learning based SEM image denoising for robust contour detection

K. Ahi
International Conference on Extreme Ultraviolet Lithography 2021 11854, 88-102
2021
Cited by: 14
[DOI Placeholder]
[Link Placeholder]
Highlight: This work introduces an unsupervised machine learning approach for highly effective denoising of SEM images, critical for achieving robust contour detection in semiconductor manufacturing.
Wafer image denoising and contour extraction for manufacturing process calibration

K. Ahi
US Patent App. 17/823,741
2023
Cited by: 2
[DOI Placeholder]
[Link Placeholder]
Highlight: This patent describes a system and method for enhancing manufacturing process calibration through advanced image denoising and precise contour extraction on wafer images.
AI-Powered Defect Detection using Deep Learning: A Pattern-Agnostic Faster R-CNN Approach for SEM Images with GPU Acceleration

K. Ahi
2025 36th Annual SEMI Advanced Semiconductor Manufacturing Conference (ASMC…)
2025
Cited by: (New publication, citations building)
[DOI Placeholder]
[Link Placeholder]
Highlight: This paper introduces a robust deep learning-based Faster R-CNN for pattern-agnostic defect detection in SEM images, significantly advancing automated quality control in semiconductor fabrication.
AI-Powered Anomaly Detection: A Robust, Pattern-Agnostic Faster R-CNN with Scalable GPU-Accelerated Deep Learning for High-Fidelity Computer Vision Defect Identification (mAP…

K. Ahi
IEEE ASMC
2025
Cited by: (New publication, citations building)
[DOI Placeholder]
[Link Placeholder]
Highlight: This research details a scalable, GPU-accelerated deep learning framework for highly accurate, pattern-agnostic anomaly and defect detection in computer vision applications.
Contour extraction of images with selection-based auto tuning

K. Ahi
US Patent App. 18/476,559
2025
Cited by: (New patent application)
[DOI Placeholder]
[Link Placeholder]
Highlight: This patent application presents an innovative method for automated and highly adaptable contour extraction from images, enhancing precision in various imaging applications.
Methods and systems for developing silhouettes, traditional-looking silhouettes, sunset silhouettes, removing shadows and objects, and improving the quality of sunset in images

K. Ahi
US Patent App. 17/163,568
2023
Cited by: (New patent application)
[DOI Placeholder]
[Link Placeholder]
Highlight: This patent application offers novel methods and systems for sophisticated image manipulation, including artistic silhouette generation and quality enhancement for diverse visual effects.
Unsupervised, Scalable Clustering, Pattern Recognition, and Graphics Processing Unit (GPU)-Accelerated Contour Extraction from Challenging High-Variability Images Using Edge…

K. Ahi
US Patent App. PCT/US2025/027,065
2025
Cited by: 4
[DOI Placeholder]
[Link Placeholder]
Highlight: This patent pending work describes a system for high-performance, GPU-accelerated image processing, enabling robust contour extraction and pattern recognition in complex and high-variability visual data.