Edge Detection in Image Processing

24 May 2024
14:00 - 15:00
B312

Dr. İlktan Ar

Kadir Has University

Abstract: Edge detection works by identifying discontinuities in brightness within an image. Various techniques exist to pinpoint these points, which are important for image segmentation and data extraction in fields such as image processing, computer vision, and machine vision. This problem has regained significance due to its practical applications in self-driving cars, augmented reality, image colorization, and medical imaging. This seminar will begin with examining the mathematical foundations of edges, including an analysis of edge types such as step, ramp, and more. The next step is to review traditional edge detection techniques based on gradients, including Prewitt and Sobel operators. This is followed by exploring advanced edge detection techniques such as Laplacian of Gaussian (LoG) and Canny. Subsequently, studies employing deep convolutional neural networks (DCNN) for edge detection, such as HED and RED-NET, are discussed. Finally, edge detection’s challenges and future directions are addressed by emphasizing performance metrics.

Speaker Biography: İlktan Ar obtained his B.Sc. degree in Computer Engineering from Kadir Has University (KHAS) in 2004. Following his undergraduate studies, he began working as a research assistant at KHAS. Concurrently, he pursued further education at Yıldız Technical University (YTU), where he completed his M.Sc. in Computer Engineering in 2007. He earned his Ph.D. in Computer Engineering from Gebze Technical University (GTU) in 2015. During his doctoral study, he was a member of the GTU-VisLab (Computer Vision Laboratory) and was funded by the TUBITAK-2211 National PhD Fellowship Program. Since January 2023, he has been a visiting scholar in the Computer Engineering Department at Kadir Has University. His research interests include computer vision, image processing, machine learning, pattern recognition, and industrial inspection.

Applications of Bayesian Reinforcement Learning in the Control Domain