Date: 2024.05.09 / 4pm
Place: ASTC 615 (첨단관 615호)
Title: Effective Edge Detection for Noisy Color Images
Speaker: Prof. Seongjai Kim (Department of Mathematics and Statistics, Mississippi State University)
Abstract:
The Canny algorithm is effective for the edge detection for various gray images, although it is sensitive to noise as most other edge detection algorithms are.
The noise removal step often weakens not only noise but also the edge strength.
This article proposes an innovative denoising operator called the reverse-transition weighting (RTW) filter, which can suppress noise without weakening the edge strength.
The RTW filter is analyzed for its stability and adopted for the noise removal step of the Canny algorithm, replacing the conventional Gaussian smoothing filter.
In order for the Canny algorithm to be applied for color images, we also consider and compare gradient-fusion methods which combine the RGB gradients into one.
The structure tensor method shows satisfactory properties for gradient-fusion.
Our goal is to formulate a robust edge detection algorithm for color images, particularly for heavily noisy images.
Various examples are given to show the effectiveness of the new RTW filter and the structure tensor method for gradient-fusion.
Organizer: Eun-Jae Park