Published Paper-2022: A Comparison of The Impact of Edge Detection Operators on The Extraction of Corner Points
Abstract:
Edge detection is an essential pre-processing procedure involved in most digital photogrammetry applications. Edge detection captures the object's physical boundaries, a crucial property of objects in the image. The amount of information provided by the edge detection operator is a crucial factor for the extraction of Corner points. Each edge map has a unique amount of information that differs depending on the used edge detection operator. Edge detection operators get optimized to solve a tremendous number of issues such as noise, edge structure, edge orientation, and computation time. Because of these issues and others, there is no single operator that fits all kinds of applications and purposes. So, it is a critical task to measure the impact of each edge detection operator on the extraction of corner points and choose the operator that yields a precise localization of corner points. This study evaluates the impact of the edge detection operator and measures its effect on the corner point localization. That is through an empirical methodology of comparatively analyzing the results of multiple edge detection operators. Implicitly, this study proposes a method to evaluate the general performance of the used edge detection operators. The proposed method is novel and evaluates edge detection operators through a specific task. In this task, the algorithm of corner point extraction considers the output of the edge detection operator as an intermediate input for the process of corner points detection. The edge detection operators to be used here are Canny, Sobel, Prewitt, Roberts, and Laplacian of Gaussian (LOG). The study concluded that the edge detection operator has a significant impact on the extraction of corner points. Also, the specific use of an edge detection operator depends on the desired application and the image under investigation. The Canny edge detection operator showed an outstanding performance.
Full text :
Comments
Post a Comment