Edge Detection Method Driven by Knowledge-Based Neighborhood Rules

Edge detection is a fundamental process, and therefore there are still demands to improve its efficiency and computational complexity. This study proposes a knowledge-based edge detection method to meet this requirement by introducing a set of knowledge-based rules. The methodology to derive the rules is based on the observed continuity properties and the neighborhood characteristics of the edge pixels, which are expressed as simple arithmetical operations to improve computational complexity. The results show that the method has an advantage over the gradient-based methods in terms of performance and computational load. It is appropriately four times faster than Canny method and shows superior performance compared to the gradient-based methods in general. Furthermore, the proposed method provides robustness to effectively identify edges at the corners. Due to its light computational requirement and inherent parallelization properties, the method would be also suitable for hardware implementation on field-programmable gate arrays (FPGA).