Comparison of Contour Detection Methods in Images on the Example of Photos with Road Surface Damage

Due to the development of technology, there are currently many information systems designed to detect various damages on the road surface. However, these systems work based on certain image processing algorithms, which include methods for contour detection. Each algorithm has its own features that can affect the operation of the information device. For a more accurate assessment of each method, a comparative analysis of its work is necessary. This article discusses the following algorithm for contour detection in images: Line Detection (LD), Point Detection (PD), Kirsch Edge Detection (KED), Canny Edge Detection (CED), Marr-Hildreth Edge Detection (MHED). The average image processing time is chosen as an indicator of the evaluation of algorithms. To calculate it, a series of experiments is carried out for 5 images, each of which is processed 10 times (a total of 50 tests for one algorithm). The experiment is performed for image of 7 different sizes (in pixels): 60000, 80000, 120000, 240000, 480000, 720000, 960000. It is established that LD and PD have linear time complexity (O(n)), KED has O(nlog(n)), while CED and MHED have quadratic complexity (O(n
²
)). MHED, CED and KED highlight clearer contours compared to LD and PD.