This paper deals with the issue of comparing selected methods for edge detection in digital images under conditions of different types of noise. The motivation for edge detection research is introduced and the broader context in which the problem of edge detection is situated is presented. The process by which edges in images are formed, their potential shapes, and the properties of their first and second derivatives are described. The mathematical methods used for edge detection, and their implementation through the Sobel, Prewitt and Laplace operators and the Canny algorithm are outlined. The images on which the edge detection methods were applied in their original form and the impact of the selected types of noise on them are described, of which speckle, Gaussian, salt-and-pepper and Poisson noise were selected for testing. The methods' implementations in code are described, and the results are presented graphically and in tabular form in the percentage form using the SSIM metric. The results showed that, according to the chosen metric, the Canny algorithm was the most successful, followed by the Sobel and Prewitt operators with similar results, while the Laplace operator was found to be the least effective. From a visual aspect, the most successful results were generated by the Sobel and Prewitt operators, with the Prewitt operator having an advantage. The Laplace operator produced good results in noise-free images but faced significant difficulties in the presence of noise. The Canny algorithm generated the least visually appealing images in most of the tested conditions.