Abstract: Modern information technologies have significantly expanded the capabilities of image and video processing, which has found wide application in medicine, transport, security, and other industries. Of particular importance are mathematical models, in particular random processes, which allow for efficient modeling of natural noise, image reconstruction, and segmentation. Markov random fields (MRF), Kalman filters, and Bayesian methods play a key role in noise filtering and video analysis. The paper analyzes the existing methods of processing images and videos based on random processes. The application of Markov random fields for modeling dependencies between pixels and the use of the Bayesian approach in segmentation and noise reduction are considered. Special attention is paid to video compression methods, which allow you to optimize the amount of data without significant loss of quality, in particular by using the discrete cosine transform (DCT). The study confirms the effectiveness of using random processes in noise filtering, image reconstruction, and video analysis tasks, and highlights the prospects for integrating machine learning to improve data processing accuracy. The results obtained can be helpful in developing more reliable and productive algorithms that will find applications in medicine, robotics, security systems, and multimedia technologies.