Background subtraction (BS) is a key component in image processing, particularly in video surveillance where accurate motion detection is essential. The Codebook and the Gaussian Mixture Models (GMM) are the most widely adopted methods for BS due to their robustness in handling dynamic environments. However, while many extensions have been introduced to improve accuracy, these enhancements often increase computational costs, thereby limiting their suitability for real-time applications. To address this challenge, we present a real-time parallel implementation of Codebook method and GMM using Open Multiprocessing (OpenMP). The implementation was tested on two platforms: an 11th Gen Intel Core i5-11400H CPU (6 cores/12 threads, 8 GB DDR4) and an NVIDIA Jetson Nano 2 GB (quad-core ARM Cortex-A57, 2 GB LPDDR4). On the Intel CPU, the optimized version achieved 90 FPS for GMM and 70 FPS for Codebook at Full-HD resolution, while on the Jetson Nano, obtained results are about 13 FPS for GMM and 4 FPS for Codebook for Full-HD resolution. These results exceed prior CPU-based benchmarks, demonstrating that careful parallelization effectively restores real-time performance.