The article discusses the implementation of the algorithm of vHGW grayscale morphology with the use of OpenMP parallel programming technology and NVIDIA CUDA. Morphological operations have low arithmetic complexity, but the use of data parallelism can improve acceleration performance with parallel processors such as graphics processors (GPU). Performing these operations for GPU provides significant acceleration compared with the central processor implementation for structuring elements of various sizes. It is shown that vHGW algorithm is a fast algorithm for computing dilation and erosion of binary and grayscale images on a serial processor. The implementation representation of vHGW algorithms for GPUs with CUDA technology improves the performance of morphological image processing. The authors demonstrate the efficiency of the algorithm implementation using CUDA technology, comparing it with the filtration OpenMP binary and grayscale images with different resolution and different size of the structuring element.