Abstract: The fractal dimension of image is an important quantitative character, which can be widely used in image analysis. But general algorithms need more computation cost and can not be computed in real-time. In order to compute the fractal dimension of an image in real-time, an algorithm is designed in this paper by using morphological cover, according to the definition of Minkowski-Bouligand dimension. The comparison with Box Counting method is done. The result of experimental computation manifests that the fractal dimension of the image can be precisely and robustly computed by using the algorithm. The architecture of real-time parallel realization is given in this paper.