This paper proposes a fast pixelated mask optimization method for extreme ultraviolet lithography. An improved fast pixelated thick mask model is utilized in mask optimization. The point pulses on the edge pixels are set according to the mask pixel size. On the basis of the dual edge evolution strategy, the optimization variables are adaptively initialized in each epoch of optimization according to the difference between the current resist pattern contour and the target pattern contour. The optimization efficiency is improved by generating the initial individuals and population using priori information. One-dimensional line-space patterns and two-dimensional complex patterns are used for simulation. Simulation results show that the imaging simulation accuracy is effectively improved. In addition, the optimization efficiency of the two complex patterns is effectively increased by the proposed method.