In this paper, we present a method of removing reflective components from a single image by utilizing deep learning-based transformer method. Specifically, the proposed method effectively combines local and global information within an image by using the hierarchical structure of top-k Transformer based on U-Net structure. Proposed method shows precise removal of reflective components from a single image and generation of high-quality images that are clean and clear. Experimental results show that the proposed method has enhanced reflection component removal performance compared to the existing method. In addition, the effect of minimizing additional undesired artifact or information loss while removing reflection components was confirmed.