We propose a method of estimating the importance of figures and tables in scientific papers by propagating importance beyond media; from language to image. In scientific papers, language and image information coordinately enable the reader to easily understand the complicated contents in detail. The proposed method propagates this importance from the sentence to figures/tables in which the position of the sentence referring to a figure/table and surrounding sentences are used to evaluate the importance of figures/tables. We conducted an experiment on estimating the importance of figures/tables by assuming figures/tables used in a presentation poster are important. The experimental results indicated that the proposed method exhibited the highest mean of the average precision (MAP) compared to comparative methods focusing on the size and caption of the figures/tables. Moreover, the proposed method exhibited more MAPs than the comparative method which did not update the sentence importance using similarity among sentences. We believe that the proposed method is effective in supporting the creation of scientific presentation posters from papers.