For the skew correction of complex scanned electronic documents with text, images, tables and formulas, a new skew detection method is proposed based on Shearlet transform and multi-scale analysis. Based on the localization and directionality of Shearlet transform, energy in all directions can be obtained. The two components with the highest energy can correspond to the horizontal and vertical directions, respectively. According to these two directions, the direction of the text line in the scanned document can be accurately detected and the skew angle of document image can be further determined. The experimental results show that the proposed method can avoid the interference of images, tables, and so on in scanned documents, and has a good anti-noise ability. The proposed method is of higher detection accuracy for document images with complex content.