Abstract: Aim at enhancing the contrast between the metal stamping character （SC） and its background, to reduce the interference of the background information to improve the accuracy and stability of the SC seg- mentation, an image fusion algorithm based on multi-directional illumination is proposed. Firstly, four circumferentially uniform distributed light sources were adopted for acquiring images under different illu- mination directions by lighting up the lights one by one. And then a background homogenization process was implemented to eliminate the light intensity uneven distribution under unidirectional illumination. Finally, the reflection features at the depression area and the flat area were well studied and a mathematical model based on calculating the image gray differences of two images in opposite illumination direction was adopted to achieve image fusion and to enhance the image contrast. Experimental results show that the pro- posed algorithm can effectively eliminate the background interference, such as oil, corrosion and so on, and enhance the contrast between the SC and its background, lay the foundation for the segmentation and recog- nition of stamping characters.