The present invention provides a method for generating images of multi-views. The method includes obtaining a 2D original image of an article and background figures of multi-views; calculating the background image range and the main body image range of the 2D original image of the article; cutting the main body image out; generating a depth model according to an equation; cutting the depth model according to the main body image range of the cut 2D image of the article; shifting every pixel in the main body image of the 2D original image of the article according to the cut depth model to obtain shifted main body images of multi-views; and synthesizing the shifted main body images of multi-views and the background figures of multi-views to obtain the final images of multi-views for 3D image reconstruction.