Based on the affine geometry theory, a new method for image invariant construction is proposed under the global affine transformation condition. At first, the image feature points are obtained by affine region cutting. Secondly, the convex hull of feature points is established, and then the vertices of the convex hull are well proportionally combined into some feature regions. At last, we use these feature regions to construct affine invariants, based on affine geometry invariant property. In order to validate the image invariant's stability and high time efficiency, experiments were performed and the proposed method was analyzed in comparison with the existent methods.