The invention discloses an image reducing method with image topology maintained. The method comprises the following steps of: S1, dividing a binary image Ik with the size of 2m-k+1*2n-k+1, wherein the image is divided into 2m-k*2n-k blocks according to the principle that each 2*2 pixels belong to one block, each block pk (I, j) contains pi and jk target pixels, I is not less than 1, but is not more than m, and I is not less than 1, but is not more than n; S2, dividing the block pk (I, j) into a smooth field and a marginal field by a human eye visual system; S3, reducing the pixel Ik+1 (I, j) in the smooth field by the human eye visual system, and reducing the pixel Ik+1 (I, j) in the marginal field according to the connectedness classification result of a neighborhood template defined by the pixel Ik+1 (I, j) in the binary image Ik; S4, carrying out iterative computation according to S1-S3 to reduce the image Ik to obtain the target image of Ia+1=2m-a* 2n-a with the reducing factor of SF=2a*2a, wherein a is smaller than min , wherein min is in the range of m to n. The invention enables the image reduced by large multiples to maintain the topology and obtains lower computation complexity and higher computation speed.