Mountain glaciers are sensitive to environment. It is important to acquire ice flow velocities over time for glacier research and hazard forecast. For this paper, cross-correlating of optical images is used to monitor ice flow velocities, and an improvement, which is called "moving grid," is made to this method. For this research, two remote-sensing images in a certain glacier area, during different times are selected. The first image is divided into grids, and we calculated the correlation coefficient of each window in the grid with the window on the second image. The window with the highest correlation coefficient is considered the counterpart one on the first image. The displacement of the two corresponding windows is the movement of the glacier, and it is used to calculate glacier surface velocity. Compared to the traditional way of dividing an image with ascertain grid, this method uses small steps to move the grid from one location to another adjacent location until the whole glacier area is covered in the image, thus increasing corresponding point density. We selected a glacier in the Tianshan Mountains for this experiment and used two remote-sensing images with a 10-year interval to determine this method.