The SURF feature detector works by applying an approximate Gaussian second derivative mask to an image at many scales.� Because the feature detector applies masks along each axis and at 45 deg to the axis it is more robust to rotation than the Harris corner.� The method is very fast because of the use of an integral image where the value of a pixel (x,y) is the sum of all values in the rectangle defined by the origin and (x,y).� In such an image the sum of the pixels within a rectangle of any size in a source image can be found as the result of 4 operations.� This allows a rectangular mask of any size to be applied with very little computing time.