“Missing” square between faces 1 and 2 is divided into four triangles of equal size, separated by the lines 2x = y, x = y, and x = 2y, as seen on the left. The two images in the middle reveal how the rotated faces 1 and 2 are separately weighted along those segments. The data of face 1 have full weight (black) on the outer triangle adjacent to face 2, and no weight (white) on the other outer triangle, while the data of face 2 are treated conversely. A smooth transition is provided by the weights on the triangles in between. The sum of the weighted faces is used to fill the gap, as demonstrated in the right-most illustration.