When using the PIL Image.open open RGB image, image.size = (w, h)
i.e. (column, row) -> (x, y)
if the carrier is converted to ndarray FIG form for ease of operation, the:
image = np.asarray (image)
Image.shape after conversion = (h, w, c) c is a number of channels, the RGB image. 3 = C
(H, W) of (row, column), that is, (y, x)