pad

pad(img: np.ndarray, pad_size: Union[int, Tuple[int, int], Tuple[int, int, int, int]], fill_value: Optional[Union[int, Tuple[int, int, int]]] = 0, pad_mode: Union[str, int, BORDER] = BORDER.CONSTANT) -> np.ndarray

• Description: Perform padding on the input image.

• Parameters:

  • img (np.ndarray): Input image to be padded.
  • pad_size (Union[int, Tuple[int, int], Tuple[int, int, int, int]]): Padding size. It can be an integer to specify the same padding amount for all sides, a tuple (pad_top, pad_bottom, pad_left, pad_right) to specify different padding amounts for each side, or a tuple (pad_height, pad_width) to specify the same padding amount for height and width.
  • fill_value (Optional[Union[int, Tuple[int, int, int]]]): Value used for padding. If the input image is a color image (3 channels), fill_value can be an integer or a tuple (R, G, B) to specify the filling color. If the input image is a grayscale image (1 channel), fill_value should be an integer. Default is 0.
  • pad_mode (Union[str, int, BORDER]): Padding mode. Available options are:
    • BORDER.CONSTANT: Pad with constant values (fill_value).
    • BORDER.REPLICATE: Pad by replicating edge pixels.
    • BORDER.REFLECT: Pad by reflecting the image around the edges.
    • BORDER.REFLECT101: Pad by reflecting the image around the edges, with a slight adjustment to avoid artificial seams. Default is BORDER.CONSTANT.

• Returns:

  • np.ndarray: Padded image.

• Example:

  import docsaidkit as D
  
  img = D.imread('lena.png')
  pad_img = D.pad(img, pad_size=20, fill_value=(255, 0, 0))
  
  # Resize the padded image to the original size for visualization
  pad_img = D.imresize(pad_img, [img.shape[0], img.shape[1]])