Obtain page boundaries

Summary

X and Y margins are subtracted from a shrunk black and white copy of the image.

Details

The story begins with the WarpedImage class in the image.py module.

• When initialised, it reads in the image file

self.cv2_img = cv2.imread(imgfile)

• A smaller version is made (stored in the .small attribute) by calling resize_to_screen().

def resize_to_screen(self, copy=False):
    height, width = self.cv2_img.shape[:2]
    scl_x = float(width) / cfg.image_opts.SCREEN_MAX_W
    scl_y = float(height) / cfg.image_opts.SCREEN_MAX_H
    scl = int(np.ceil(max(scl_x, scl_y)))
    if scl > 1.0:
        inv_scl = 1.0 / scl
        img = cv2.resize(
            self.cv2_img, (0, 0), None, inv_scl, inv_scl, cv2.INTER_AREA
        )
    elif copy:
        img = self.cv2_img.copy()
    else:
        img = self.cv2_img
    return img

• The SCREEN_MAX_W and SCREEN_MAX_H parameters are configured through the global cfg object. They assume a roughly HD screen size for viewing intermediate results (1280x700)
  • cfg is constructed by loading defaults from a TOML config file which ships with the package (see options.py) but all of these settings can be modified using command line flags (see cli.py)

The page boundaries are obtained by the call to calculate_page_extents, where the height of the shrunk page (to fit a HD screen) is offset by an x and y margin (each on both sides).

• An all-zero page mask [8-bit integer numpy array, i.e. 0-255] is created
  • (i.e. to initialise the mask as "non-page" or "off" but as a colour: black),
• a rectangle is overlaid (in-place) onto the pagemask according to the margin size, with colour 255 (i.e. white), with a negative value for thickness which is a sentinel value meaning 'filled' rather than outline of a rectangle

height, width = self.small.shape[:2]
xmin = cfg.image_opts.PAGE_MARGIN_X
ymin = cfg.image_opts.PAGE_MARGIN_Y
xmax, ymax = (width - xmin), (height - ymin)
self.pagemask = np.zeros((height, width), dtype=np.uint8)
rectangle(self.pagemask, (xmin, ymin), (xmax, ymax), color=255, thickness=-1)
self.page_outline = np.array(
    [[xmin, ymin], [xmin, ymax], [xmax, ymax], [xmax, ymin]]
)

Note that the page outline is a list (matrix) of the corners clockwise from bottom left (BL, TL, TR, BR).