Trim the whitespace of the grayscale image

Question

I get a images set, and the images are like this.

How can I remove the beneath white part of the image using python, which doesn't contain any useful content?

I read the image to numpy array in python.

My code is like this:

data_dir = "/Users/leon/Projects/inpainting/data/"
images = []
files = glob.glob (data_dir + "*.jpg")
for file in files:
    image = cv2.imread(file, 0)
    images.append(image)

I'm thinking using two for loop to check the pixel element, but I got 5,000 images like this, will it be to slow to do it in this way? — Hao Chen, May 29 '18 at 20:40
Possible duplicate of https://stackoverflow.com/questions/10615901/trim-whitespace-using-pil/10616717 — DYZ, May 29 '18 at 20:42
https://www.physicsforums.com/threads/removing-0s-from-2d-arrays.884140/ see last part — Mark Setchell, May 29 '18 at 20:46
thank you guys but I find a better way to do it, see my answer blew. — Hao Chen, May 29 '18 at 22:43

filippo · Answer 1 · 2018-05-29T21:04:52.033

1

This trims white space row-wise both above and below (actually it trims any full white row):

trimmed = image[np.where(~np.all(image == 255, axis=1))]

If you need to trim just the top and bottom margins you can do:

empty_row_mask = np.all(image == 255, axis=1)
top = np.searchsorted(~empty_row_mask, True)
bottom = np.searchsorted(empty_row_mask, True)
trimmed = image[top:bottom]

edited May 29 '18 at 21:04

answered May 29 '18 at 20:56

filippo

5,197
2
21
44

thank you, I find a way which can find the most suitable window for trimming the undesired white part both sides – Hao Chen May 29 '18 at 22:46

score 0 · Answer 2 · answered May 29 '18 at 22:44

I find a way to do it using openCV3 and python3.6

image_neg = cv2.bitwise_not(image) # invert the root to white
coords = cv2.findNonZero(image_neg) # find all the non-zero points (root)
x, y, w, h = cv2.boundingRect(coords) # find minimum spanning bounding box
rect = image[y:y+h, x:x+w]

Trim the whitespace of the grayscale image

2 Answers2