35

Suppose you have the following image:

Example:

Now I want to extract each of the independent letters into individual images. Currently, I've recovered the contours and then drew a bounding box, in this case for the character a:

Bounding box for the character 'a'

After this, I want to extract each of the boxes (in this case for the letter a) and save it to an image file.

Expected result:

Result

Here's my code so far:

import numpy as np
import cv2

im = cv2.imread('abcd.png')
im[im == 255] = 1
im[im == 0] = 255
im[im == 1] = 0
im2 = cv2.cvtColor(im,cv2.COLOR_BGR2GRAY)
ret,thresh = cv2.threshold(im2,127,255,0)
contours, hierarchy = cv2.findContours(thresh,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)

for i in range(0, len(contours)):
    if (i % 2 == 0):
       cnt = contours[i]
       #mask = np.zeros(im2.shape,np.uint8)
       #cv2.drawContours(mask,[cnt],0,255,-1)
       x,y,w,h = cv2.boundingRect(cnt)
       cv2.rectangle(im,(x,y),(x+w,y+h),(0,255,0),2)
       cv2.imshow('Features', im)
       cv2.imwrite(str(i)+'.png', im)

cv2.destroyAllWindows()

Thanks in advance.

coffeewin
  • 182
  • 3
  • 6
  • 26
Edgar Andrés Margffoy Tuay
  • 1,337
  • 5
  • 16
  • 29

3 Answers3

47

The following will give you a single letter

letter = im[y:y+h,x:x+w]
Abid Rahman K
  • 51,886
  • 31
  • 146
  • 157
Andrey Kamaev
  • 29,582
  • 6
  • 94
  • 88
4

Here's an approach:

  • Convert image to grayscale
  • Otsu's threshold to obtain a binary image
  • Find contours
  • Iterate through contours and extract ROI using Numpy slicing

After finding contours, we use cv2.boundingRect() to obtain the bounding rectangle coordinates for each letter. 

x,y,w,h = cv2.boundingRect(c)

To extract the ROI, we use Numpy slicing

ROI = image[y:y+h, x:x+w]

Since we have the bounding rectangle coordinates, we can draw the green bounding boxes

cv2.rectangle(copy,(x,y),(x+w,y+h),(36,255,12),2)

Here's the detected letters

enter image description here

Here's each saved letter ROI

enter image description here

import cv2

image = cv2.imread('1.png')
copy = image.copy()
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
thresh = cv2.threshold(gray,0,255,cv2.THRESH_OTSU + cv2.THRESH_BINARY)[1]

cnts = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]

ROI_number = 0
for c in cnts:
    x,y,w,h = cv2.boundingRect(c)
    ROI = image[y:y+h, x:x+w]
    cv2.imwrite('ROI_{}.png'.format(ROI_number), ROI)
    cv2.rectangle(copy,(x,y),(x+w,y+h),(36,255,12),2)
    ROI_number += 1

cv2.imshow('thresh', thresh)
cv2.imshow('copy', copy)
cv2.waitKey()
nathancy
  • 42,661
  • 14
  • 115
  • 137
  • 1
    How can I apply this method to extract images of words instead of alphabets @nathancy ? – Raj May 10 '20 at 17:35
  • @Raj same process, just perform image processing till you obtain a binary image then you can use this example. It could work with anything, shapes, objects, word clusters, blobs just as long as the foreground object you're trying to extract is different from the background. In image processing, we generally want the desired object to be in white with the background in black – nathancy Oct 23 '20 at 21:44
0
        def bounding_box_img(img,bbox):
            x_min, y_min, x_max, y_max = bbox
            bbox_obj = img[y_min:y_max, x_min:x_max]
            return bbox_obj

        img = cv2.imread("image.jpg")
        cropped_img = bounding_box_img(img,bbox)
        cv2.imshow(cropped_img)

this returns cropped image (bounding box)

in this aproach, bounding box coordinates bases on pascal-voc annotation formats like here

livan3li
  • 119
  • 2
  • 8