Extracting text OpenCV

Question

I am trying to find the bounding boxes of text in an image and am currently using this approach:

// calculate the local variances of the grayscale image
Mat          


        

Answer 1

You can utilize a python implementation SWTloc.

Full Disclosure : I am the author of this library

To do that :-

First and Second Image

Notice that the text_mode here is 'lb_df', which stands for Light Background Dark Foreground i.e the text in this image is going to be in darker color than the background

from swtloc import SWTLocalizer
from swtloc.utils import imgshowN, imgshow

swtl = SWTLocalizer()
# Stroke Width Transform
swtl.swttransform(imgpaths='img1.jpg', text_mode = 'lb_df',
                  save_results=True, save_rootpath = 'swtres/',
                  minrsw = 3, maxrsw = 20, max_angledev = np.pi/3)
imgshow(swtl.swtlabelled_pruned13C)

# Grouping
respacket=swtl.get_grouped(lookup_radii_multiplier=0.9, ht_ratio=3.0)
grouped_annot_bubble = respacket[2]
maskviz = respacket[4]
maskcomb  = respacket[5]

# Saving the results
_=cv2.imwrite('img1_processed.jpg', swtl.swtlabelled_pruned13C)
imgshowN([maskcomb, grouped_annot_bubble], savepath='grouped_img1.jpg')

---

---

Third Image

Notice that the text_mode here is 'db_lf', which stands for Dark Background Light Foreground i.e the text in this image is going to be in lighter color than the background

from swtloc import SWTLocalizer
from swtloc.utils import imgshowN, imgshow

swtl = SWTLocalizer()
# Stroke Width Transform
swtl.swttransform(imgpaths=imgpaths[1], text_mode = 'db_lf',
              save_results=True, save_rootpath = 'swtres/',
              minrsw = 3, maxrsw = 20, max_angledev = np.pi/3)
imgshow(swtl.swtlabelled_pruned13C)

# Grouping
respacket=swtl.get_grouped(lookup_radii_multiplier=0.9, ht_ratio=3.0)
grouped_annot_bubble = respacket[2]
maskviz = respacket[4]
maskcomb  = respacket[5]

# Saving the results
_=cv2.imwrite('img1_processed.jpg', swtl.swtlabelled_pruned13C)
imgshowN([maskcomb, grouped_annot_bubble], savepath='grouped_img1.jpg')

---

You will also notice that the grouping done is not so accurate, to get the desired results as the images might vary, try to tune the grouping parameters in swtl.get_grouped() function.