qurator-spk/eynollah
1
0
Fork 0
mirror of https://github.com/qurator-spk/eynollah.git synced 2025-08-29 11:59:55 +02:00
Code Issues Projects Releases Wiki Activity

Implement hyphenated textline merging in OCR engine and a bug fixed for curved textline OCR

Browse source
This commit is contained in:
vahidrezanezhad 2025-05-21 14:39:31 +02:00
parent c0835665a9
commit f94fc9973b
1 changed files with 71 additions and 86 deletions
Download patch file Download diff file Expand all files Collapse all files

81
src/eynollah/eynollah.py
View file

@ -5500,7 +5500,6 @@ class Eynollah_ocr:
def get_orientation_moments_of_mask(self, mask): def get_orientation_moments_of_mask(self, mask):
mask=mask.astype('uint8') mask=mask.astype('uint8')
print(mask.shape)
contours, _ = cv2.findContours(mask[:,:,0], cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) contours, _ = cv2.findContours(mask[:,:,0], cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
largest_contour = max(contours, key=cv2.contourArea) if contours else None largest_contour = max(contours, key=cv2.contourArea) if contours else None
@ -5547,39 +5546,24 @@ class Eynollah_ocr:
def break_curved_line_into_small_pieces_and_then_merge(self, img_curved, mask_curved): def break_curved_line_into_small_pieces_and_then_merge(self, img_curved, mask_curved):
peaks_4 = self.return_splitting_point_of_image(img_curved) peaks_4 = self.return_splitting_point_of_image(img_curved)
if len(peaks_4)>0:
img_0 = img_curved[:, :peaks_4[0], :]
img_1 = img_curved[:, peaks_4[0]:peaks_4[1], :]
img_2 = img_curved[:, peaks_4[1]:peaks_4[2], :]
img_3 = img_curved[:, peaks_4[2]:peaks_4[3], :]
img_4 = img_curved[:, peaks_4[3]:, :]
mask_0 = mask_curved[:, :peaks_4[0], :]
mask_1 = mask_curved[:, peaks_4[0]:peaks_4[1], :]
mask_2 = mask_curved[:, peaks_4[1]:peaks_4[2], :]
mask_3 = mask_curved[:, peaks_4[2]:peaks_4[3], :]
mask_4 = mask_curved[:, peaks_4[3]:, :]
cv2.imwrite("split0.png", img_0)
cv2.imwrite("split1.png", img_1)
cv2.imwrite("split2.png", img_2)
cv2.imwrite("split3.png", img_3)
or_ma_0 = self.get_orientation_moments_of_mask(mask_0)
or_ma_1 = self.get_orientation_moments_of_mask(mask_1)
or_ma_2 = self.get_orientation_moments_of_mask(mask_2)
or_ma_3 = self.get_orientation_moments_of_mask(mask_3)
or_ma_4 = self.get_orientation_moments_of_mask(mask_4)
imgs_tot = [] imgs_tot = []
imgs_tot.append([img_0, mask_0, or_ma_0] )
imgs_tot.append([img_1, mask_1, or_ma_1]) for ind in range(len(peaks_4)+1):
imgs_tot.append([img_2, mask_2, or_ma_2]) if ind==0:
imgs_tot.append([img_3, mask_3, or_ma_3]) img = img_curved[:, :peaks_4[ind], :]
imgs_tot.append([img_4, mask_4, or_ma_4]) mask = mask_curved[:, :peaks_4[ind], :]
elif ind==len(peaks_4):
img = img_curved[:, peaks_4[ind-1]:, :]
mask = mask_curved[:, peaks_4[ind-1]:, :]
else:
img = img_curved[:, peaks_4[ind-1]:peaks_4[ind], :]
mask = mask_curved[:, peaks_4[ind-1]:peaks_4[ind], :]
or_ma = self.get_orientation_moments_of_mask(mask)
imgs_tot.append([img, mask, or_ma] )
w_tot_des_list = [] w_tot_des_list = []
w_tot_des = 0 w_tot_des = 0
@ -5622,22 +5606,9 @@ class Eynollah_ocr:
for ind in range(len(w_tot_des_list)): for ind in range(len(w_tot_des_list)):
img_final_deskewed[:,w_indexer:w_indexer+w_tot_des_list[ind],:] = imgs_deskewed_list[ind][:,:,:] img_final_deskewed[:,w_indexer:w_indexer+w_tot_des_list[ind],:] = imgs_deskewed_list[ind][:,:,:]
w_indexer = w_indexer+w_tot_des_list[ind] w_indexer = w_indexer+w_tot_des_list[ind]
#cv2.imwrite('final.png', img_final_deskewed)
#print(or_ma_0, or_ma_1, or_ma_2, or_ma_3, or_ma_4, 'orients')
##cv2.imwrite("split4.png", img_curved[:, peaks_4[3]:peaks_4[4], :])
##cv2.imwrite("split5.png", img_curved[:, peaks_4[4]:peaks_4[5], :])
##cv2.imwrite("split6.png", img_curved[:, peaks_4[5]:peaks_4[6], :])
##cv2.imwrite("split7.png", img_curved[:, peaks_4[6]:peaks_4[7], :])
##cv2.imwrite("split8.png", img_curved[:, peaks_4[7]:peaks_4[8], :])
##cv2.imwrite("split9.png", img_curved[:, peaks_4[8]:peaks_4[9], :])
#cv2.imwrite("split4.png", img_4)
#sys.exit()
return img_final_deskewed return img_final_deskewed
else:
return img_curved
def run(self): def run(self):
ls_imgs = os.listdir(self.dir_in) ls_imgs = os.listdir(self.dir_in)
@ -6144,7 +6115,21 @@ class Eynollah_ocr:
text_by_textregion = [] text_by_textregion = []
for ind in unique_cropped_lines_region_indexer: for ind in unique_cropped_lines_region_indexer:
extracted_texts_merged_un = np.array(extracted_texts_merged)[np.array(cropped_lines_region_indexer)==ind] extracted_texts_merged_un = np.array(extracted_texts_merged)[np.array(cropped_lines_region_indexer)==ind]
if len(extracted_texts_merged_un)>1:
text_by_textregion_ind = ""
next_glue = ""
for indt in range(len(extracted_texts_merged_un)):
if extracted_texts_merged_un[indt].endswith('⸗') or extracted_texts_merged_un[indt].endswith('-'):
text_by_textregion_ind = text_by_textregion_ind + next_glue + extracted_texts_merged_un[indt][:-1]
next_glue = ""
else:
text_by_textregion_ind = text_by_textregion_ind + next_glue + extracted_texts_merged_un[indt]
next_glue = " "
text_by_textregion.append(text_by_textregion_ind)
else:
text_by_textregion.append(" ".join(extracted_texts_merged_un)) text_by_textregion.append(" ".join(extracted_texts_merged_un))
#print(text_by_textregion, 'text_by_textregiontext_by_textregiontext_by_textregiontext_by_textregiontext_by_textregion')
indexer = 0 indexer = 0
indexer_textregion = 0 indexer_textregion = 0