我有下面的图片在Tesseract OCR中使用:
我处理图片的代码是:
# HOCR
with image[450:6200, 840:3550] as cropped:
imgPage = wi(image = cropped)
imageBlob = imgPage.make_blob('png')
horas = gerarHocr(imageBlob)
def gerarHocr(imageBlob):
image = Image.open(io.BytesIO(imageBlob))
markup = pytesseract.image_to_pdf_or_hocr(image, lang='por', extension='hocr', config='--psm 6')
soup = BeautifulSoup(markup, features='html.parser')
spans = soup.find_all('span', {'class' : 'ocrx_word'})
listHoras = []
...
return listHoras
例如,尽管我的OCR有时会感到困惑,并且将8
与3
复制,并返回07:44/14:183
而不是07:44/14:13
。
[我认为,如果我使用魔杖去除灰线,则可以提高OCR的置信度。请问我该怎么做?
谢谢,
我将使用cv2
和/或numpy.array
将浅灰色转换为白色
img[ img > 128 ] = 255
将深灰色转换为黑色
img[ img < 128 ] = 0
import cv2
folder = '/home/user/images/'
# read it
img = cv2.imread(folder + 'old_img.png')
# convert ot grayscale
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# reduce colors
img[ img > 128 ] = 255
img[ img < 128 ] = 0
# save it
cv2.imwrite(folder + 'new_img.png', img)
# display result
#cv2.imshow('window', img)
#cv2.waitKey(0) # press any key in window to close it
#cv2.destroyAllWindows()
结果