Tesseract OCR engine: Tesseract is an open source OCR engine, you need to install it first. You can download the installer or source code for your operating system from the Tesseract official website (https://github.com/tesseract-ocr/tesseract), and install it according to the official documentation.
Tesseract OCR may not be accurate for low-resolution or blurry images. Try using higher resolution and sharper images to improve the accuracy of the recognition results. For screenshots on Mac, they are generally very clear, so this shortcoming has little effect.
On Mac, install using the method recommended by the official website :
brew install tesseract
The tesseract directory can then be found using brew info tesseract, e.g.
/usr/local/Cellar/tesseract/5.3.2/bin/tesseract
demo:
import pytesseract
from PIL import Image
# 可以写一个函数 crop_picture 将原图裁剪一下，只保留想要识别文本的部分，这样识别更加准确一些。
def crop_picture(picture_path, crop_box: list):
"""
crap picture with crop_box
:param picture_path: picture to be crapped
:param crop_box: crop region, eg: [100, 200, 300, 350]
:return: path of crapped picture
"""
dirname = os.path.dirname(picture_path)
basename = os.path.basename(picture_path)
new_basename = ''.join([basename.split('.')[0], '_new.', basename.split('.')[1]])
picture_origin = Image.open(picture_path)
picture_origin_size = picture_origin.size
if crop_box[2] is None:
crop_box[2] = picture_origin_size[0]
if crop_box[3] is None:
crop_box[3] = picture_origin_size[1]
picture_new = picture_origin.crop(tuple(crop_box))
picture_new_path = os.path.join(dirname, new_basename)
picture_new.save(picture_new_path)
return picture_new_path
def get_text_from_picture(picture_path, crop_box: list):
"""
get text from picture
:param picture_path: picture to be crapped
:param crop_box: crop region, eg: [100, 200, 300, 350]
:return: text
"""
pytesseract.pytesseract.tesseract_cmd = r'/usr/local/Cellar/tesseract/5.3.2/bin/tesseract'
picture_new_path = crop_picture(picture_path, crop_box=crop_box)
image = Image.open(picture_new_path)
text = pytesseract.image_to_string(image, lang='eng')
print(text)
return text
if __name__ == '__main__':
get_text_from_picture('my_picture_path', crop_box=[585, 360, None, 800])