OCR on PDF files using Python

Hi there folks! You might have heard about OCR using Python. The most famous library out there is tesseract which is sponsored by Google. It is very easy to do OCR on an image. The issue arises when you want to do OCR over a PDF document.

I am working on a project where I want to input PDF files, extract text from them and then add the text to the database. I had to search a lot before I stumbled over the final solution. So without wasting any time, lets begin.

Installing Tesseract

It is very easy to install tesseract on various operating systems. For the sake of simplicity I will be using Ubuntu as an example. In Ubuntu you simply have to run the following command in the terminal:

sudo apt-get install tesseract-ocr

It will install Tesseract along with the support for three languages.

Installing PyOCR

Now we need to install the Python bindings for tesseract. Fortunately, there are some pretty nice bindings out there. We will be installing a latest one:

pip install git+

Installing Wand and PIL

We need to install two other dependencies as well before we can move on. First one is Wand. It is the Python bindings for Imagemagick. We will be using it for converting PDF files to images:

pip install wand

We will be using PIL as well because PyOCR needs it. You can take a look at the official docs on how to install it on your operating system.

Warming up

Let’s start writing our script. First of all, we will be importing the required libraries:

from wand.image import Image
from PIL import Image as PI
import pyocr
import io

Note: I imported Image from PIL as PI because otherwise it would have conflicted with the Image module from wand.image.

Get Going

Now we need to get the handle of the OCR library (in our case, tesseract) and the language which will be used by pyocr.

tool = pyocr.get_available_tools()[0]
lang = tool.get_available_languages()[1]

We used the second language in the tool.get_available_languages() because the last time I checked, it was English.

Now we need to setup two lists which will be used to hold our images and final_text.

req_image = []
final_text = []

Next step is to open the PDF file using wand and convert it to jpeg. Let’s do it!

image_pdf = Image(filename="./PDF_FILE_NAME", resolution=300)
image_jpeg = image_pdf.convert('jpeg')

Note: Replace PDF_FILE_NAME with a valid PDF file name in the current path.

wand has converted all the separate pages in the PDF into separate image blobs. We can loop over them and append them as a blob into the _reqimage list.

for img in image_jpeg.sequence:
    img_page = Image(image=img)

Now we just need to run OCR over the image blobs. It is very easy.

for img in req_image: 
    txt = tool.image_to_string(,

Now all of the recognized text has been appended in the _finaltext list. You can use it in any way you want. I hope this tutorial was helpful for you guys!

If you have any comments and suggestions then do let me know in the comments section below.

Till next time! 🙂



If you liked what you read then I am sure you will enjoy a newsletter of the content I create. I send it out every other month. It contains new stuff that I make, links I find interesting on the web, and occasional discount coupons for my book. Join the 5000+ other people who receive my newsletter:

I send out the newsletter once every other month. No spam, I promise + you can unsubscribe at anytime

✍️ Comments

Cristi Vlad

This does not work on windows as easily because it’s tricky to get tesseract up and running. Or, i do not have the knowledge to make it run :)

This One Person
In reply to Cristi Vlad

I ran across this on and wanted to try this on Windows as well. Here’s what I did if it helps anybody out there.

Step #1

Easiest way to obtain tesseract for Windows is here:

I did this with the tesseract-ocr-setup-3.05.00dev.exe download.

Step #2 Works as described.

pip install git+

If you get a git not found error. Install git.

And then go to System-> Advanced system settings -> Environment Variables and add to the PATH variable the location of the git binary. Mine was ‘C:\Program Files\Git\bin’

Leave Environment Variables window open for now.

Step #3 Works as described

Step #4 Works as described

Step #5

You need to verify you have TESSDATA_PREFIX in your System Variables window in the Environment Variables window. It should lead to the installation directory of Tesseract from step #1.

Mine is C:\Program Files (x86)\Tesseract-OCR

I had to make one change in pyocr. Even after mapping the value to a PATH variable I couldn’t get it call tesseract correctly. I fixed it by changing the TESSERACT_CMD value to what is below

TESSERACT_CMD = os.environ[“TESSDATA_PREFIX”] + ‘tesseract.exe’ if == ‘nt’ else ‘tesseract’

You should be to run the code from Step #5 after that. I did have to change

tool.get_available_languages()[1] because that wasn’t english on my system.


What’s do you think about OCR on PDF forms? dot matrix printed invoices with dates, invoice numbers, line items (qty, item, stock #, price) each in designated spaces. Other than the obvious irony that it would be easier now to just get a damn xml file.

Stanley Denman

Why did you convert the pdf to an image file? In my project Ihave the choice of using a tiff or pdf file as the source of text to be extracted and stored in a DB. Which would be better-tiff or pdf - as the source?


Could you say anything about the database you selected to store the test output in?



I am attempting to get this code to work with various PDFs that I have that have been OCR’d from printed documents. I am running on Mac OSX El Capitan with the latest tesseract installed via Homebrew.

I am running it in a Python 2.7.11 environment

My issue is that when I run a PDF document through this procedure it does not generate any text. It is always blank.

What might be causing this?


Very useful, thank you for taking the time to write this down!

Gordon Bruce Nuttall

very helpful, thank you. What can do to increase the accuracy of characters read?

Stefan Brot

Works like a charm, but wand/imagemagick currently might have some security issue (see

I got it to work with pdf2image instead of wand:

#from wand.image import Image # This requires change in /etc/ImageMagick-6/profile.xml and a potential security issue
import pdf2image
from PIL import Image as PI
import pyocr
import io
tool = pyocr.get_available_tools()[0]
lang = tool.get_available_languages()[0]
final_text = []
PDF_PATH = "./ASDF.pdf"
DPI = 300
FORMAT = 'jpg'
STRICT = False
image_jpeg = pdf2image.convert_from_path(PDF_PATH, dpi=DPI, output_folder=OUTPUT_FOLDER, first_page=FIRST_PAGE, last_page=LAST_PAGE, fmt=FORMAT, thread_count=THREAD_COUNT, userpw=USERPWD, use_cropbox=USE_CROPBOX, strict=STRICT)
for img in image_jpeg: 
    txt = tool.image_to_string(

Say something

Send me an email when someone comments on this post.

Thank you!

Your comment has been submitted and will be published once it has been approved. 😊