PDF-Image Highlighter

I am making a highlighter for a group of image in a PDF file. It is a long journey.

Prerequisites

• Install Google tesseract in your PC. Follow the installation tutorial here. This project used tesseract with configuration --psm 3 --oem 1 (LSTM model and made for small text detection). To install tesseract in Ubuntu machine:

sudo apt install tesseract-ocr
sudo apt install libtesseract-dev

• Python 3.x.
• You will need to install pillow, pytesseract, pdf2image and textdistance.

pip install pillow
pip install pytesseract
pip install pdf2image
pip install textdistance

Example #1: Detect and read the text.

I made some example! I placed my example files in example directory. First, change your active directory to example.

cd example

After that, run this command:

# python example.py <image_file>
python example.py textimage2.png

You will get the result (shortened version):

The text extraction result is: 
Attn: Pearlene Then
Sub-BU: Technology Department - PUBO9
To: Public Utilities Board
Accounts Payable Section, Finance Department,
...

Data: 
5;1;1;1;1;1;6;22;35;13;93;Attn:
5;1;1;1;1;2;49;22;68;13;92;Pearlene
5;1;1;1;1;3;122;22;39;13;96;Then
5;1;1;1;2;1;7;43;64;13;90;Sub-BU:
...

Example #2: Highlight text in an image

In this example, I am trying to highlight a spesific text inside an image. Do the following steps to try it:

1. You can specify the text that you want to search in the image by changing the soe variable located at line 128 on process.py. Feel free to change soe value. By default, i put my soe with some sentences like salta and El zorro. That means, I want to search those sentences inside the given image.

soe = ["salta","El zorro"]

2. Run process.py followed by the path of the image that you want to detect.

# python process.py <image_path>
python process.py example/textimage.png

3. You will see the result of text detection process according to the sentences that you put inside soe.

My upcoming updates

My next plan for this project is:

• Read pdf file, convert it to a pile of images and detect the defined soe texts inside it.
• I planned to make an API of this project by using Flask.