prakharguptaujjain/Scene-Text-Binarization

This project focuses on scene text binarization, which aims to enhance the readability and usability of text found in natural scenes, by converting it into a binary format.

Scene Text Binarization and Text Recognition

The following code is used perform text binarization on the given scenes and later apply text recognition to recognize the text in the image.

Specifications

• scene_text_binarization.py - Takes image as input and return text binarized images.
• utils.py - Contains the necessary utility functions used in scene_text_binarization.py
• code.ipynb - The following is being used to showcase the output images after appying binarization on the images.
• frozen_east_text_detection.pb - Pre-trained model used for text-detection.
• Optical Character Recognition.ipynb - After getting the text-binarized images, easy-ocr is being applied on the images to recognize the text. The file applies easy-ocr on the image and plots the output after text recognition.

Compilation and Execution

• Place the image in input_images folder
• run python scene_text_binarization.py
• Output images will be saved in output_images/<image_name>
• The results can also be reviewed by running the cells of the code.ipynb and Optical Character Recgognition.ipynb.

Example

Input Image:

After binarization: