textbsr 0.0.24

a simple version for blind text image super-resolution (current version is only for English and Chinese)

This is a simple text image super-resolution package.

More details can be found in our Project Page: https://github.com/csxmli2016/textbsr

It can post-process the text region with a simple commond, i.e.,

textbsr -i [LR_TEXT_PATH] -b [BACKGROUND_SR_PATH]

Quick Start

Dependencies and Installation

• numpy
• opencv-python
• torch>=1.8.1
• torchvision>=0.9

# Install with pip
pip install textbsr

Basic Usage

# On the terminal command
textbsr -i [LR_TEXT_PATH]

or

# On the python environment
from textbsr import textbsr
textbsr.bsr(input_path='./testsets/LQs')

Parameter details:

parameter name	default	description
-i, --input_path	-	The lr text image path. It can be a full image or a text region only
-b, --bg_path	None	The background sr path from other methods. If None, we only super-resolve the text region.
-o, --output_path	None	The save path for text sr result. If None, we save the results on the same path with [input_path]_TIMESTAMP
-a, --aligned	False	action='store_true'. If True, the input text image contains only text region. If False, we use CnSTD to detect and restore the text region.
-s, --save_text	False	action='store_true'. If True, save the LR and SR text layout.
-d, --device	None	Device, use 'gpu' or 'cpu'. If None, we use torch.cuda.is_available to select the device.

Example for post-processing the text region

# On the terminal command
textbsr -i [LR_TEXT_PATH] -b [BACKGROUND_SR_PATH] -s

or

# On the python environment
from textbsr import textbsr
textbsr.bsr(input_path='./testsets/LQs', bg_path='./testsets/RealESRGANResults', save_text=True)

When [BACKGROUND_SR_PATH] is None, we only restore the text region and paste back to the LR input, with the background region unchanged.

---

Example for super-resolving the aligned text region

# On the terminal command
textbsr -i [LR_TEXT_PATH] -a

or

# On the python environment
from textbsr import textbsr
textbsr.bsr(input_path='./testsets/LQs', aligned=True)

If you find this package helpful, please kindly consider to cite our paper:

@InProceedings{li2023marconet,
author = {Li, Xiaoming and Zuo, Wangmeng and Loy, Chen Change},
title = {Learning Generative Structure Prior for Blind Text Image Super-resolution},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year = {2023}
}