LSFM-destriping

Introduction

This is a simple version of testing code for LSFM image destriping task.

Quick Start

1.Prepare data

Put the testing images to folder "./data/images/".
The size of images is H * W * 3, where 3 is the channel number of the images. And the height H and width W of the images can be any size. But H and W of each image should be consistent.
The data is named 'image' + number + 'tif'. For example, 'image1.tif'. The format can be PNG, JPG, etc.

Run Load_data.py, and the prepared hdf5 dataset will be saved in path "./data/".

2.Download model

Two well-trained models is provided in https://drive.google.com/drive/folders/1HaMBDyng2Pp0EbmpiFM3sH7Ir0cPgS9z?usp=sharing.
After downloading the models, put it under the path "./Model/".

3.Run testing

Some parameters need to be modified according to the testing dataset. And run run_testing.py to test the data.

Requirment

python 3.6
keras 2.1.6
tensorflow 1.13.1

Reference

Zechen Wei, Xiangjun Wu, Wei Tong, Suhui Zhang, Xin Yang, Jie Tian, and Hui Hui, "Elimination of stripe artifacts in light sheet fluorescence microscopy using an attention-based residual neural network," Biomed. Opt. Express 13, 1292-1311 (2022)