A Hazy Image Database with Analysis of the Frequency Magnitude

The database is available here: Download

Abstract

Image haze removal has been extensively studied, but there has been no such an image database regarding the haze effect. As a result, it is hard to carry out fair comparison, since different haze removal methods may be useful for different types of images, and meanwhile, it is not convenient for readers to verify the statistical priors that are widely used for haze removal. To solve these problems, we collected a database of more than 4,000 images of different kinds of outdoor scenes. The images of this database are grouped into 4 classes. Along with the database, we also present the analysis and comparison of the amplitude spectra of these 4 classes. We call this the outdoor image database with haze classified. Our purpose is to help develop image haze removal, as well as verify and discovery statistical priors of images.

 

BibTex

@article{article,
author = {Wang, Shuhang and Tian, Yu and Pu, Tian and Wang, Patrick},
year = {2017},
month = {09},
pages = {},
title = {A Hazy Image Database with Analysis of the Frequency Magnitude},
booktitle = {International Journal of Pattern Recognition and Artificial Intelligence}
}

Database

Our image database consists of 2000 high quality haze-free images and 1513 hazy images. Note that the total hazy image number in the paper (1464) is not correct. We classify these images into 4 classes by subjective assessment.

haze-free images:

 

images of light haze:

 

images of mild haze:

 

images of dense haze:

 

Advertisements
Categories:

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s