Regular article
Infrared small target detection based on local intensity and gradient
properties
Hong Zhang, Lei Zhang, Ding Yuan
⇑
, Hao Chen
Image Processing Center, Beihang University, Beijing 100191, China
highlights
Two local properties of infrared small targets from the perspective of intensity and gradient are characterized.
The proposed model combines the intensity information and gradient information.
Based on the two properties, an infrared small target detection algorithm is proposed.
The proposed method yields good performance for suppressing background clutter.
article info
Article history:
Received 15 July 2017
Revised 25 December 2017
Accepted 29 December 2017
Available online 30 December 2017
Keywords:
Small target detection
Gradient
Intensity
Infrared images
abstract
Infrared small target detection is a challenging task for computer vision due to the factors such as scale
variations of the targets and strong clutters. Inspired by the Gaussian-like shape of the small target, we
characterize two local properties of the small target from the perspective of intensity and gradient to
address this problem. The two properties are that the intensity value of the target pixels is greater than
the value of its locally neighboring pixels, and the gradients towards the target center often distribute
regularly around the target. First, based on the two properties, the local intensity and gradient (LIG)
map is calculated from the original infrared image in order to enhance the targets and suppress clutters.
Next, we can obtain the targets conveniently via segmentation from the LIG map. Extensive evaluations
on real data demonstrate that the proposed algorithm has satisfactory results in terms of clutter suppres-
sion and robustness.
Ó 2017 Elsevier B.V. All rights reserved.
1. Introduction
Infrared small target detection is widely applied in practical
infrared imaging systems, especially in remote sensing and surveil-
lance. Infrared small targets usually occupy a few pixels in images,
and lack textural information. Since there is usually low contrast
between the small targets and backgrounds in infrared images,
the small targets are easily immersed in complex backgrounds
[1]. Besides, the size and brightness variations of infrared small tar-
gets, which are caused by imaging distances, thermodynamic
states of targets and environments, also raise the difficulty. Conse-
quently, infrared small target detection is still a valuable research
problem.
In the past decades, various methods have been proposed for
detecting the small target accurately and efficiently [2–4]. The
existing algorithms can be divided into three categories roughly.
The first category is the target-focused method that focuses on
the target characteristics and distinguishes the small targets from
infrared backgrounds. High-pass filters in frequency domain have
been applied to detect the targets in [5,6]. Wang et al. [5] devel-
oped two directional high-pass filtering templates by the least
squares support vector machine for the detection task. Kim [6]
decomposed the Laplacian of Gaussian filter into four filters and
applied the minimum filter to obtain the final spatial filtering
image. These methods focused on removing the low frequency
clutters, but failed to filter out noise and strong clutters in the
high-frequency components. A Facet-based model [7] was pro-
posed to detect the targets based on the idea that the maximum
extreme point was more likely the target, but their method
neglected to preserve the shape of targets. Inspired by the human
vision system (HVS), some researchers have proposed related
methods. Based on local contrast, Chen et al. utilized the brightness
discrepancy between the target and its neighboring area to detect
the target [1]. This method could work effectively, nevertheless, it
could be affected by noise of high brightness [8]. Some methods
https://doi.org/10.1016/j.infrared.2017.12.018
1350-4495/Ó 2017 Elsevier B.V. All rights reserved.
⇑
Corresponding author.
E-mail address: dyuan@buaa.edu.cn (D. Yuan).
Infrared Physics & Technology 89 (2018) 88–96
Contents lists available at ScienceDirect
Infrared Physics & Technology
journal homepage: www.elsevier.com/locate/infrared