514 IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 17, NO. 3, MARCH 2020
Infrared Small Target Detection Using
Homogeneity-Weighted Local Contrast Measure
Peng Du and Askar Hamdulla
Abstract— Detecting small targets in infrared (IR) image
sequences is an important task in IR guidance systems. The
clutter of complex backgrounds often submerges small targets,
making detection difficult. Achieving high detection and low false
alarm rates with complex backgrounds is a primary problem.
We propose an IR small target detection method using our
new homogeneity-weighted local contrast measure (HWLCM).
Inspired by the ability of the human visual system (HVS) to
determine saliency characteristics, we implement our method to
use the local contrast features of the central and surrounding
regions and the weighted homogeneity characteristics of the
surrounding regions to enhance the target while suppressing the
complex background. Our method divides each image into blocks
with a sliding window for which the HWLCM is calculated. The
HWLCM enhances the actual target and suppresses interference
simultaneously. We apply an adaptive threshold to target region
extraction to further refine the results. Our experimental results
show that our proposed method is more effective than six com-
parable methods, especially in terms of the signal-to-clutter gain
(SCRG) and background suppression factor (BSF) indicators.
Index Terms— Homogeneity-weighted local contrast measure
(HWLCM), human visual system (HVS), infrared (IR) small
target.
I. INTRODUCTION
I
NFRARED search and tracking (IRST) technologies are
widely used in a variety of fields including termin al pre-
cision guidance, early warning systems, space-based surveil-
lance, and geological analysis [1], [2]. The main challenge
in these areas is the need to have prior knowledge about
the size, shape, and textural features when tracking small
targets [3]. In addition, the presence of complex background
edges or highlight background areas often overpower weak
target intensities or hide targets in clutter and noise. Thus, it is
very difficult to detect and separate infrared (IR) small targets
under low signal-to-clutter ratio (SCR) and complex back-
ground conditions. Improving the detection rate and reducing
the false alarm rate (FAR) for small targets h as attracted
significant research attention.
Target detection methods are divided between
single-frame [3]–[5] and multi-frame [6] methods. Usually,
multi-frame detection algorithms that build on single-frame
detection are not suitable for stable detection on some
high-speed motion detection platforms [7]. Therefore,
Manuscript received March 14, 2019; revised May 27, 2019; accepted
June 6, 2019. Date of publication July 3, 2019; date of current version
February 26, 2020. This work was supported by the National Natural
Science Foundation of China under Grant 61563049. (Corresponding author:
Askar Hamdulla.)
The authors are with the Institute of Information Science and Engineer-
ing, Xinjiang University, Ürümqi 830046, China (e-mail: hysddp@163.com;
askar@xju.edu.cn).
Color versions of one or more of the figures in this letter are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/LGRS.2019.2922347
high-performance single-frame algorithms have received the
most interest.
Existing state-of-the-art single-frame detection methods for
small targets fall into two classes. The first consists of
target-based methods that analyze the target characteristics
and separate the target from the background by enhancing
the target. Enhancement methods include the Laplacian of
Gaussian (LoG) filter [8], high-pass filters, the difference
of Gaussian (DoG) filter [9], and the improved difference
of gabor (IDoGb) filter [10]. These methods remove low-
frequency clutter but do not filter noise and strong clut-
ter effectively in the high-frequency range. A facet-based
model detects targets using extreme value theory, but it is
ineffective with weak targets near the edges of a bright
background. Several algorithms implement aspects of the
human visual system (HVS). These include the local con-
trast measure (LCM) [2], the improved LCM (ILCM) [11],
the novel LCM (NLCM) [12], the multi-scale patch-based
contrast measure (MPCM) [ 13], the multi-scale local homo-
geneity measure (MLHM) [4], and the multi-scale relative
LCM (RLCM) [5]. These algorithms perform well with mildly
fluctuating backgrounds but have poor detection rates and
robustness with complex backgrounds because these algo-
rithms either remove the background or enhance the target
instead of doing the two things at once.
The second class consists of background-based methods that
utilize predictions or estimations of background. The poten-
tial target is detected by calculating the differences between
the original image and the estimated background using an
algorithm such as the top-hat transform. Bai and Zhou [14]
eliminated backgrounds with a new top-hat transformation,
morphology filtering, max-median or max-mean filters, and
Bayes estimation. These background-based methods are sen-
sitive to noise or the structural elements’ shape and size.
In this letter, we propose a new approach for IR detection of
small targets inspired by the HVS and called the homogeneity-
weighted LCM (HWLCM) as shown in Fig. 1.
Compared with the MLHM algorithm, which only calcu lates
the homogeneity of the central region of the target and does
not enhance the target effectively, our approach both enhances
targets and suppresses backgrounds.
We characterize two properties from the local contrast and
homogeneity to suppress all types of background clutter and
enhance the target simultaneously.
For contrast, we define a formula to suppress the back-
ground and enhance the target by taking advantage of the ratio
of the gray differences between the central and surrounding
regions. In addition, in most cases, we take advantage of the
fact that the surroundings of the target tend to be homoge-
neous, and thus, we define the mean of the eight-direction
standard deviation to suppress the complex background and
enhance the target further. We point out that this aspect of
1545-598X © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: Xinjiang University. Downloaded on March 23,2020 at 10:46:44 UTC from IEEE Xplore. Restrictions apply.
资源评论
