Source-linked AI summary
Fast and Robust Small Infrared Target Detection Using Absolute Directional Mean Difference Algorithm
Saed Moradi, Payman Moallem, Mohamad Farzan Sabahi
TL;DR
Small infrared target detection is difficult in noisy scenes containing high-intensity structural backgrounds, while existing approaches may neglect directional information or produce clutter responses. The paper proposes ADMD, a directional extension of AAGD that combines directional contrasts to suppress structural backgrounds. On real infrared images, the authors report effective target enhancement and background-clutter elimination, with ADMD strongly outperforming other methods in SCR and BSF.
Problem
Small infrared target detection is challenging in noisy, complicated backgrounds, and AAGD can intensify high-intensity edges and structural clutter.
Method
ADMD extends AAGD with directional processing and uses the minimum of eight directional contrasts to improve clutter rejection while preserving target enhancement.
Results
ADMD strongly outperforms other methods from both SCR and BSF points of view on the test images.
Takeaways & Limitations
Directional minimum selection suppresses structural backgrounds because small targets have positive contrast across directions while sharp edges can have zero contrast along the edge direction.
Takeaways & Limitations
Multi-frame alternatives require frame registration and cannot identify targets with radial motion, while AAGD can intensify structural clutter.
Abstract
from arXiv · showhide
Infrared small target detection in an infrared search and track (IRST) system is a challenging task. This situation becomes more complicated when high gray-intensity structural backgrounds appear in the field of view (FoV) of the infrared seeker. While the majority of the infrared small target detection algorithms neglect directional information, in this paper, a directional approach is presented to suppress structural backgrounds and develop a more effective detection algorithm. To this end, a similar concept to the average absolute gray difference (AAGD) is utilized to construct a novel directional small target detection algorithm called absolute directional mean difference (ADMD). Also, an efficient implementation procedure is presented for the proposed algorithm. The proposed algorithm effectively enhances the target area and eliminates background clutter. Simulation results on real infrared images prove the significant effectiveness of the proposed algorithm.
1. Introduction
Small infrared target detection remains difficult in noisy, complicated backgrounds, motivating robust single-frame methods and a directional ADMD algorithm.
- Small infrared target detection remains a major challenge when distant targets are embedded in noisy and complicated backgrounds.
- Machine-learning detection performance and false-alarm rates depend strongly on the training samples and features.
- Single-frame filtering methods are practical, but background-subtraction approaches can produce high false-alarm rates because of imperfect background estimation.
- Multi-frame methods require frame registration and cannot identify targets with radial motion because their locations do not change across successive frames.
- The paper presents a directional absolute directional mean difference algorithm to develop more robust and effective small-target detection.
2. Background and motivation
AAGD uses nested-window average differences to enhance targets and suppress noise, but structural edges can also generate strong responses. ADMD adds directional processing to improve clutter rejection.
- AAGD compares averages from nested target and background windows to construct a saliency map, using local averaging to suppress background noise.
- AAGD enhances low-contrast targets but can intensify high-intensity edges and structural clutter, including regions with either positive or negative contrast.
- ADMD incorporates a directional approach into AAGD to improve clutter rejection and background suppression.
- Figure 1 depicts the internal and external windows used by AAGD.
- Figure 2 contrasts a target with a high-intensity edge in a real infrared scenario.
3. The proposed algorithm
ADMD extends AAGD with directional processing and minimum-response selection to suppress structural clutter while preserving positive-contrast targets, then reformulates the computation for efficient implementation.
- Window construction: The method assigns nested target and background windows, divides the external window into eight directional cells, and computes their average intensities.The external window is three times the target-window size and is divided into nine equal cells, including the target cell.
- Directional design: ADMD uses directional information and minimum directional response because small targets have positive contrast in all directions, unlike structural clutter.The method combines directional AAGD responses and suppresses negative values to improve clutter rejection.
- Clutter rejection mechanism: On a sharp cloud edge, ADMD limits the red-rectangle response to 0.6, whereas AAGD reaches about 60 after normalization.The AAGD response is approximately 100 times larger in this example.
- Clutter rejection mechanism: AAGD intensifies sharp edges as edge contrast increases and can produce strong responses in non-target regions, including negative-contrast scenarios.When the target-window center lies exactly on the edge, AAGD produces a zero response, creating a dark region between parallel responses.
- Efficient implementation: To reduce computational complexity, the implementation obtains the maximum directional background value through a morphological dilation structural element instead of separately computing all directional filters.The single-scale formulation otherwise requires eight directional AAGD computations.
4. Simulation results
Simulation results on real infrared images show that ADMD improves target–background separation, suppresses clutter, and offers an efficient single-scale implementation.
- Detection ability analysis: Pre-thresholding results compare ADMD with Top-Hat, AAGD, LCM, LoG, and PCM across real infrared images.The comparison uses four scales for AAGD, LCM, PCM, and ADMD, while MS-LoG uses 12 scales.
- Detection ability analysis: ADMD strongly outperforms the comparison methods on both SCR and BSF across the test images.These metrics quantify target enhancement and background-clutter rejection, respectively.
- Detection ability analysis: ADMD has the lowest false-alarm rate versus global threshold among the evaluated algorithms.The false-alarm curves cover six test images and assess post-thresholding clutter rejection.
- Execution time analysis: The efficient ADMD implementation is almost 27 times faster than original ADMD, although no evaluated multi-scale method meets real-time requirements on the reported hardware.For single-scale processing, TopHat, AAGD, and ADMDeff can operate in real time under the tested setup.
5. Conclusion
The conclusion presents ADMD as a directional method for robust small-target detection and improved background suppression. Its minimum directional-contrast selection preserves target enhancement while eliminating structural clutter responses.
- 5. Conclusion: ADMD uses a directional approach to improve suppression of structural backgrounds in small infrared target detection.The method is motivated by the need for robust detection in difficult infrared scenes.
- 5. Conclusion: ADMD selects the minimum contrast across eight directions because small targets have positive contrast in every direction.For isotropic small targets, the directional contrasts are nearly equal, so minimum selection does not reduce target enhancement.
- 5. Conclusion: Structural backgrounds respond differently by direction, allowing ADMD to suppress them while retaining small-target responses.A sharp edge has zero contrast along the edge direction, which supports its elimination by minimum selection.