[Paper Review] Fast and Robust Small Infrared Target Detection Using Absolute Directional Mean Difference Algorithm

Saed Moradi, Payman Moallem+1 more|arXiv (Cornell University)|Oct 7, 2018

Infrared Target Detection MethodologiesEngineering44 references165 citations

TL;DR

This paper proposes the Absolute Directional Mean Difference (ADMD) algorithm for fast and robust detection of small infrared targets in complex backgrounds. By computing directional contrast differences across eight directions and applying minimum selection, ADMD suppresses structural clutter while enhancing targets. The method achieves significantly higher signal-to-clutter ratio (SCR) and background suppression factor (BSF), with a 27× speedup via an efficient implementation, outperforming AAGD and other baselines in real-time performance and false alarm reduction.

ABSTRACT

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.

Motivation & Objective

• Address the limitation of existing small infrared target detection methods in suppressing high-intensity structural backgrounds and edges.
• Overcome the false alarm problem in AAGD when structural clutter has high local contrast.
• Develop a directional approach that enhances target detection by exploiting isotropic contrast properties of small targets.
• Ensure real-time applicability through an efficient algorithmic implementation with minimal computational overhead.
• Demonstrate superior performance over AAGD and other baseline methods in both quantitative metrics and visual results.

Proposed method

• Compute directional contrast differences using eight 3×3 directional cells around a central pixel, replacing the single-window AAGD approach.
• Apply absolute difference between the central cell and each of the eight surrounding directional cells to compute directional mean differences.
• Use minimum selection across all eight directional contrasts as the final output to suppress structural backgrounds with inconsistent directional responses.
• Formulate the ADMD as a combination of local averaging and absolute difference operations to ensure low computational complexity.
• Introduce an efficient implementation (ADMDeff) using morphological dilation with a specific 3×3 cross-shaped structuring element to accelerate computation by 27×.
• Leverage multi-scale processing with [3×3, 5×5, 7×7, 9×9] windows to handle varying target sizes across different image scales.

Experimental results

Research questions

• RQ1 Can directional contrast analysis improve background suppression in small infrared target detection compared to non-directional methods like AAGD?
• RQ2 Does minimum selection across directional contrasts preserve target enhancement while eliminating structural clutter with inconsistent directional responses?
• RQ3 Can the proposed ADMD algorithm achieve real-time performance despite increased computational load from eight directional operations?
• RQ4 How does the proposed method compare to AAGD and other baselines in terms of signal-to-clutter ratio (SCR) and background suppression factor (BSF)?
• RQ5 What is the impact of the efficient implementation (ADMDeff) on execution time and feasibility for real-time deployment?

Key findings

• ADMD achieves a background suppression factor (BSF) of 114.23 on image 9a, significantly outperforming AAGD (8.66) and other baselines.
• On image 9aj, ADMD achieves a BSF of 45.29 and SCR of 14.77, demonstrating superior clutter rejection in complex scenes.
• The false alarm rate (Pfa) of ADMD is consistently lower than all other methods across all test images, especially at low threshold levels.
• The efficient implementation (ADMDeff) reduces execution time by approximately 27× compared to the original ADMD algorithm.
• Single-scale ADMD and ADMDeff achieve real-time performance on the given hardware, with ADMDeff being suitable for practical deployment.
• In scenarios with homogeneous background and noise (e.g., image 9v), ADMD performs comparably to AAGD, confirming no degradation in simple conditions.