Infrared Small Target Detection Based on Spatial-Temporal Enhancement Using Quaternion Discrete Cosine Transform

Abstract

Infrared small target detection plays an important role in the infrared search and track system. However, infrared small target images often suffer from low contrast. In this paper, we propose an infrared small target detection method that improves the target contrast and suppresses background clutters based on spatial-temporal enhancement using the quaternion discrete cosine transform (QDCT). The proposed method is twofold: 1) we propose to detect the infrared small target by constructing the quaternion feature map for infrared images. The quaternion integrates four feature maps, including the kurtosis feature, two directional feature maps extracted by steerable filtering in spatial domain and motion feature in the temporal domain. 2) Then the quaternion is input into QDCT, and the saliency maps of each feature channel can be obtained by sign function processing. The final detection result is obtained by inverse QDCT to the quaternion. Compared to several state-of-the-art algorithms, the proposed method has a lower false alarm rate when the same positive detection rate is achieved. INDEX TERMS Infrared (IR) small target detection, kurtosis feature, steerable filter, quaternion discrete cosine transform (QDCT).The associate editor coordinating the review of this manuscript and approving it for publication was Sudipta Roy.

A Method for Infrared Sea-Sky Condition Judgment and Search System: Robust Target Detection via PLS and CEDoG

“…Under Sea-sky Background Extraction Direction Method [1]- [4] Wavelet Transform Method [5]- [7] Based on Gray and Contrast [8]- [9] Visual Attention Model [10]- [14] Multiscale Fuzzy Metric [15]- [16] To comprehensively show the superiority of the presented method, more recent infrared small target detection methods are systematically reviewed in the Table 2. Chen et al [17] applied a method based on local contrast, they used the brightness difference between the target and its neighboring area to detect the target.…”

“…Local Contrast Method [17]- [20] Spatio-temporal Saliency Approach [21]- [22] Low-rank Tensor Completion [23]- [24] Based on Deep Learning [25]- [29] It can be known from above scientific research situation that these current methods are more suitable for targets with a single feature in the entire image [30][31][32][33][34]. On the other word, sea-sky background clutter and wave noise in the space or transform domain are irrelevant to the target [35][36][37][38][39][40][41].…”

Infrared Dim and Small Target Detection Based on Greedy Bilateral Factorization in Image Sequences

“…In addition, IPI and NRAM also have high computational complexity. QDCT [26] is a sequential detection method. FKRW [11] based on image segmentation is one of the best methods at present.…”

“…A new method based on in-frame and interframe information is proposed [25]. A small IR target detection method based on spatial-temporal enhancement using the quaternion discrete cosine transform (QDCT) is proposed by fusing the kurtosis feature, two-directional feature maps, and motion feature of image sequences [26]. Similarly, temporal and spatial information of sequential images is used to extract objects of interest in [27] and [28].…”

Classification of Infrared Objects in Manifold Space Using Kullback-Leibler Divergence of Gaussian Distributions of Image Points

“…Many different approaches have been proposed for infrared object tracking such as saliency extraction [9], multiscale patch-based contrast measure and a temporal variance filter [14], feature learning and fusion, reliability weight estimation based on nonnegative matrix factorization [15], Poisson reconstruction and the Dempster-Shafer theory [16], three-dimensional scalar field [17], a double-layer region proposal network (RPN) [18], Siamese convolution network [19], a mixture of Gaussians with modified flux density [20], spatial-temporal total variation regularization and weighted tensor [21], two-stage U-skip context aggregation network [22], histogram similarity map based on the Epanechnikov kernel function [23], quaternion discrete cosine transform [24], non-convex optimization [25], Mexican-hat distribution of pixels [26], and Schatten regularization with reweighted sparse enhancement [27].…”