目的 红外弱小目标检测是红外图像处理领域中难度大且实际意义相当重要的一项研究热点问题，其在侦察预警系统、飞行器跟踪系统与导弹制导系统中扮演了十分重要的角色。自然背景下的红外图像一般具有较低信噪比，其中背景占据着绝大部分面积，而目标则尺寸很小且不具有明显形状和纹理信息，这为红外图像中进行弱小目标检测增加了难度。本文提出一种将Facet方向导数特征与稀疏表示相结合的红外弱小目标检测算法。方法 首先利用Facet模型提取原红外图像在0°、90°、45°和-45°四个方向上的一阶导数特征，然后通过稀疏表示方法，在方向导数信息基础上对图像进行由上至下从左到右分块逐一处理，利用求解出的稀疏系数和导数图像块的重建残差构建检测数值图，最后分割出小目标所在具体位置。结果 通过对4组不同红外图像序列进行实验验证，绘制了检测率与虚警率ROC曲线图。从结果上可以看出，本文提出的算法相较于对比算法在小目标检测中具有较高检测率。结论 本文算法将Facet方向导数特征与稀疏表示相结合，在红外弱小目标检测上具有较高检测精度和较强抗噪声干扰能力，相比于传统检测算法具有一定优势，同时可根据不同检测背景训练出相应背景字典，从而得到较好检测效果，在实际工程应用中具有良好针对性。
Infrared small target detection algorithm based on derivative characteristics of Facet combined with sparse representation(NCIG2018)
Objective Infrared dim and small target detection conforms as a hot topic of research in the field of infrared image processing, which is very difficult but rather practical. It plays an important role in reconnaissance and warning systems, aircraft tracking systems and missile guidance systems as well. Generally speaking, the infrared image among the natural background has a low signal-to-noise ratio, wherein the background occupies most of the area, and the targets, however, come to be so small that they are short of shape and texture information, which makes it difficult to work out the detection of weak targets in infrared images. This paper proposes an infrared small target detection algorithm that combines Facet directional derivative features with sparse representation. Method Firstly, Facet model is used to extract the first derivative features of the original infrared image in four directions of 0°, 90°, 45°, and ?45°. Then, through sparse representation, the blocks separated from the image are processed from top to bottom and left to right based on the directional derivative information. The sparse coefficients and the reconstruction residuals of the derivative image blocks are used to generate a detected-value map. Finally the specific location of the small target is segmented. Result Through the experiments of four different infrared image sequences, the ROC(receiver operating characteristic) curves of detection rate and false alarm rate are plotted. From the results, we can see that the proposed algorithm has a higher detection rate than the comparison algorithm in infrared small target detection. Conclusion The algorithm proposed combines Facet directional derivative features with sparse representation, and owns high detection accuracy and great anti-noise ability in infrared small target detection. Compared with traditional algorithms, it has better performance obviously. In addition，its background dictionary can be trained according to varied circumstances so as to get a better effect, which may put on the pertinence in practical applications.