为解决天空背景下红外弱小目标检测问题，提出了一种基于图像复杂度和方向梯度的检测方法。利用信息熵对图像复杂度进行描述，引入了图像方差和像素局部变化率对信息熵进行加权，使云内部和云边界区域得到抑制。以复杂度为描述对象，建立多级多方向梯度模型，在背景局部复杂度高于目标复杂度的情况下，仍能够有效分割出目标。实验证明，该方法能够在复杂云背景情况下检测出弱小目标。

In order to solve the problem of detecting small dim targets in infrared images with the complicated air background, an improved algorithm based on image complexity and orientation gradient is put forward. The information entropy reflects the complexity of the image. The weighted information entropy with the variance and the change rate of local pixels suppresses both the inside and the edge of clouds. Since the image of the complexity is obtained, targets can be extracted by the means of multi-degree and multi-orientation gradient under the circumstance that the complexity of the background is higher than that of the targets. Experiments indicate that the algorithm can detect small dim targets effectively with complicated cloud background.