基于向量小波多尺度纹理分析的红外小目标检测

Multiwavelet Multi-Resolution Texture Analysis Based Small Det ection in Infrared Image

向量小波具有一般标量小波所不具备的许多优点。以背景纹理分析为基础 ,首先利用向量小波对不同尺度下各通道图像进行分解 ,然后利用能量法提取其局部纹理特征 ,并计算各点的特征向量与中心向量间的距离 ,即可得到一个相关的多尺度距离像 ,根据该距离像进行直方图统计 ,从而实现目标的检测。实验表明 ,本方法对小目标的检测取得了较好的效果 ,检测率高 ,且能较精确地检测出小目标的位置。尤其是本方法所具有的运行速度快 。

Multiwavelt has many good qualities. Through the ba ckground texture analysis, this paper picks up the part texture characteristic f rom the maps after multi-scale multiwavelet, and calculates the distance betwee n characteristic vector and center characteristic. This can gain a correlative m ulti-scale distance map. The histogram of distance map is employed for the auto matic selection of threshold value. Experiments show that this approach can achi eve quite satisfactory-results, and spends less time of experiment.