基于多尺度局部方差的海面红外舰船检测

方差可以表示局部区域中各像素之间的关系：一般情况下，不同的像素相同的邻域大小计算出的方差不同，相同的像素不同的邻域大小计算出的方差也不同。本文构造多尺度局部方差，来计算每个像素的方差随尺度变化的情况。目标像素与其邻域像素的方差随尺度变化大，背景像素与其邻域像素的方差随尺度变化小；最后运用行列投影求出方差变化大的像素作为检测结果。实验结果表明，本文算法可以有效减弱背景中各种干扰的影响，正确检测出红外图像中舰船目标的位置。

Infrared warship target detection based on multi-scale variance

Variance can express the relationship among the pixels in a small area.Generally speaking,the variance of different pixel is different when the area size is same,and the variance of same pixel is different when the area size is different.In this paper,a Multi-scale Variance algorithm is proposed to compute the change of variances resulting from different scale.That change in the target is big and that change in the background is small.Finally,it uses rows and columns projection to find the big change as the target position.The experimental results show that the proposed algorithm is successful to suppress the noise and get the position of infrared warship target.