基于最简视觉显著性的红外目标快速提取方法

在红外深空目标跟踪系统中,为了能够从深空红外图像中快速提取微小目标,通过分析红外深空图像的特点,提出一种基于最简视觉显著性的红外目标快速提取方法。该方法在传统的视觉显著性的基础上,通过计算局部灰度最大值和目标像素的灰度平均值与邻域像素的灰度加权值的对比度组成特征向量,构造显著性模型,抑制背景并凸显目标,使之不但能够减少运算耗时,而且能够保证提取精度。通过对红外深空图像进行处理,实验结果表明该算法的运算时间仅为传统的视觉显著性算法的28%,且有较好的处理结果,证明了该算法的有效性。

Rapid infrared target extraction based on simplest visual saliency

In the system of infrared deep space target tracking,by analyzing the feature of the infrared images,a rapid infrared target extraction method based on simplest visual saliency is proposed to achieve fast infrared targets extraction.Compared with traditional visual saliency,this method mainly constructs saliency model through computing eigenvector constituted by regional maximum gray value and contrast between mean grey value of target and weight value of every pixels in neighborhood,which can restrain background and highlight targets.After processing the infrared deep space images,the experimental results show that the runtime of this method is only 28% of that of the traditional visual saliency,which proves the validity of this method.