中文摘要：

提出一种对图像矩阵中的每行（或每列）进行搜索，寻找出各行（或各列）相邻像素之间灰度差值的最大值与最小值，并将达到极值处的像素灰度值作为该行（或该列）的边界灰度值，使用得到的列边界灰度值或行边界灰度值计算出阈值从而提取红外图像边缘的算法；此外，利用黄金分割方法，并采用多阈值及限定区域的思想设计提取图像边缘的另一种算法；分析了文中算法的时间复杂度。实验结果表明，文中算法能够快速有效地提取红外图像的边缘，并能满足红外自动目标识别的实际需要。

英文摘要：

The searching algorithm of edge grey levels in column and row was proposed. The basic principle of the algorithm is as follows. Starting from the pixel of highest grey level and computing the disparity in grey level of each two adjacent pixels within the same column of image, the grey levels of pixels where the disparity reaches maximum and minimum are taken as the upper and lower thresholds of image edges in that column and that row. Thresholds are searched in each column and row of image. Besides that, the other edge detection algorithm was presented based on golden section, together with multi-threshold and limited area. Computational complexity of new algorithm was analyzed. Experimental results show that new algorithm extracts the edge of thermal image quickly and meets the practical demands of automatic target recognition.