基于修正后直方图不变矩的符号图像变化检测  

Detection of Changed Symbol Images Based on Revised Histogram Invariant Moments

作　　者：周军其[1] 薛存金[1] 孙家抦[1] 

机构地区：[1]武汉大学遥感信息工程学院,武汉市珞喻路129号430079

出　　处：《武汉大学学报（信息科学版）》2005年第2期150-153,共4页Geomatics and Information Science of Wuhan University

摘　　要：提出了基于修正后的直方图不变矩对符号图像进行变化检测的算法。该算法的主要思想是首先计算 图像直方图不变矩,然后用两幅符号图像之间的相关系数来修正,利用修正后的不变矩来检测符号图像变化。 该方法解决了Hu不变矩在检测符号图像变化时难以克服的灰度变化问题。结果表明,该算法对符号图像的 对比度和灰度有较强的抵抗能力,能够提高符号图像变化的检测率和控制误测率。As the Hu's invariant moments cannot deal with the changed gray during detecting changed image, a revised histogram invariant moment's algorithm is put forward based on the histogram invariant moments. The invariant moments of image histogram filtered is calculated and revised with the related coefficient of the two images. Two symbol images, a reference image and an aim image, are tested to detect the change. The experiment shows that the method can resist the effect of the brightness and the contrast change of images.

关 键 词：HU不变矩 修正的直方图不变矩 符号图像变化检测 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]