掩膜边缘图及带掩膜图像窗口的统计特征快速算法  

Edge Image of Mask and Rapid Algorithm for Calculating Statistical Characteristics of Image Windows with Mask

作　　者：邵平[1] 王俊波[1] 

机构地区：[1]肇庆学院计算机学院,广东肇庆526061

出　　处：《肇庆学院学报》2015年第5期1-6,共6页Journal of Zhaoqing University

基　　金：广东省教育厅科学研究基金资助项目(201012MS185);广东省人才引进基金资助项目(2050205);肇庆学院教授科学研究启动基金资助项目(611271);肇庆学院教学研究基金资助项目(JGZB201201;sjjx201418);肇庆学院青年科学研究基金资助项目(201433)

摘　　要：应用图像窗口的统计特征进行多尺度目标检测时,往往需要计算所有可能窗口的灰度均值、方差、三阶中心矩等统计特征,十分耗时.应用传统的积分图虽然可以实现快速计算,但在某些需要用掩膜掩去窗口四周一些无关区域的应用场合,将无法使用积分图来加速,只能使用经典公式法.提出掩膜边缘图,并结合广义行积分图法,实现了一种带掩膜图像窗口的统计特征快速算法.该算法只需扫描1遍窗口,且扫描时不需要遍历窗口像素.实验表明,本算法的计算时间仅为经典公式法的1/4;若掩膜的形状具有对称性,则本算法的计算速度更快.While statistical characteristics of image windows are used to detect multi-scale objects, it is often very time-consuming to calculate these statistical characteristics such as the gray average, variance and third cen- tral moment for all possible image windows. The time of calculation can be greatly shortened with traditional in- tegral image. But in applications that masks are used to eliminate the irrelevant or easily disturbed areas around each window, rapid algorithm with integral image can＇ t be used anymore. There is no other way and only the classic formulas method can be used. The edge image of mask is presented, and a rapid algorithm is realized for calculating statistical characteristics of image windows with mask by combining it with the extended row integral image. Each image window is only needed to scan once by this algorithm, and not all the pixels of each window are needed to scan. Experiments show that the calculating time of this algorithm is only a quarter of the classic formulas method. When the shape of mask is symmetry, the calculating time of this algorithm can be decreased further.

关 键 词：掩膜边缘图 带掩膜图像窗口 统计特征 快速算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]