摘要
分析了双门滤波算法抑制噪声能力强的特性,针对其在计算目标门和背景门均值和方差时计算量大的缺点,提出了一种快速双门滤波算法,该方法利用两个与图像一样大的二维数组来保存中间结果,这两个数组的元素值分别代表图像左上角所有像素的和与平方和,则计算目标门和背景门的均值和方差时可将M×N次加法和一次除法简化为两次减法、一次加法和一次除法运算,从而大大减少运算量,本算法的计算量从理论上讲仅为传统双门滤波运算量的3/M×N,且计算量不会随目标门、背景门尺寸的增加而增加(M、N分别代表背景门的高度和宽度)。运用快速双门滤波抑制背景和噪声,初步检测出目标,然后再利用目标运动的连续性剔除虚假目标,实现运动目标的最终检测。试验结果表明:该算法能够对小目标进行有效的检测。
The double-gate filtering algorithm is powerful to restrain noise yet time consuming. In this paper, a fast double-gate algorithm is proposed. This algorithm stores temporary results using two arrays which are the same size with the image to represent the sum and square sum of all left-top pixels respectively. Therefore, the operation process of mean and variance computing for target windows and background windows are greatly reduced from MxN additions and one division to two subtractions, one addition and one division. Theoretically, the computation is only 3/MxN compared with traditional algorithm and not increasing with the increase of target and background window's size (M and N are the height and width of the background window respectively). After using double-gate filtering algorithm to restrain background and noise and to detect targets roughly, the continuity of moving target is used to eliminate false targets and to realize target detection. Experimental results show that this algorithm is able to detect small target effectively.
出处
《光电工程》
EI
CAS
CSCD
北大核心
2007年第3期16-18,25,共4页
Opto-Electronic Engineering
关键词
目标检测
目标识别
双门滤波
形态学滤波
信噪比
Target detection
Target recognition
Double-gate filter
Morphological filter
Signal-noise ratio