摘要
对于星空观测CCD图像,除星空背景成像为大面积起伏背景噪声外,其余均为点状或近似点状小目标,分析背景的统计特性并建立起精确的数学模型来描述图像中的起伏背景,对小目标检测、识别是至关重要的。首先结合实际图像分析了星空图像的性质。然后,提出了一种基于局部直方图高斯拟合的背景参数估计方法,并利用KL散度定量的衡量估计直方图与原始图像直方图之间的相似度。实验结果证明,算法能够克服恒星的干扰,对星空背景图像的统计特性进行精确的估计。
For the CCD star image, except for the wide undulate background, the other part image is point-like or approximate point-like object. It is an essential !mportant for target detection and identification to analyze the statistics characteristic of background image and establish a precious model to describe background image. The characteristics of star-sky image are investigated. Then, an algorithm based on local-histogram Gaussian fitting is proposed to estimate the background parameters. Finally, the KL divergence is used to evaluate the fitness of original histogram and the estimated histogram. The results show that the algorithm is accurate and can get over the influence of stars.
出处
《红外技术》
CSCD
北大核心
2008年第4期230-233,共4页
Infrared Technology
关键词
直方图
高斯拟合
星图
KL散度
histogram
gaussian fitting
star-sky image
KL divergence