摘要
提出一种采用剪切波变换的红外弱小目标图像预处理新方法.该方法首先采用剪切波变换对图像进行分解,针对高低频子带的不同特点,对低频子带系数采用基于贝叶斯统计的方法进行处理,推导出剪切波系数为拉普拉斯先验分布的最大后验估计表达式和低频子带阈值;而高频子带则对其系数构成分析后,利用图像不同组成元素能量值的不同设定阈值;对处理后的子带系数进行重构以获得预测的背景图像,将其与原图相减,可得到突出目标且背景被抑制的图像;采用经典的自适应阈值分割法对预处理后的图像进行分割,很好地实现了目标检测.实验结果表明,该方法处理的图像信噪比值高,在客观评价指标与主观视觉两方面均表现出良好的效果.
An image pre-processing method based on shearlet transform was proposed. First, image was decomposed by shearlet transform, then to get the subband coefficients, Bayesian statistics and threshold shrinkage methods were used in the low-frequency component and high--frequency components, through shearlet reconstruction, the predicted background of the image was obtained, the original image is subtracted by the predicted background image and we can get the result image, Finally, in order to test the new method, classical adaptive threshold segmentation was used to segregate the targets. Theoretical analysis and experimental results show that, compared with several commonly used methods , ISNR of result image which using the new method is higher than the others, and have better effects both in objective evaluation and the subjective vision.
出处
《微电子学与计算机》
CSCD
北大核心
2013年第4期5-9,共5页
Microelectronics & Computer
基金
国家自然科学基金项目(60902067)
关键词
红外图像预处理
剪切波变换
贝叶斯统计
阈值收缩
infrared image pre-processing
shearlet transform
Bayesian statistics
threshold shrinkage
