摘要
针对多聚焦图像的特点,提出了一种基于清晰度计算的非抽样轮廓波变换(Non-SubsampledContourlet Transform,NSCT)域多聚焦图像融合算法。该算法首先对源图像进行NSCT分解,以此克服传统Contourlet变换不具平移不变性的缺点。在分析光学成像中散焦表现形式的基础上,对分解后的低频子带和高频方向子带分别以"邻域梯度"及"合成邻域模值"作为清晰度指标,采用自适应选择法实现对多聚焦图像的融合处理。实验结果表明,该方法不仅能有效融合图像中的"伪影"和"振铃效应",视觉效果明显优于传统小波和Contourlet方法,且融合图像的熵、交叉熵及均方根交叉熵等客观评价指标也有明显提高。
According to the feature of multi-focus image, a multi-focus image fusion algorithm based on the calculation of clarity in non-subsampled contourlet transform （NSCT） domain is proposed. This algorithm firstly decomposes the source images into low-frequency subbands and detail subbands in the first place. Then it analyzes the presentation of defocus ira- age. We adopt neighborhood gradient and synthetic neighborhood modulus values respectively as the clarity index to the low and high frequency subbands via the adaptive selection. The experimental results show that the method not only eliminates the artifact and ringing effect of traditional wavelet and contourlet method, but also the objective evaluation （entropy, cross entropy and root cross entropy） elevates obviously.
出处
《光学与光电技术》
2010年第2期7-10,共4页
Optics & Optoelectronic Technology
基金
重庆市自然科学基金(CSTC2009BB2188)资助项目
关键词
图像融合
多聚焦图像
非抽样轮廓波变换
邻域合成模值
image fusion
multi-focus image
non-subsampled contourlet transform
synthetic neighborhood modulus
