基于人眼视觉特性与自适应PCNN的医学图像融合算法

发布时间：2018-03-17 06:10

本文选题：医学图像融合　切入点：非下采样Contourlet变换(NSCT)　出处：《光电子·激光》2017年07期 　论文类型：期刊论文

【摘要】：针对多尺度变换的图像特征,提出了一种基于人眼视觉特性与自适应脉冲耦合神经网络(PCNN)的医学图像融合新方法。首先,对经配准的源图像进行非下采样Contourlet变换(NSCT),得到低频、高频子带系数;然后,考虑到低频子带系数中保留了绝大部分源图像能量和图像轮廓特征,提出区域能量(RE)和梯度奇异值度量(GSVM)相结合的方法;考虑到图像全局特征,将PCNN用于高频子带系数中,提出区域视觉对比度(SLVC)模拟人眼视觉特性作为PCNN的外部刺激输入,设定PCNN的链接强度随视觉对比敏感度(VCS)自适应变化,同时考虑到PCNN的迭代次数,利用Sigmoid函数计算其点火输出幅值的显著性度量;最后,对获得的融合系数进行逆NSCT得到融合图像。通过实验对比分析表明,本文算法不仅可以保留源图像信息的同时,还得到较好的客观评价指标和视觉效果。
[Abstract]:A new method of medical image fusion based on human visual characteristics and adaptive impulse coupled neural network (PCNN) is proposed for image features of multi-scale transform. Firstly, a non-downsampling Contourlet transform is applied to the registered source image to obtain low frequency. Then, considering that most of the source image energy and image contour features are retained in the low frequency subband coefficients, a method combining regional energy with gradient singular value measurement (GSVM) is proposed, and considering the global features of the image, Using PCNN in high-frequency subband coefficients, a regional visual contrast scale (RVVC) is proposed to simulate human visual characteristics as the external stimulus input of PCNN. The link strength of PCNN is set to change adaptively with visual contrast sensitivity (VCSs), and the number of iterations of PCNN is taken into account. The Sigmoid function is used to calculate the significance measure of the ignition output amplitude. Finally, the fusion image is obtained by inverse NSCT of the obtained fusion coefficient. The experimental results show that the proposed algorithm can not only preserve the source image information, but also preserve the source image information at the same time. Also get better objective evaluation index and visual effect.
【作者单位】：浙江工商大学信息与电子工程学院;浙江理工大学自动化研究所;
【基金】：国家自然科学基金(61374022)资助项目
【分类号】：TP391.41

【相似文献】