基于压缩感知和NSCT-PCNN的PET/CT医学图像融合算法

PET / CT Medical Image Fusion Algorithm Based on Compressive Sensing and NSCT-PCNN

针对非下采样Contourlet变换(NSCT)后计算复杂度高以及医学融合图像质量差等问题,提出一种基于压缩感知和脉冲耦合神经网(PCNN)的图像融合方法。首先将源图像进行NSCT单层分解;其次,对计算量较大的高频子带采用高斯随机测量矩阵进行压缩测量,融合规则选用绝对值取大的方法,对融合后的高频图像采用正交匹配追踪算法(OMP)进行重构;然后对低频子带采用基于PCNN的融合规则,将低频子带系数作为信号激励PCNN网络,根据低频图像的特性选择较大点火次数的系数作为低频子带融合系数;最后对高频融合图像和低频融合图像通过NSCT逆变换,得到最终的融合图像。实验结果表明:该算法无论从人眼视觉效果还是客观评价指标上均优于其他算法,且具有较强的鲁棒性。

For the high computation complexity and poor medical fusion images under non-subs ampled contourlet transform（ NSCT）,a method of image fusion based on compressed sensing and PCNN was proposed. Firstly,the source images were decomposed in monolayer with NSCT. Secondly,the Gauss random matrix in high frequency which has large calculation was used for compression measure-ment,and a method based on maximum value was utilized to fuse the high-frequency components respectively. High frequency measurement was reconstructed using the orthogonal matching pursuit（ OMP） method after fusion. Third,a fusion rule based on PCNN was adopt in low frequency subband coefficient,and it was input into PCNN network as the signal,and we chose the bigger ignition frequency coefficient as the fusion low-frequency subband coefficients according to the characteristics of low frequency images. Finally,the final fusion image was acquired through the NSCT inverse transformation. The experiment results show that the proposed algorithm is superior to other algorithms in both the visual effect and the objective evaluation index,and it has strong robustness.