基于梯度融合规则的医学图像融合方法*

为了更好地满足临床诊断和治疗的需要，本文提出了一种在图像融合阶段对测量值进行自适应梯度加权和图像重建时采用CoSaMP重建算法相结合的方法。该算法首先对两幅源图像分块并进行稀疏表示，同时利用观测矩阵进行测量。在测量数据融合阶段引入图像梯度来反应图像本身的边界信息，先计算每幅分块子图像的梯度；然后利用自适应梯度加权的融合规则得到融合的测量数据，并对融合测量数据进行随机压缩采样；最后通过CoSaMP算法对采样数据进行信息重构实现测量数据的恢复。该方法克服了图像融合时信息畸变的缺陷，并且可以根据不同融合区域自动调整融合规则的权重系数,有效地避免了设置固定权重系数造成的融合误差。实验结果和评价指标验证了本文算法的有效性和先进性。

Medical image fusion method based on gradient fusion rules

To better meet the demands of clinical diagnosis and treatment, we propose a method of combining the adaptive gradient weighting of the measured values and the CoSaMP reconstruction algorithm. Firstly, the source images are compressed by compressive sensing. In the measured data fusion phase, the gradient-based weights are obtained by compositing the gradients of each image, and then the measurement matrix of the fused image are obtained by the self-adaptive gradient . Finally, it uses a CoSaMP-based image reconstruction algorithm to achieve the recovery of the data. The method overcomes the defects of the information distortion in the image fusion, and can automatically adjust the weight coefficient of the fusion rules according to the different fusion areas.The experimental results and evaluation results show that the proposed algorithm is effective and advanced.