基于稀疏表示和脉冲耦合神经网络的医学影像融合算法

为满足医学图像辅助诊断的需要，提出一种基于稀疏表示和脉冲耦合神经网络（PCNN）的CT和MR影像融合算法。首先，原始图像通过滑动窗方法构成联合矩阵，通过K-SVD算法得到该联合矩阵的冗余字典，采用正交匹配追踪算法得到该联合矩阵的稀疏系数；然后，根据稀疏系数的特点，采用脉冲耦合神经网络来融合稀疏系数；最后，由融合后的稀疏系数和冗余字典得到融合矩阵，反变换得到融合图像。实验图像为10组配准的脑部CT和MR图像，采用5种性能指标来评价融合图像的质量，同2种流行的医学影像融合算法进行比较，结果显示算法除Q^AB/F指数外，其他4项指标均为最优，Piella指数、Q^AB/F指数和BSSIM指数的均值分别为0.760 4、0.877 1和0.537 3，融合图像的纹理和边缘清晰，对比度高。主观和客观分析显示，算法的融合性能比较优越。

Medical Image Fusion Algorithm Based on Sparse Representation and Pulse Coupled Neural Network

A novel fusion algorithm for medical image based on sparse representation and pulse coupled neural network (PCNN) was proposed to meet the demand of computer aided diagnosis from the medical images. First, the K-SVD algorithm was used to obtain the redundant dictionary of the joint matrix which was obtained by sliding window technique. Next, sparse coefficients for the joint matrix were set up through orthogonal matching pursuit (OMP) algorithm. Then, the sparse coefficients were fused by a PCNN based on their characteristics. At last, the fused image was obtained by transforming the fused matrix which was got by the fused sparse coefficients and redundant dictionary. Ten groups of co aligned medical images were tested by experiments and the quality of the fused image was evaluated by five kinds of commonly used objective criterions. Comparing with the other two popular medical image fusion algorithms, the proposed algorithm was optimal for the four object indexes except for  Q^AB/Findex, the mean of Piella, Q^AB/F and BSSIM indexes were 0.760 5, 0.877 1 and 0.537 3 respectively. The texture, edge and contrast of fused image were optimal. Subjective and objective analysis of the results showed the advantages of the proposed algorithm.