Medical Image Fusion Based on Adaptive Pulse Coupled Neural Networks and Nonsubsampled Contourlet Transform

Multi-modal medical image fusion can effectively integrate imaging information from multiple sensors and has become a research hotspot in medical image processing in recent years. In this paper, a new multimodal medical image fusion method based on adaptive pulse-coupled neural network (PCNN) and nonsubsampled contour wave transform (NSCT) is proposed. In the proposed, the NSCT decomposition is first performed on medical source images to obtain high-frequency subbands and low-frequency subbands. Secondly, the fusion strategy of phase coherence model for feature extraction of low frequency sub-bands can effectively retain the energy information of the source image. In addition, a new high-frequency subband fusion rule is designed to customize the decomposed high-frequency sublayer, and further optimized for the different frequency band characteristics of the decomposed obtained sublayer. And a new input parameter adaption is used to improve the PCNN model as a strategy for high frequency coefficient fusion. Finally,the fused image is reconstructed by performing inverse NSCT on the fused high-frequency and low-frequency bands. Experiments are conducted by performing on a publicly accessible brain image dataset. The experimental results show that the algorithm achieves positive performance in terms of visual quality and the objective indicators of overall Cross entropy (OCE), structural similarity (SSIM), feature mutual information (FMI) and visual fidelity (VIFF) have reached the best level.