Refined Multi-focus Image Fusion Technique Utilizing Pulse Coupled Neural Networks

Conventional multi-focus image fusion techniques often fail to adequately capture and integrate image characteristics, leading to unclear boundaries and a decrease in texture clarity. This research introduces a novel fusion technique utilizing the Non-Subsampled Shearlet Transform (NSST) alongside an advanced Pulse-Coupled Neural Network (PCNN) framework. The first phase comprises applying the NSST algorithm to the grayscale image, thereby dividing it into its high-frequency and low-frequency components. For the low-frequency component, a fusion criterion is implemented that combines regional variance and regional energy through their product. Meanwhile, the other high sub-band is processed using an improved adaptive PCNN fusion method. The method is that the PCNN parameters are adaptively set by the Grey Correlation Degree. The grey correlation matrix is used as the link strength of the PCNN and MSMG is used to regulate the PCNN. The final stage of the process involves the application of the NSST inverse transform to obtain a fused image. Compared with alternative fusion methods, this method has a high fusion evaluation index, clear images.