Medical image fusion using fuzzy adaptive reduced pulse coupled neural networks

This paper addresses a novel neuro-fuzzy-based approach to set the weighted linking strength of parameter adaptive reduced pulse coupled neural networks. In reduced PCNN based medical image fusion algorithms, it is quite essential to evaluate the prominence of each pixel in an image. The fusion performance in turn depends on the linking factor, internal activity. Thus, we need to set these values of reduced PCNN in a more adaptive manner with fewer complications and uncertainties. For this, the weighted linking strength i.e., lambda of the reduced PCNN neurons is attentively set by a fuzzy-based approach. Here, lambda of neurons is represented as fuzzy membership values using the activity level measures such as local information entropy and energy. Finally, a new model called-Fuzzy adaptive reduced pulse coupled neural networks is developed by reducing the number of parameters and fuzzy adaptive settings of them. This leads to a very less complicated network and more computational efficacy, which is a prominent part of health care requirements. The proposed scheme is free from the shortcomings such as loss of boundaries, structural details, unwanted artifacts, degradations, etc. Subjective and objective evaluations show better performance of this new approach compared to the existing techniques.