In this paper, we provide a brand-new architecture to deal with fusion issues between infrared images and visible images. Our encoding network, in variance to standard and traditional network architectures, consists of encoding convolution layer and fusion convolution layers where the output from each layer is cascaded to the forthcoming layers. The convolution layers create feature maps, which are merged. This architecture extracts more key salient features from the input images as they are processed through the encoder and merges the extracted features; and eventually, a decoder reconstructs the merged image. The suggested fusion approach delivers an efficient and advanced performance when compared to existing algorithms.