Multispectral Image Fusion using Deep Neural Network A Novel Approach

Abstract

Polychromatic imaging is widely used in various fields including remote sensing, medical imaging, and industrial inspection. However, due to the limitations of imaging sensors, polychromatic images often suffer from low geospatial definition and poor optical fidelity. To address these issues, we propose a novel method for polychromatic image fusion based on deep neuronal system (DNN). Our method involves building a training set of high-definition and low-definition image blocks, utilizing an improved sparse denoising self-encoding encoder learning training neuronal system model for pre-training, and finely adjusting the frameworks of the DNN model. The proposed method is capable of preserving both the high geospatial definition and optical information of the polychromatic image. Experimental results demonstrate that our method outperforms state-of-the-art methods in terms of both objective and subjective quality measures.