Abstract
In the fields of pattern recognition and image processing, deep learning (DL) technology has been intensively studied. According to the characteristics of multi-modal medical images, medical diagnostic technology, and practical implementation, according to the practical requirements for medical diagnosis, a multi-mode medical image fusion with DL will be suggested. In this study, we present a straightforward but efficient deep learning medical image fusion technique based on convolution neural network with three input images fused by three distinct conventional methods to make use of the gradient information to achieve the multi-exposure image composition in various medical image techniques. The proposed technique is capable of creating a pleasing tone mapped-like high resolution image from many input images utilizing various existing methods by seamlessly compositing them under the direction of gradient-based quality assessment. Particularly, this technique was created based on observations of gradient changes between various exposures and is utilized to improve the fused image instead of other conventional techniques. The usefulness of the suggested approach is demonstrated through experiments with various medical imaging techniques.