Infrared and visible image fusion via parallel scene and texture learning

作者：

Highlights：

• We propose an efficient IR/VIS fusion method based on spatially variant RNNs.

• It extracts coarse- and fine-grained information via scene and texture learning.

• It models the internal structures of images on detail branch in an explicit manner.

• Experiments are conducted to show the superiority of our method for IR/VIS fusion.

• Our method has the potential for improving the performance of object detection.

摘要

•We propose an efficient IR/VIS fusion method based on spatially variant RNNs.•It extracts coarse- and fine-grained information via scene and texture learning.•It models the internal structures of images on detail branch in an explicit manner.•Experiments are conducted to show the superiority of our method for IR/VIS fusion.•Our method has the potential for improving the performance of object detection.

论文关键词：Image fusion,Infrared,Scene and texture learning,Recurrent neural network

论文评审过程：Received 30 March 2022, Revised 30 June 2022, Accepted 21 July 2022, Available online 23 July 2022, Version of Record 28 July 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108929