Learning a Coordinated Network for Detail-Refinement Multiexposure Image Fusion

Abstract

Nowadays, deep learning has made rapid progress in the field of multi-exposure image fusion. However, it is still challenging to extract available features while retaining texture details and color. To address this difficult issue, in this paper, we propose a coordinated learning network for detail-refinement in an endto-end manner. Firstly, we obtain shallow feature maps from extreme over/under-exposed source images by a collaborative extraction module. Secondly, smooth attention weight maps are generated under the guidance of a self-attention module, which can draw a global connection to correlate patches in different locations. With the cooperation of the two aforementioned used modules, our proposed network can obtain a coarse fused image. Moreover, by assisting with an edge revision module, edge details of fused results are refined and noise is suppressed effectively. We conduct subjective qualitative and objective quantitative comparisons between the proposed method and twelve state-ofthe-art methods on two available public datasets, respectively. The results show that our fused images significantly outperform others in visual effects and evaluation metrics. In addition, we also perform ablation experiments to verify the function and effectiveness of each module in our proposed method. The source code can be achieved at https://github.com/lok-18/LCNDR.

Image fusion using Y-net-based extractor and global-local discriminator

“…Consequently, some useful multi-layer information is lost in the deep cascaded network, resulting in unfriendly visual perception. In addition, some non-end-to-end methods [ 11 , [15] , [16] , [17] , 27 , 28 ] generate unsatisfied fusion results due to unreasonable fusion rules. To this end, in this work, we focus on developing more effective GAN frameworks that explicitly deal with the scale-space problems faced by visible and infrared image fusion task in an end-to-end fashion.…”

“…Zheng [ 13 ] achieved feature extraction at different scales and levels using HINBlock. Reference [ [14] , [15] , [16] , [17] ] employed convolution kernels of different sizes to extract common and unique features of source images. Reference [ [18] , [19] , [20] ] captured the multilevel features of the source images via residual learning.…”

Towards Robust Image Stitching: An Adaptive Resistance Learning against Compatible Attacks

“…In this setup, the three conventional down-sampling and three up-sampling encoding stages in the traditional U-Net are supplanted by the communal cells. To boost the robustness of our stitching framework, the communal cell is adapted with the similar candidate operator to the homography estimation and the inter connection is determined through the search learning (Li et al 2022(Li et al , 2023. A detailed depiction of the structure can be seen in Fig.…”

A 3D Cuboid Image Encryption Algorithm Based on Controlled Alternate Quantum Walk of Message Coding

“…Differential attack is to change specific elements of the plaintext image, corresponding to the degree of influence of different ciphertexts, to obtain as much of the key as possible. The ability of an encryption algorithm to resist differential attacks can be measured by two important parameters: number of pixel change rate (NPCR) and uniform average change intensity (UACI) [44], which are calculated as Equation ( 49)- (51):…”