FDNet: An end-to-end fusion decomposition network for infrared and visible images

Jing Di; Li Ren; Jizhao Liu; Wenqing Guo; Huaikun Zhange; Qidong Liu; Jing Lian

doi:10.1371/journal.pone.0290231

Peer Review History

Original SubmissionApril 5, 2023
22 May 2023 Decision Letter - Chiranjibi Sitaula, Editor PONE-D-23-09775FDNet:An end-to-end fusion decomposition network for infrared and visible imagesPLOS ONE Dear Dr. Ren, Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Please submit your revised manuscript by Jul 06 2023 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file. Please include the following items when submitting your revised manuscript: A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter. If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols. We look forward to receiving your revised manuscript. Kind regards, Chiranjibi Sitaula Academic Editor PLOS ONE Journal Requirements: When submitting your revision, we need you to address these additional requirements. 1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf 2. Please note that PLOS ONE has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, all author-generated code must be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse. 3. We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match. When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section. 4. Thank you for stating the following in your Competing Interests section: "we no competing interests exist" Please complete your Competing Interests on the online submission form to state any Competing Interests. If you have no competing interests, please state "The authors have declared that no competing interests exist.", as detailed online in our guide for authors at http://journals.plos.org/plosone/s/submit-now This information should be included in your cover letter; we will change the online submission form on your behalf. 5. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide. 6. PLOS requires an ORCID iD for the corresponding author in Editorial Manager on papers submitted after December 6th, 2016. Please ensure that you have an ORCID iD and that it is validated in Editorial Manager. To do this, go to ‘Update my Information’ (in the upper left-hand corner of the main menu), and click on the Fetch/Validate link next to the ORCID field. This will take you to the ORCID site and allow you to create a new iD or authenticate a pre-existing iD in Editorial Manager. Please see the following video for instructions on linking an ORCID iD to your Editorial Manager account: https://www.youtube.com/watch?v=_xcclfuvtxQ Additional Editor Comments: The paper requires clarity in the presentation and novelty based on the reviewers' suggestions. Please address them one by one. Also, the following papers describe the CBAM attention's application in different areas, so the discussion of these works could enhance the readability of the paper as the current work is based on CBAM attention mostly. https://link.springer.com/article/10.1007/s10489-020-02055-x https://arxiv.org/pdf/2305.00679.pdf Reviewers' comments: Reviewer's Responses to Questions Comments to the Author 1. Is the manuscript technically sound, and do the data support the conclusions? The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented. Reviewer #1: Partly Reviewer #2: Yes ******** 2. Has the statistical analysis been performed appropriately and rigorously? Reviewer #1: No Reviewer #2: Yes ****** 3. Have the authors made all data underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified. Reviewer #1: Yes Reviewer #2: Yes ****** 4. Is the manuscript presented in an intelligible fashion and written in standard English? PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here. Reviewer #1: No Reviewer #2: Yes ****** 5. Review Comments to the Author Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters) Reviewer #1: The manuscript proposes a novel infrared and visible image fusion algorithm based on fusion decomposition network. Although the experimental results show certain advantages over the compared methods, further improvement and validation are needed to make the proposed algorithm more robust and practical. 1. The writing of the manuscript needs improvement. For instance, some figures are not clear enough, some tables are not properly formatted using the standard three-line table style, and some equations are not centered. 2. The manuscript claims that the proposed fusion decomposition strategy and the adaptive fusion loss are innovative contributions to the field of infrared and visible image fusion. However, these strategies have already been proposed in SDNet. 3. Deep learning-based methods used in the comparison are not the latest or most advanced ones in the literature, and the experimental results are not convincing enough to support the superiority of the proposed algorithm. In addition, the recent SOTA methods U2Fusion, SwinFusion and SuperFusion should be discussed. 4. The authors claim that the proposed algorithm can promote high-level visual tasks. However, the manuscript lacks experimental evidence or analysis to support this claim. 5. The experimental results shown in Figures 7, 8, 9, and 10 are not easy to distinguish subjectively. Authors should consider highlighting the differences between the experimental results by zooming in on small areas of interest. 6. The authors claim in the Introduction section that all CNN-based methods for image fusion require a reference image. However, this statement is inaccurate and may mislead the readers. In the context of infrared and visible image fusion tasks, there is no reference image available. Therefore, all CNN-based IR-VIS fusion algorithms in this field are essentially unsupervised learning methods. 7. In the Abstract section, the authors claim that existing fusion algorithms have limitations in extracting multi-source information but do not specify what these limitations are. Authors should consider revising the abstract to provide more specific and concise information about the proposed algorithm and its contributions. Reviewer #2: The work of this paper is logical. After reading this paper, I think this paper is complete and contains all the necessary components. The structure is reasonable and logical, and the design principle is clearly and moderately illustrated. To further improve this manuscript, I have the following suggestions: 1. It is recommended that the authors further refine the description of the contributions of this paper in the introduction to make it more concise and precise. By refining the description, the core contributions of this research can be more clearly conveyed, and lengthy or vague statements can be avoided. 2. The FDNet network architecture proposed in Figure 3 consists of fusion and decomposition networks. Please give a detailed explanation and justification to clarify why decomposition is needed after fusion and what is the intent of the decomposition network design. 3. Provide a brief explanation of the relationship and differences between CBAM and I_CBAM and what are the advantages of the improved I_CBAM attention mechanism. 4. Please have the authors draw a diagram to explain this paper's overall image fusion framework and the relationship between the designed gradient loss and intensity loss. 5. The designed gradient loss adaptive weight block is processed by Gaussian low-pass filtering for noise reduction, and then the optimization target of the corresponding pixel of the source image is evaluated based on the gradient richness. This process lacks visual representation, and it is recommended to insert relevant images in the manuscript to further illustrate this process for better reader comprehension. 6. I would suggest that the authors include 1-2 additional recent comparative experiments to validate the superior performance of the proposed network model. These supplementary comparative experiments can further evaluate and demonstrate the performance advantages of the proposed network model compared to other methods or models. 7. There are some errors in the manuscript's figures, equations, and other elements. Please carefully review and check them. In summary, my recommendation for this paper is to accept it with minor revisions. Moreover, I hope the author will adjust according to the above recommendations. ****** 6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy. Reviewer #1: No Reviewer #2: No ******** [NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.] While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step. https://doi.org/10.1371/journal.pone.0290231.r001
Revision 1
27 Jun 2023 Author Response To Reviewer #1: 1.The writing of the manuscript needs improvement. For instance, some figures are not clear enough, some tables are not properly formatted using the standard three-line table style, and some equations are not centered. Thank you for your valuable comments and suggestion. According to your comments, we have used professional typesetting LaTeX software to reorganize the pictures, tables, and equations in the paper. 2. The manuscript claims that the proposed fusion decomposition strategy and the adaptive fusion loss are innovative contributions to the field of infrared and visible image fusion. However, these strategies have already been proposed in SDNet. Thank you for your valuable comments. The overall framework of the SDNet network is divided into a fusion and a decomposition network. The fusion network performs feature extraction using dense blocks, and the decomposition network utilizes a dual-branch network to decompose the fused image, which contains rich scene information. Over the course of designing the loss functions, the intensity loss selects useful intensity information by adjusting the proportional coefficient, and the gradient loss introduces an adaptive gradient decision block. The decomposition loss utilizes the pixel-wise difference to guide the approximation between the decomposed image and the raw image. Our design idea inspired by the SDNet is different from the SDNet. Infrared images have a strong anti-interference ability, and visible images provide better texture detail and have high spatial resolution. Our fusion network is designed not only based on the characteristics of infrared and visible imagery, but also takes into account the different properties of raw images captured by different sensors. Our proposed fusion network is divided into three steps in the feature extraction process, which has an obvious difference than the SDNet acquired from dense blocks. The first step is to utilizes multi-scale convolution for shallow feature extraction. This addresses the issue of insufficient feature extraction with only a single-scale convolution kernel and enhances the network capability to extract shallow features. The second step is to use depth-separable convolution to consider the channel and spatial information in the image region separately. We employ deep convolution operations and point-by-point convolution operations to deepen the network depth without changing the size of the feature map itself. This allows for enhancing the network expression capability and constructing a lightweight network. The third step uses the improved I-CBAM attention mechanism to focus on the salient information of infrared and visible images from both the channel and spatial aspects. This process suppresses irrelevant channel information and ensures that all salient features can be effectively utilized during the fusion process. Our proposed decomposition network takes into account the presence of certain noise in the fused image, constructs single and double channels to extract and decompose different fusion features, and uses the constructed SSIM (structure similarity) loss function to calculate the structural similarity between the decomposition result and the raw image in terms of structural distortion, contrast distortion and luminance distortion. Hereinto, an improved Frobenius norm is chosen in the fusion loss to effectively adjust the weights between the fused image and the two raw images. This adjustment is achieved by measuring the distance between the pixel matrices of the fused image and the raw image. It prompts the network training to select more effective information. To demonstrate that our method is superior to the SDNet, we add the SDNet in our comparative experiments. The results can be observed in the subjective evaluation in Section 4.3.1 and the objective evaluation in Section 4.3.2. 3. Deep learning-based methods used in the comparison are not the latest or most advanced ones in the literature, and the experimental results are not convincing enough to support the superiority of the proposed algorithm. In addition, the recent SOTA methods U2Fusion, SwinFusion and SuperFusion should be discussed. Thank you for your valuable comments and suggestions. According to your suggestions, we have added two state-of-the-art infrared and visible image fusion algorithms, including U2Fusion and SwinFusion. However, the third SuperFusion method needs a large number of computing resources to train and tune the network. on the research, we face resource constraints and could not afford the computational cost, so we regrettably could not include the SuperFusion method in the manuscript. We sincerely apologize for this limitation and deeply understand that this is an important aspect of perfecting our research. In future research, we will continue to explore the SuperFusion method and will use it as a valuable direction to improve our work. In addition, we also added two latest comparison algorithms, including a progressive infrared and visible image fusion network based on illumination aware (PIAFusion) proposed by Tang et al. and a fast and lightweight infrared and visible image fusion network via feature flow and edge compensation for salient information (FLFusion) proposed by Xue et al. They can be seen in the subjective evaluation in Section 4.3.1 and the objective evaluation in Section 4.3.2. 4. The authors claim that the proposed algorithm can promote high-level visual tasks. However, the manuscript lacks experimental evidence or analysis to support this claim. Thank you for your valuable comments. There was an error describing our method to facilitating "high-level visual tasks" in the Abstract of the manuscript, and I would appreciate your help about pointing out this error to me. We careful checked the statement regarding "high-level visual tasks", and concluded that it should not be described as "high-level visual tasks" here, but should be described as "human eye visual characteristics". Because the subjective vision of the fused image and the labeling of the significant target regions of the fused image are primarily described in the related experiments of Section 4. This further emphasizes the high conformity of the fused image with human eye visual characteristics. 5. The experimental results shown in Figures 7, 8, 9, and 10 are not easy to distinguish subjectively. Authors should consider highlighting the differences between the experimental results by zooming in on small areas of interest. Thank you for your valuable comments and suggestions. Based on your suggestion, we have enlarged the regions of interest in Figs. 7, 8, 9, 10 to highlight the differences between the experiments. Due to the addition of a schematic diagram illustrating the adaptive weighting process in the revised manuscript and a graphical description in the Introduction, the corresponding figure numbers have changed. Therefore, Fig. 7 becomes Fig. 9, Fig. 8 becomes Fig. 10, and so on. The revised images have been inserted into Sections 4.2.1, 4.2.2, 4.2.3, and 4.2.4 of the revised version. 6. The authors claim in the Introduction section that all CNN-based methods for image fusion require a reference image. However, this statement is inaccurate and may mislead the readers. In the context of infrared and visible image fusion tasks, there is no reference image available. Therefore, all CNN-based IR-VIS fusion algorithms in this field are essentially unsupervised learning methods. Thank you for your valuable comments. There are indeed errors in the manuscript. Not all CNN-based image fusion algorithms are based on supervised learning. The well-known supervised learning method was only proposed in 2017 for multi-focus image fusion. In 2019, Li et al. proposed DenseFuse, which train an image reconstruction network using the COCO dataset and achieve image fusion goal. After consulting relevant references, different from the multi-focus image fusion task, the infrared and visible image fusion task cannot generate usable labeled data for supervised training. In order to solve the problem, the network proposed later uses the CNN trained offline as a feature extractor, but lacks the ability to select or fuse deep features adaptively. This section has been revised in the revised manuscript. The revised content is given as follows: Deep learning-based methods extract and combine image features based on strong feature learning capabilities of neural networks, and could be classified into supervised learning-based methods and unsupervised learning-based methods. Liu et al. [18] adopted the Convolutional Neural Network (CNN) for image fusion and made significant progresses comparing with traditional algorithms, but the CNN requires supervised training. For infrared and visible image fusion tasks, it is impossible to generate usable labeled data. In other words, it is impossible to artificially construct fusion images that can be referenced for supervised training. To address this problem, Li Hui et al [19] proposed to use the pre-trained VGG network for fusing infrared and visible images. This algorithm enables the extraction and fusion of multi-level deep features from the source images. Later, ResNet-50 [20] was proposed to extract and fuse depth features from the source images. However, a significant drawback of these models is their reliance on pre-trained CNN models as offline feature extractors. This limitation prevents adaptive extraction and fusion of features from the source images. Subsequently, scholars designed an end-to-end network framework specifically for image fusion. Prabhakar et al [21]. proposed an unsupervised end-to-end convolutional neural network learning framework, which does not require manually setting complex fusion strategies than other image fusion methods. The novel framework has the flexibility and versatility than previously experienced, but its performance evaluation results are not optimal for specific image fusion tasks. 7. In the Abstract section, the authors claim that existing fusion algorithms have limitations in extracting multi-source information but do not specify what these limitations are. Authors should consider revising the abstract to provide more specific and concise information about the proposed algorithm and its contributions. Thank you for your valuable comments and suggestions. Based on the reviewer suggestions, we have added specific limitations of the existing fusion algorithms in extracting multi-source information in Abstract. The modifications are written as follows: However, most existing fusion methods for extracting features from infrared and visible images are based on convolutional neural networks (CNNs). These methods often fail to make full use of the salient objects and texture features in the raw image, leading to problems such as insufficient texture details and low contrast in the fused image. To Reviewer #2: 1.It is recommended that the authors further refine the description of the contributions of this paper in the introduction to make it more concise and precise. By refining the description, the core contributions of this research can be more clearly conveyed, and lengthy or vague statements can be avoided. Thank you for your valuable comments and suggestions. In the manuscript, the introduction section consisted of a list of each innovation point, lacking conciseness, and failing to highlight the core contributions. Based on your advice, in the revised manuscript, we have condensed and simplified the core contributions. 2.The FDNet network architecture proposed in Figure 3 consists of fusion and decomposition networks. Please give a detailed explanation and justification to clarify why decomposition is needed after fusion and what is the intent of the decomposition network design. Thank you for your valuable comments. The proposed decomposition network in this paper aims to further improve the quality of fused images. The existing fused images have noise and redundancy to a certain extent, resulting in the lack of scene information in the fused images. The decomposition network designed in this paper constructs the structural similarity loss function SSIM by extracting and decomposing different fused feature sources, and compares the fused images with the raw images, so that the loss function is minimized and vice versa to promote the optimal training of the overall framework and finally output the final fused images. 3.Provide a brief explanation of the relationship and differences between CBAM and I-CBAM and what are the advantages of the improved I-CBAM attention mechanism. Thank you for your valuable comments. The original Convolutional Block Attention Module (CBAM) combines spatial attention and channel attention within the convolutional module, creating a sequential attention structure from channels to spatial dimensions. However, for image fusion tasks, this sequential structure acquired channel attention feature maps based on spatial attention learning mechanism. Which partly affects the features learned by the attention module and naturally reduces the performance of the spatial attention module. The improved attention mechanism in this design breaks away original sequential structure, allowing the combined spatial attention module and channel attention module to learn the original input features simultaneously. This eliminates the need for the sequential order of spatial attention and channel attention, enabling direct learning of the original input feature maps. Secondly, since the performance of spatial attention is determined by the size of the convolutional kernel, the spatial attention module chooses a large 7×7 convolutional kernel to aggregate spatial features. Compared to using a smaller 3×3 convolutional kernel, this increases the size of the receptive field and the number of parameters in the module. Therefore, compared to other same receptive fields, we design a spatial attention module using dilated convolution to complete feature aggregation to reduce the number of module parameters. 4.Please have the authors draw a diagram to explain this paper's overall image fusion framework and the relationship between the designed gradient loss and intensity loss. Thank you for your valuable comments. Based on your suggestion we have drawn a diagram to illustrate the correlation between the overall fusion framework, gradient loss and intensity loss. 5.The designed gradient loss adaptive weight block is processed by Gaussian low-pass filtering for noise reduction, and then the optimization target of the corresponding pixel of the source image is evaluated based on the gradient richness. This process lacks visual representation, and it is recommended to insert relevant images in the manuscript to further illustrate this process for better reader comprehension. Thank you for your valuable comments and suggestions. Based on your suggestion, we have added schematic diagram of adaptive weight block corresponding to the gradient loss from Fig. 3 in Section 3.4.1. 6. I would suggest that the authors include 1-2 additional recent comparative experiments to validate the superior performance of the proposed network model. These supplementary comparative experiments can further evaluate and demonstrate the performance advantages of the proposed network model compared to other methods or models. Thank you for your valuable comments and suggestions. Based on your suggestion we have added 2 latest deep learning comparison algorithms, including a progressive infrared and visible image fusion network based on illumination aware (PIAFusion) in Section 4.2.1 and a fast and lightweight infrared and visible image fusion network via feature flow and edge compensation for salient information (FLFusion) in Section 4.2.2. 7.There are some errors in the manuscript's figures, equations, and other elements. Please carefully review and check them. Thank you for your valuable comments and suggestions. Based on your suggestions, we have rechecked the images, formulas, and tables in the text carefully and reformatted them. Attachments Attachment Submitted filename: Response to Reviewers.doc https://doi.org/10.1371/journal.pone.0290231.r002
7 Aug 2023 Decision Letter - Gulistan Raja, Editor FDNet:An end-to-end fusion decomposition network for infrared and visible images PONE-D-23-09775R1 Dear Dr. Ren, We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements. Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication. An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org. If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org. Kind regards, Gulistan Raja Academic Editor PLOS ONE Additional Editor Comments (optional): Reviewers' comments: Reviewer's Responses to Questions Comments to the Author 1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation. Reviewer #1: All comments have been addressed Reviewer #2: All comments have been addressed ******** 2. Is the manuscript technically sound, and do the data support the conclusions? The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented. Reviewer #1: Yes Reviewer #2: Yes ****** 3. Has the statistical analysis been performed appropriately and rigorously? Reviewer #1: Yes Reviewer #2: Yes ****** 4. Have the authors made all data underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified. Reviewer #1: Yes Reviewer #2: Yes ****** 5. Is the manuscript presented in an intelligible fashion and written in standard English? PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here. Reviewer #1: Yes Reviewer #2: Yes ****** 6. Review Comments to the Author Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters) Reviewer #1: The authors have addressed most of my concerns. In this case, I recommend accepting this paper as is. Reviewer #2: all comments have been addressed.Once the author has revised the article to meet the publishing requirements, it is advisable to accept and publish it. ****** 7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy. Reviewer #1: No Reviewer #2: No ******** https://doi.org/10.1371/journal.pone.0290231.r003
Formally Accepted
7 Sep 2023 Acceptance Letter - Gulistan Raja, Editor PONE-D-23-09775R1 FDNet: An end-to-end fusion decomposition network for infrared and visible images Dear Dr. Ren: I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department. If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org. If we can help with anything else, please email us at plosone@plos.org. Thank you for submitting your work to PLOS ONE and supporting open access. Kind regards, PLOS ONE Editorial Office Staff on behalf of Dr. Gulistan Raja Academic Editor PLOS ONE https://doi.org/10.1371/journal.pone.0290231.r004

Open letter on the publication of peer review reports

PLOS recognizes the benefits of transparency in the peer review process. Therefore, we enable the publication of all of the content of peer review and author responses alongside final, published articles. Reviewers remain anonymous, unless they choose to reveal their names.

We encourage other journals to join us in this initiative. We hope that our action inspires the community, including researchers, research funders, and research institutions, to recognize the benefits of published peer review reports for all parts of the research system.

Learn more at ASAPbio .