Article
Computer Science, Artificial Intelligence
Xingyu Hu, Junjun Jiang, Xianming Liu, Jiayi Ma
Summary: Multi-focus image fusion (MFF) is a challenging task due to the difficulty in distinguishing different blur levels and the lack of real supervised data. In this study, we propose a novel deep learning-based framework named ZMFF, which captures the deep prior of the fused image and the focus map using deep image prior and deep mask prior networks, respectively. Our method achieves promising performance, generalization, and flexibility on both synthetic and real-world datasets without the need for extensive training data.
INFORMATION FUSION
(2023)
Article
Computer Science, Information Systems
Nana Yu, Jinjiang Li, Zhen Hua
Summary: This paper proposes a dual-path fusion network with attention mechanism for multi-focus image fusion, aiming to address the problems of traditional image fusion methods. The experimental results demonstrate the strong robustness and effectiveness of the proposed method.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Qizhi Xu, Yuan Li, Jinyan Nie, Qingjie Liu, Mengyao Guo
Summary: This study proposes an iterative network based on spectral and textural loss constrained GAN for pansharpening. By generating mean difference images and using a coarse-to-fine fusion framework, as well as embedding loss functions for fidelity preservation, better fusion performance is achieved.
INFORMATION FUSION
(2023)
Article
Computer Science, Artificial Intelligence
Kaijun Wu, Yuan Mei
Summary: This paper presents a multi-focus image fusion model based on unsupervised learning, which balances the information of an image with different focus positions through a two-stage processing. The proposed method achieves good results in preserving texture details and edge information, and outperforms other algorithms in fusion performance.
MACHINE VISION AND APPLICATIONS
(2022)
Article
Physics, Multidisciplinary
Bingzhe Wei, Xiangchu Feng, Kun Wang, Bian Gao
Summary: A novel fusion method that combines CNN and SR for multi-focus image fusion has been proposed, resulting in a more accurate and informative fused image. Experimental results demonstrate that this method clearly outperforms existing methods in terms of visual perception and objective evaluation metrics, while also significantly reducing computational complexity.
Review
Computer Science, Artificial Intelligence
Gaurav Choudhary, Dinesh Sethi
Summary: Image fusion is a well-established field of study in digital image processing. Multi-focus image fusion (MFIF) is a widely used application, but it faces challenges such as low contrast, color distortion, and fusion losses. This study proposes an experimental and comprehensive approach to address these challenges by applying pre-hand enhancement criteria for both grayscale and color images. The results demonstrate that using the enhancement method as a preprocessing step can improve the outcomes of MFIF algorithms in both objective and subjective evaluations.
ARTIFICIAL INTELLIGENCE REVIEW
(2023)
Article
Computer Science, Artificial Intelligence
Limai Jiang, Hui Fan, Jinjiang Li
Summary: This study proposes a new end-to-end network for multi-focus image fusion, which uses a residual atrous spatial pyramid pooling module and a disparities attention module to extract multi-level features and reduce information loss. By introducing a new dataset for supervised learning, the network's performance is improved.
APPLIED INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Odysseas Bouzos, Ioannis Andreadis, Nikolaos Mitianoudis
Summary: The limited depth of field of optical lenses makes multi-focus image fusion (MFIF) algorithms crucial. Recently, Convolutional Neural Networks (CNN) have been widely used in MFIF methods, but their predictions often lack structure and are limited by the size of the receptive field. To address this, a novel robust to noise Convolutional Neural Network-based Conditional Random Field (mf-CNNCRF) model is introduced. The model utilizes the mapping between input and output of CNN networks and the long-range interactions of CRF models to achieve structured inference.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Engineering, Biomedical
Kun Tang, Lihui Wang, Xingyu Huang, Xinyu Cheng, Yue-Min Zhu
Summary: Deformable medical image registration is crucial for clinical applications. We propose a multi-dilation spherical graph transformer (MD-SGT) that combines convolutional and graph transformer blocks to effectively distinguish differences between reference and template images at various scales, improving registration performance.
COMPUTERIZED MEDICAL IMAGING AND GRAPHICS
(2023)
Article
Chemistry, Analytical
Shaodi Yang, Yuqian Zhao, Miao Liao, Fan Zhang
Summary: An improved unsupervised learning-based framework is proposed for multi-organ registration on 3D abdominal CT images, incorporating RCN modules and a topology-preserving loss in the total loss function for more accurate registration results. Experimental results demonstrate the superiority of the proposed method over existing methods and its potential to meet clinical registration requirements.
Article
Computer Science, Information Systems
Derya Avci, Eser Sert, Fatih Ozyurt, Engin Avci
Summary: A new fusion method called Multi-Focus Image Fusion Based on Discrete Wavelet Transform with Deep Convolutional Neural Network (MFIF-DWT-CNN) is proposed to reduce spatial artifacts and blurring effects in edge details and increase the robustness of multifocal image fusion. The MFIF-DWT-CNN approach collects the required features from the main image to create a new merged image, resulting in a clearer image. Experimental results show that the proposed method outperforms other methods in relevant metrics, demonstrating its effectiveness.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Hao Zhang, Zhuliang Le, Zhenfeng Shao, Han Xu, Jiayi Ma
Summary: This paper introduces a new method for multi-focus image fusion, utilizing a generative adversarial network with adaptive and gradient joint constraints to address the issue of detail loss in existing methods. The proposed method demonstrates superiority in both subjective visual effect and quantitative metrics over the state-of-the-art, while also being approximately one order of magnitude faster.
INFORMATION FUSION
(2021)
Article
Computer Science, Artificial Intelligence
Yu Liu, Lei Wang, Huafeng Li, Xun Chen
Summary: An innovative DL-based multi-focus image fusion method is proposed in this paper, which combines the advantages of transform domain methods and spatial domain methods. The method can effectively preserve the original focus information and prevent visual artifacts around boundary regions.
INFORMATION FUSION
(2022)
Article
Computer Science, Hardware & Architecture
Wenyi Zhao, Huihua Yang, Jie Wang, Xipeng Pan, Zhiwei Cao
Summary: The study proposes a region- and pixel-based method for multi-focus image fusion, which outperforms other methods in visual perception and object metrics, with a 80% improvement compared to conventional CNN-based methods.
MOBILE NETWORKS & APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Chengchao Wang, Yuanyuan Pu, Xue Wang, Chaozhen Ma, Rencan Nie
Summary: This paper proposes a novel conditional focus probability learning model, MCNN, for multi-focus image fusion (MFIF). The MCNN can generate conditional focus probabilities for a pair of source images, which are then converted into binary focus masks to directly produce an all-focus image without postprocessing. The MCNN includes a fully convolutional encoder with two mutually coupled Siamese branches and a hybrid loss with a structural sparse fidelity loss and a structural similarity loss, which enable accurate conditional focus probability learning.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Information Systems
Pourya Shamsolmoali, Masoumeh Zareapoor, Huiyu Zhou, Jie Yang
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS
(2020)
Article
Computer Science, Artificial Intelligence
Deepak Kumar Jain, Masoumeh Zareapoor, Rachna Jain, Abhishek Kathuria, Shivam Bachhety
NEURAL COMPUTING & APPLICATIONS
(2020)
Article
Computer Science, Artificial Intelligence
Pourya Shamsolmoali, Masoumeh Zareapoor, Linlin Shen, Abdul Hamid Sadka, Jie Yang
Summary: This paper proposes a method to address imbalanced image datasets by introducing a competitive game between generator and discriminator networks, which improves learning from imbalanced data. The generator is trained with a feature matching loss function to prevent the generation of outliers and maintain the majority class space.
Article
Computer Science, Artificial Intelligence
Donghao Shen, Masoumeh Zareapoor, Jie Yang
Summary: The paper introduces a novel multimodal image fusion algorithm focusing on transferring salient structures and maintaining spatial consistency. The algorithm selects features to transfer using a graph cut algorithm, with spatial varying smoothness cost formulated based on the independence between local features.
IMAGE AND VISION COMPUTING
(2021)
Article
Engineering, Mechanical
Masoumeh Zareapoor, Pourya Shamsolmoali, Jie Yang
Summary: This paper introduces a new adversarial network model for simultaneous classification and fault detection. By generating faulty samples from a mixture of data distribution to restore balance in imbalanced datasets, the proposed model performs well in experiments, particularly in recognizing faulty samples.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Lu Wang, Jie Yang, Masoumeh Zareapoor, Zhonglong Zheng
Summary: This paper introduces a novel framework for projecting original data points from different modalities into low-dimensional latent space and finding cluster centroid points using Cluster-wise Unsupervised Hashing (CUH). The framework aims to jointly learn compact hash codes and corresponding linear hash functions, showing superior effectiveness in unsupervised cross-modal hashing tasks compared to state-of-the-art methods.
PATTERN RECOGNITION
(2021)
Article
Computer Science, Information Systems
Masoumeh Zareapoor, Jie Yang
Summary: Image-to-Image translation faces challenges such as lack of paired datasets, multimodality, and diversity. A new variation of generative models using a trainable transformer aims to address these issues by explicitly allowing spatial manipulation of data within training.
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS
(2021)
Article
Geochemistry & Geophysics
Pourya Shamsolmoali, Masoumeh Zareapoor, Huiyu Zhou, Ruili Wang, Jie Yang
Summary: A new model is introduced to apply structured domain adaption for synthetic image generation and road segmentation, incorporating a feature pyramid network into generative adversarial networks to minimize the difference between the source and target domains and improve road extraction accuracy and completeness.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Computer Science, Artificial Intelligence
Pourya Shamsolmoali, Masoumeh Zareapoor, Eric Granger, Huiyu Zhou, Ruili Wang, M. Emre Celebi, Jie Yang
Summary: Generative Adversarial Networks (GANs) have shown great success in various fields, particularly in image synthesis. This survey provides a comprehensive review of adversarial models for image synthesis, summarizing methods and discussing future research directions. Additionally, all software implementations and datasets of these GAN methods have been collected and made available, which is a unique feature of this review.
INFORMATION FUSION
(2021)
Article
Geochemistry & Geophysics
Pourya Shamsolmoali, Masoumeh Zareapoor, Jocelyn Chanussot, Huiyu Zhou, Jie Yang
Summary: The study introduces a novel image pyramid network based on rotation equivariance convolution to tackle the challenge of extracting features for small-scale objects in current object detectors. The proposed model combines a single-shot detector with a lightweight image pyramid module, allowing for feature extraction across various scales and orientations in an optimized manner.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Pourya Shamsolmoali, Jocelyn Chanussot, Masoumeh Zareapoor, Huiyu Zhou, Jie Yang
Summary: In this study, a new architecture called MPFP-Net is proposed to address the challenges of object detection in remote sensing images. By dividing patches into class-affiliated subsets and designing a sequence of smooth loss functions, the model is improved to better collect small object parts. The network utilizes bottom-up and crosswise connections to fuse features of different scales for enhanced accuracy, while also being more efficient than baseline models.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Automation & Control Systems
Pourya Shamsolmoali, Masoumeh Zareapoor, Swagatam Das, Salvador Garcia, Eric Granger, Jie Yang
Summary: Image-to-image translation is crucial in generative adversarial networks. Convolutional neural networks have limitations in capturing spatial relationships, making them unsuitable for image translation tasks. Capsule networks are proposed as a remedy, capturing hierarchical spatial relationships. In this paper, a new framework for capsule networks is presented, which can be applied to generator-discriminator architectures without computational overhead. A Gromov-Wasserstein distance is used as a loss function to guide the learned distribution. The proposed method, called generative equivariant network, is evaluated on I2I translation and image generation tasks and shows a principled connection between generative and capsule models.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Computer Science, Artificial Intelligence
Pourya Shamsolmoali, Masoumeh Zareapoor, Huiyu Zhou, Dacheng Tao, Xuelong Li
Summary: Deep generative models can learn non-linear data distributions using latent variables and a non-linear generator function. However, the weak projection of the latent space into the data space can result in poor representation learning. This paper proposes a Variational spatial-Transformer AutoEncoder (VTAE) that minimizes geodesics on a Riemannian manifold to improve representation learning.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Pourya Shamsolmoali, Masoumeh Zareapoor, Jie Yang, Eric Granger, Huiyu Zhou
Summary: This study proposes a dilated scale-wise feature fusion network based on convolution factorization for skin lesion detection in dermoscopic images. The proposed model can extract features at different scales and fuse them for better lesion detection.
2022 26TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR)
(2022)
Article
Computer Science, Information Systems
Lu Wang, Masoumeh Zareapoor, Jie Yang, Zhonglong Zheng
Summary: Cross-modal hashing (CMH) is a method that can learn and retrieve similarity across different modalities. However, existing methods have limitations in fully exploiting the underlying properties of multi-modal data and often suffer from significant quantization errors. This paper proposes a novel Asymmetric Correlation Quantization Hashing (ACQH) method to address these challenges, which learns projection matrices and constructs hash codes using semantic similarity preservation and label regression.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)