Yuping Huang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yuping Huang has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Weisheng Li
Chongqing University of Posts and Telecommunications
4 shared publications - Lijian Yang
Chongqing University of Posts and Telecommunications, Chongqing, China
2 shared publications - Yucheng Shu
Chongqing University of Posts and Telecommunications, Chongqing, China
2 shared publications - Bin Xiao
School of Artificial Intelligence, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
1 shared publication - Dan He
School of Computer Science and Technology, Chongqing University of Posts and Telecommunications, Chongqing, China
1 shared publication Feifei Chao
Chongqing Key Laboratory of Image Cognition, Chongqing University of Posts and Telecommunications, Chongqing, 400065, China
1 shared publication
Latest Publications
- 16 December 2025
- Published by Elsevier in Pattern Recognition
- Vol. 173
Rethinking normalization strategies and convolutional kernels for multimodal image fusion
- 16 December 2025
- Published by Elsevier in Pattern Recognition
- Vol. 173
Rethinking normalization strategies and convolutional kernels for multimodal image fusion
- Open Access
- 28 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 2814-2823
ShiftMorph: A Fast and Robust Convolutional Neural Network for 3D Deformable Medical Image Registration
Deformable image registration (DIR) is crucial for many medical image applications. In recent years, learning-based methods utilizing the convolutional neural network (CNN) or the Transformer have demonstrated their superiority in image registration, dominating a new era for DIR. However, very few of these methods can satisfy the demands of real-time applications due to the high spatial resolution of 3D volumes and the high complexity of 3D operators. To tackle this, we propose losslessly downsampling by shifting the strided convolution. A grouping strategy is then used to reduce redundant computations and support self-consistency learning. As an inherent regularizer of the network design, self-consistency learning improves the deformation quality and enables halving the proposed network after training. Furthermore, the proposed shifted connection converts the decoding operations into a lower-dimensional space, significantly reducing decoding overhead. Extensive experimental results on medical image registration demonstrate that our method is competitive with state-of-the-art methods in terms of registration performance, and additionally, it achieves over 3× the speed of most of them.
- 28 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 2814-2823
Open AccessShiftMorph: A Fast and Robust Convolutional Neural Network for 3D Deformable Medical Image Registration
Deformable image registration (DIR) is crucial for many medical image applications. In recent years, learning-based methods utilizing the convolutional neural network (CNN) or the Transformer have demonstrated their superiority in image registration, dominating a new era for DIR. However, very few of these methods can satisfy the demands of real-time applications due to the high spatial resolution of 3D volumes and the high complexity of 3D operators. To tackle this, we propose losslessly downsampling by shifting the strided convolution. A grouping strategy is then used to reduce redundant computations and support self-consistency learning. As an inherent regularizer of the network design, self-consistency learning improves the deformation quality and enables halving the proposed network after training. Furthermore, the proposed shifted connection converts the decoding operations into a lower-dimensional space, significantly reducing decoding overhead. Extensive experimental results on medical image registration demonstrate that our method is competitive with state-of-the-art methods in terms of registration performance, and additionally, it achieves over 3× the speed of most of them.
- 1 February 2022
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 141, 105005
A multiscale double-branch residual attention network for anatomical–functional medical image fusion
- 1 February 2022
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 141, 105005
A multiscale double-branch residual attention network for anatomical–functional medical image fusion
- 20 November 2020
- Published by Wiley in International Journal of Imaging Systems and Technology
- Vol. 31 (3) , 1246-1264
Anatomical‐functional image fusion based on deep convolution neural networks in local Laplacian pyramid domain
Medical image fusion technology makes clinical diagnosis and treatment more accurate. This technique can solve the existing problem that single mode medical images conveying insufficient information. Therefore, the key to this technology is to retain as much as possible information in the original medical image of multiple modes. However, the existing methods often lose source image detail with a low contrast and cause color distortion. This article proposed a novel multi‐modal medical image fusion framework. Our framework has four key steps. First, extracting the super‐resolution images from anatomical images via a deep neural network is performed. Second, decomposing the source images into detail‐enhanced approximate and residual images, and then processing by local Laplacian filtering. Third, a convolution neural network is used to complete the mapping of the source image to the feature map and the local energy‐based strategy is used to integrate the coefficients. Finally, the inversed local Laplacian pyramid is adopted to reconstruct the fused image. The experimental results prove that the fused image via the proposed method has the characteristics of high contrast, high brightness, and high color saturation and great advantages in retaining the color and structural information.
- 20 November 2020
- Published by Wiley in International Journal of Imaging Systems and Technology
- Vol. 31 (3) , 1246-1264
Anatomical‐functional image fusion based on deep convolution neural networks in local Laplacian pyramid domain
Medical image fusion technology makes clinical diagnosis and treatment more accurate. This technique can solve the existing problem that single mode medical images conveying insufficient information. Therefore, the key to this technology is to retain as much as possible information in the original medical image of multiple modes. However, the existing methods often lose source image detail with a low contrast and cause color distortion. This article proposed a novel multi‐modal medical image fusion framework. Our framework has four key steps. First, extracting the super‐resolution images from anatomical images via a deep neural network is performed. Second, decomposing the source images into detail‐enhanced approximate and residual images, and then processing by local Laplacian filtering. Third, a convolution neural network is used to complete the mapping of the source image to the feature map and the local energy‐based strategy is used to integrate the coefficients. Finally, the inversed local Laplacian pyramid is adopted to reconstruct the fused image. The experimental results prove that the fused image via the proposed method has the characteristics of high contrast, high brightness, and high color saturation and great advantages in retaining the color and structural information.
- 1 October 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 7569-7573
Simulations of friction models for linear motors
In this paper, three different models of friction in presliding regime on linear motor platform are deduced according to basic physical laws, which are named as the Kelvin model, the modified Kelvin model and the four-element model. To validate the effectiveness of these models, the particle swarm optimization is applied to determine the coefficients in each model. Then experiments of ramp input force and step-like input force are conducted to collect the tiny displacement of the moving stage; simulations are carried out to calculate the models under input forces. Through analysis on the error and comparison between experiments and simulations, we can draw a conclusion that the modified Kelvin model is the most precise friction model on linear motor platforms.
- 1 October 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 7569-7573
Simulations of friction models for linear motors
In this paper, three different models of friction in presliding regime on linear motor platform are deduced according to basic physical laws, which are named as the Kelvin model, the modified Kelvin model and the four-element model. To validate the effectiveness of these models, the particle swarm optimization is applied to determine the coefficients in each model. Then experiments of ramp input force and step-like input force are conducted to collect the tiny displacement of the moving stage; simulations are carried out to calculate the models under input forces. Through analysis on the error and comparison between experiments and simulations, we can draw a conclusion that the modified Kelvin model is the most precise friction model on linear motor platforms.