检索结果-内蒙古大学图书馆

Image depth estimation assisted by multi-view projection

COMPLEX & INTELLIGENT SYSTEMS 2025年第1期11卷 1-16页

作者： Liu, Liman Tian, Jinshan Luo, Guansheng Xu, Siyuan Zhang, Chen Hu, Huaifei Tao, Wenbing South Cent Minzu Univ Sch Biomed Engn Key Lab Cognit Sci Hubei Prov Key Lab Med Informat Anal & Tumor Diag Minzu Rd Wuhan 430074 Hubei Peoples R China Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Natl Key Lab Sci & Technol Multispectral Informat Luoyu Rd Wuhan 430074 Hubei Peoples R China

In recent years, deep learning has significantly advanced the development of image depth estimation algorithms. The depth estimation network with single-view input can only extract features from a single 2D image, often neglecting the information contained in neighboring views, resulting in learned features that lack real geometrical information in the 3D world and stricter constraints on the 3D structure, leading to limitations in the performance of image depth estimation. In the absence of accurate camera information, the multi-view geometric cues obtained by some methods may not accurately reflect the real 3D structure, resulting in a lack of multi-view geometric constraints in image depth estimation algorithms. To address this problem, a multi-view projection-assisted image depth estimation network is proposed, which integrates multi-view stereo vision into a deep learning-based encoding-decoding image depth estimation framework without pre-estimation of view bitmap. The network estimates optical flow for pixel-level matching across views, thereby projecting the features of neighboring views to the reference viewpoints for self-attentive feature aggregation, compensating for the lack of stereo geometry information in the image depth estimation framework. Additionally, a multi-view reprojection error is designed for supervised optical flow estimation to effectively constrain the optical flow estimation process. In addition, a long-distance attention decoding module is proposed to achieve effective extraction and aggregation of features in distant areas of the scene, which enhances the perception capability for outdoor long-distance. Experimental results on the KITTI dataset, vKITTI dataset, and SeasonDepth dataset demonstrate that our method achieves significant improvements compared to other state-of-the-art depth estimation techniques. This confirms its superior performance in image depth estimation.

关键词： multi-view projection Depth estimation Neural network Optical flow

来源：评论

学校读者我要写书评

暂无评论

multi-view projection Based Joint Geometry and Color Hole Repairing Method for G-PCC Trisoup Encoded Color Point Cloud

引用

IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE 2024年第1期8卷 892-902页

作者： Tao, Wenxu Jiang, Gangyi Yu, Mei Zhang, Yun Jiang, Zhidi Ho, Yo-Sung Ningbo Univ Fac Informat Sci & Engn Ningbo 315211 Peoples R China Zhejiang Prov United Key Lab Embedded Syst Hangzhou 310032 Peoples R China Sun Yat Sen Univ Sch Elect & Commun Engn Shenzhen 518107 Peoples R China Ningbo Univ Coll Sci & Technol Ningbo 315300 Peoples R China Gwangju Inst Sci & Technol Sch Elect Engn & Comp Sci Gwangju 61005 South Korea

Efficient compression of color point clouds (CPC) is crucial for immersive visual systems. The G-PCC standard by Moving Picture Experts Group offers effective compression solutions for static CPCs. G-PCC Trisoup, one of the G-PCC solutions, achieves high coding performance using a pruned octree, combined with triangle soup (trisoup) surface encoding. However, encoding errors introduce distortions like triangular holes, degrading visual quality. This article proposes a multi-view projection based joint geometry and color hole repairing method for G-PCC trisoup encoded color point cloud, significantly improving reconstruction quality. Firstly, a G-PCC Trisoup encoded CPC is projected onto multi-view color and geometry maps. A multi-view projection-based scheme detects triangular holes by identifying geometry discontinuities. A geometric weighting factor, considering direction, distance, and level set distance, repairs geometry maps. Additionally, a novel gated hole repairing network repairs color maps by leveraging semantic information and reducing background interference. Finally, the joint geometry and color repaired results are inversely projected to obtain the final CPCs. The proposed method is validated on distorted CPCs encoded with G-PCC Trisoup at different levels. Experimental results demonstrate effective improvement in geometry and color objective quality, leading to enhanced visual perception.

关键词： Color point cloud geometry based point cloud compression hole repairing multi-view projection gated hole repairing network

来源：评论

学校读者我要写书评

暂无评论

Representative Slice Selection and multi-view projection Learning for Pulmonary Tuberculosis Infectiousness Identification Using CT Volume

引用

IEEE ACCESS 2024年 12卷 1374-1385页

作者： Bai, Qiushun Gao, Yi Chen, Feng Zhang, Yiwen Yang, Yuan Zhong, Liming Yang, Wei Yang, Yi Southern Med Univ Sch Biomed Engn Guangzhou 510515 Peoples R China Guangdong Prov Key Lab Med Image Proc Guangzhou 510515 Peoples R China Hainan Med Univ Hainan Affiliated Hosp Hainan Gen Hosp Dept Infect Dis Haikou 570311 Peoples R China Southern Med Univ Nanfang Hosp Dept Infect Dis Guangzhou 510515 Peoples R China Southern Med Univ Nanfang Hosp Hepatol Unit Guangzhou 510515 Peoples R China Hainan Med Univ Hainan Gen Hosp Hainan Affiliated Hosp Dept Radiol Haikou 570311 Peoples R China

Pulmonary tuberculosis (PTB) is a major global health threat. Diagnosing PTB infectiousness is vital for clinical decision-making, but existing etiological examination methods do not meet the requirements for speed, accuracy, and cost effectiveness. Developing deep learning models for infectiousness identification based on computed tomography (CT) volume measurements holds promise for meeting these requirements. However, with limited samples and coarse annotations, the large amount of information in the CT volume poses a challenge for models to distinguish information related to patient-level labels, which often leads to severe model overfitting. In this study, A dual-branch framework is developed for identifying PTB infectiousness using CT volume. To address the issue of imbalance between the CT volume information and the patient-level labels, we propose a method for selecting representative slices to reduce redundant information and adopt a multiple-instance learning framework to improve label supervision. Furthermore, we incorporate multi-view projection information to compensate for the deficiency of global information caused by using single-dimensional slices as the input. Experimental results demonstrate that our strategy effectively mitigates overfitting and achieves desirable performance on an external test set, with an area under the receiver operating characteristic curve of 80.48%. This performance is superior to that obtained for models using the 3D CT volume or 2D projection images alone as the input.

关键词： Pulmonary tuberculosis infectiousness computed tomography slice selection multi-view projection multiple instance learning

来源：评论

学校读者我要写书评

暂无评论

MVMS-RCN: A Dual-Domain Unified CT Reconstruction With multi-Sparse-view and multi-Scale Refinement-Correction

IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING

引用

IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING 2024年 10卷 1749-1762页

作者： Fan, Xiaohong Chen, Ke Yi, Huaming Yang, Yin Zhang, Jianping Xiangtan Univ Sch Math & Comp Sci Xiangtan 411105 Peoples R China Hunan Key Lab Comp & Simulat Sci & Engn Xiangtan 411105 Peoples R China Zhejiang Normal Univ Coll Math Med Jinhua 321004 Peoples R China Univ Liverpool Ctr Math Imaging Tech Dept Math Sci Liverpool L6972L England Univ Strathclyde Dept Math & Stat Glasgow G1 1XH Scotland Hunan Natl Appl Math Ctr Xiangtan 411105 Peoples R China Minist Educ Key Lab Intelligent Comp & Informat Proc Xiangtan 411105 Peoples R China

X-ray Computed Tomography (CT) is one of the most important diagnostic imaging techniques in clinical applications. Sparse-view CT imaging reduces the number of projection views to a lower radiation dose and alleviates the potential risk of radiation exposure. Most existing deep learning (DL) and deep unfolding sparse-view CT reconstruction methods: 1) do not fully use the projection data;2) do not always link their architecture designs to a mathematical theory;3) do not flexibly deal with multi-sparse-view reconstruction assignments. This paper aims to use mathematical ideas and design optimal DL imaging algorithms for sparse-view CT reconstructions. We propose a novel dual-domain unified framework that offers a great deal of flexibility for multi-sparse-view CT reconstruction through a single model. This framework combines the theoretical advantages of model-based methods with the superior reconstruction performance of DL-based methods, resulting in the expected generalizability of DL. We propose a refinement module that utilizes unfolding projection domain to refine full-sparse-view projection errors, as well as an image domain correction module that distills multi-scale geometric error corrections to reconstruct sparse-view CT. This provides us with a new way to explore the potential of projection information and a new perspective on designing network architectures. The multi-scale geometric correction module is end-to-end learnable, and our method could function as a plug-and-play reconstruction technique, adaptable to various applications. Extensive experiments demonstrate that our framework is superior to other existing state-of-the-art methods.

关键词： Computed tomography Image reconstruction Imaging Noise reduction Convolutional neural networks X-ray imaging Mathematical models Error correction Computer architecture Transforms Deep learning (DL) unfolding explainable network multi-scale geometric correction multi-view projection sparse-view CT reconstruction plug-and-play

来源：评论

学校读者我要写书评

暂无评论

High-resolution spatial image display with multiple UHD projectors

High-resolution spatial image display with multiple UHD proj...

引用

Conference onThree-Dimensional Imaging, Visualization, and Display

作者： Watanabe, Hayato Kawakita, Masahiro Okaichi, Naoto Sasaki, Hisayuki Kano, Masanori Mishina, Tomoyuki NHK Japan Broadcasting Corp Sci & Technol Res Labs Setagaya Ku 1-10-11 Kinuta Tokyo 1578510 Japan

ISBN: (数字)9781510618442

ISBN: (纸本)9781510618442

Light field displays can provide a naturally viewable three-dimensional (3D) image without the need for using special glasses. However, improving in the resolution of 3D images is difficult because considerable image information is required. Therefore, we propose two new light field display methods that use multiple ultra-high definition projectors to realize a reproduction of a high-resolution spatial image. One of the two proposed methods is based on integral imaging. multi-elemental images are superimposed onto a lens array using multiple projectors placed at optimal positions. An integral 3D image with enhanced resolution and viewing angle can be reproduced by projecting each elemental image as collimated light rays at different predetermined angles. We prototyped a display system having six projector units and realized a resolution of approximately 100,000 pixels and viewing angle of approximately 30 degrees. The other proposed method aiming at further resolution enhancement is based on multi-view projection. By constructing a new display optical system to reproduce a full parallax light field and by developing a special 3D screen with isotropic narrow diffusion characteristics of non-Gaussian shape, optical 3D images could be reconstructed, which was difficult with conventional methods. We prototyped a display system comprising two projector units and realized higher resolution of approximately 330,000 pixels as compared to our previous full parallax light field display systems.

关键词： three-dimensional display high resolution full parallax light field integral imaging spatial imaging elemental images multi-view projection

来源：评论

学校读者我要写书评

暂无评论

Improvement of camera arrangement in computer-generated holograms synthesized from multi-view images

Improvement of camera arrangement in computer-generated holo...

引用

Conference on Practical Holography XXV - Materials and Applications

作者： Hayashi, Noriyuki Sakamoto, Yuji Honda, Yasuhiro Hokkaido Univ Grad Sch Informat Sci & Technol Kita Ku Sapporo Hokkaido 0600814 Japan

ISBN: (纸本)9780819484949

Computer-generated hologram (CGH) is a technique that enables recording optical systems for holography by computer simulation, and can reconstruct ideal 3D images. Since CGH requires the model data of objects, it is necessary for reacting 3D images of real existing objects to model them precisely. However, it is often difficult to obtain this data. To solve this problem, there has been much research into multi-view projection image based methods that do not require 3D model data. CGH using integral photography and CGH using multi-view projection obtained by lens array are examples of this research. However, these methods have some problems. First, their optical recording systems are complex and expensive, and therefore, realizing these systems is difficult. Another problem is that eormous projections are required to generate a hologram with sufficient pixel size. In the past, a method to generate holograms from multi-view images taken with digital cameras was proposed. Additionally, to generate CGHs using multi-view images fro randomaly arranged cameras, we proposed a method that uses distance transformations, rotations of light waves, and an algorithm that sets up hologram parameters automatically by using camera arrangement information. In this paper, the results of numerical reconstructions are compared, and both the motion parallax and resolution of the reconstructed images are discussed. The proper number of multi-view projections needed will be derived from this discussion.

关键词： Computer generated hologram multi-view projection arbitrary viewpoint real object

来源：评论

学校读者我要写书评

暂无评论

Aktina Vision: Full-parallax light field display system with resolution of 330,000 pixels using top-hat diffusing screen

Aktina Vision: Full-parallax light field display system with...

引用

Conference on Three-Dimensional Imaging, Visualization, and Display

作者： Watanabe, Hayato Okaichi, Naoto Omura, Takuya Kano, Masanori Sasaki, Hisayuki Kawakita, Masahiro NHK Japan Broadcasting Corp Sci & Technol Res Labs Setagaya Ku 1-10-11 Kinuta Tokyo 1578510 Japan

ISBN: (纸本)9781510626607

Light field displays technologies are popular glasses-free three-dimensional (3D) display methods, whereby natural 3D images can be viewed by precisely reproducing light rays from the objects. However, sufficient display performances cannot be obtained with conventional display techniques because reproduction of a great number of high-density light rays is required for high quality 3D images. Therefore, we develop a novel light field display method named Aktina Vision, which consists of a special 3D screen with isotropic narrow diffusion characteristics and a display optical system for projecting high-density light rays. In this method, multi-view images with horizontal and vertical parallaxes are projected onto the 3D screen at various angles in a superposed manner. The 3D screen has narrow diffusion angle and top-hat diffusion characteristics for optimal widening of the light rays according to the discrete intervals between the rays. 3D images with high resolution and depth-reproducibility can be displayed by suppressing crosstalk between light rays and reproducing them with continuous luminance distribution. We prototype a display system using 14 exclusively designed 4K projectors and develop a light field calibration technique. The reproduction of 3D images with a resolution of approximately 330,000 pixels, which is three times higher than that of conventional display methods using a lens array, and viewing angles of 35.1 degrees in the horizontal direction and 4.7 degrees in the vertical direction is realized by projecting 350 multi-view images in a superposed manner.

关键词： Three-dimensional display high resolution full parallax light field spatial imaging multi-view projection

来源：评论

学校读者我要写书评

暂无评论

Computer-generated hologram of actual objects from arbitrary viewpoints with range sensors and digital cameras

Computer-generated hologram of actual objects from arbitrary...

引用

28th SPIE Conference on Practical Holography - Materials and Applications

作者： Tai, Keigo Ichikawa, Tsubasa Sakamoto, Yuji Hokkaido Univ Grad Sch Informat Sci & Technol Kita Ku Sapporo Hokkaido 0600814 Japan

ISBN: (纸本)9780819499196

In computer-generated holograms (CGHs), various methods to generate holograms of actual objects have been proposed. One method uses the 3D information of actual objects obtained by one range sensor to generate a hologram. However, this method is unable to express occlusion because the 3D information is measured from only one viewpoint. Therefor, re-capturing 3D information is necessary to make CGHs from another viewpoint in this method. To overcome these problems, we have proposed a method that generates models of actual objects in a computer using two or more range sensors. In this method, CGHs are calculated by using multi-view images (MVIs) rendered from object models, which enables the occlusion to be expressed correctly. However, determination brightness values is unable to be performed on the whole object model because the luminance images are insufficient. This makes it difficult to generate CGHs from arbitrary viewpoints at one time. In this paper, we propose a method using additional digital cameras with range sensors to generate CGHs of actual objects from arbitrary viewpoints. This method determines the brightness values of the whole model using pictures taken by the additional cameras. The proposed method is advantageous in that re-capturing 3D information is unnecessary when CGHs are generated from another viewpoint. This method calculates CGHs with MVIs rendered from the model having brightness values. Results of computer simulations and optical reconstruction showed that it was possible to generate CGHs of actual objects with occlusion from arbitrary viewpoints.

关键词： Computer-generated hologram multi-view projection arbitrary viewpoint range sensor

来源：评论

学校读者我要写书评

暂无评论

The Design of an Achromatic Two Dimensional Deflection Prism

The Design of an Achromatic Two Dimensional Deflection Prism

引用

IEEE 6th International Conference on Advanced Infocomm Technology (ICAIT)

作者： Shih, Wen-Tse Lu, Wei-Fu Yang, Bor-Wen Minghsin Univ Sci & Technol Dept Optoelect Syst Engn Hsinchu Taiwan Asia Univ Dept Comp Sci & Informat Engn Taichung Taiwan

An achromatic two dimensional deflection prism was designed for applications of autostereoscopic or multi-view projection display systems. The relation between the beam deflection and the wedges was analyzed theoretic... 详细信息

ISBN: (纸本)9781479904648

关键词： two dimensional deflection prism achromatic biprism autostereoscopic projection multi-view projection display

来源：评论

学校读者我要写书评

暂无评论

Computer-generated holograms at arbitrary positions using multi-view images

Computer-generated holograms at arbitrary positions using mu...

引用

Conference on Practical Holography XXVI - Materials and Applications

作者： Ohsawa, Yusuke Sakamoto, Yuji Hokkaido Univ Grad Sch Informat Sci & Technol Sapporo Hokkaido 0600814 Japan

ISBN: (纸本)9780819489241

Computer-generated holograms (CGHs), which are generated by simulating the recording process of a hologram in a computer, are noted as an ideal three-dimensional (3D) display technology. However, with CGHs it is necessary to create precise 3D model data based on objects that already exist, and it is difficult to do this. To solve this problem, there has been much research on generating CGHs using multi-view images (MVIs). MVIs make it possible to generate CGHs from real-existing objects in natural light. A method using ordinary digital cameras resulted in high-resolution reconstructed images without the need for any special devices, but with this method it is necessary to capture a huge number of images or to use a huge number of cameras to ensure a sufficient continuous motion parallax. This is simply not realistic for the construction of 3D display applications. In this paper, we describe a method of generating voxel models from captured images and then using the MVIs obtained by the models to generate CGHs. We generate voxel models by SFS, determine voxel value using the captured images, and render voxel models into MVIs. Using this method enables us to arrange holograms at arbitrary positions in the range in which MVIs are generated correctly. We can also obtain a sufficient continuous motion parallax by generating MVIs obtained from voxel models in spite of capturing only a small number of images. Results of optical experiments demonstrated the effectiveness of the proposed method.

关键词： Computer generated hologram multi-view projection voxel model arbitrary viewpoint

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：