检索结果-内蒙古大学图书馆

HDR image reconstruction from a single exposure using deep CNNs

ACM TRANSACTIONS ON GRAPHICS 2017年第6期36卷 1-15页

作者： Eilertsen, Gabriel Kronander, Joel Denes, Gyorgy Mantiuk, Rafal K. Unger, Jonas Linkoping Univ Linkoping Sweden Univ Cambridge Cambridge England

Camera sensors can only capture a limited range of luminance simultaneously, and in order to create high dynamic range (HDR) images a set of different exposures are typically combined. In this paper we address the problem of predicting information that have been lost in saturated image areas, in order to enable HDR reconstruction from a single exposure. We show that this problem is well-suited for deep learning algorithms, and propose a deep convolutional neural network (CNN) that is specifically designed taking into account the challenges in predicting HDR values. To train the CNN we gather a large dataset of HDR images, which we augment by simulating sensor saturation for a range of cameras. To further boost robustness, we pre-train the CNN on a simulated HDR dataset created from a subset of the MIT Places database. We demonstrate that our approach can reconstruct high-resolution visually convincing HDR results in a wide range of situations, and that it generalizes well to reconstruction of images captured with arbitrary and low-end cameras that use unknown camera response functions and post-processing. Furthermore, we compare to existing methods for HDR expansion, and show high quality results also for image based lighting. Finally, we evaluate the results in a subjective experiment performed on an HDR display. This shows that the reconstructed HDR images are visually convincing, with large improvements as compared to existing methods.

关键词： HDR reconstruction inverse tone-mapping deep learning convolutional network

来源：评论

学校读者我要写书评

暂无评论

Real-time slip angle measurements using digital image correlation 19

Real-time slip angle measurements using digital image correl...

引用

19th International and 14th European-African Regional conference of the International Society for Terrain-Vehicle, ISTVS 2017

作者： Johnson, Devin K. Botha, Theunis R. Els, P. Schalk University of Pretoria 1 Lynnwood Road Pretoria0002 South Africa

ISBN: (纸本)9781942112495

In vehicle dynamics, all the external forces (apart from aerodynamic forces) are generated at the tyre-road interface. To fully understand the dynamics that govern these forces, parameters such as terrain profile, side-slip angle and longitudinal slip are critical. The ability to directly measure these parameters in real-time will aid and improve many driver assist systems such as ABS and traction control, especially over rough terrain. In this paper, the use of image correlation is investigated to measure the side-slip angle in real time. Digital image correlation is the process of comparing and analysing changes in sequential images by applying software algorithms to these images. Previous research has proven that digital image correlation can be used to accurately measure these critical parameters over rough off-road terrain using inexpensive, off-the-shelf cameras. However, this was achieved in post processing and not implemented in real time due to the large computation times of the algorithms. Commercial side-slip angle sensors are available but they are costly. They are also restricted to low side-slip angle, give unsatisfactory results at lower speeds and have trouble measuring over uneven terrain. In this paper, the maximum obtainable sampling frequency and maximum operating speeds are investigated using this method. Tests were conducted on flat concrete surfaces and showed that the vehicle side-slip angle could be measured in real-time at nearly highway speeds. The method proposed provides an inexpensive alternative to commercial sensors and estimation to provide a direct measurement of the side-slip angle in real-time.

关键词： Landforms

来源：评论

学校读者我要写书评

暂无评论

FFTLasso: Large-Scale LASSO in the Fourier Domain 30

FFTLasso: Large-Scale LASSO in the Fourier Domain

引用

30th IEEE/CVF conference on Computer Vision and Pattern Recognition (CVPR)

作者： Bibi, Adel Itani, Hani Ghanem, Bernard King Abdullah Univ Sci & Technol KAUST Thuwal Saudi Arabia

ISBN: (纸本)9781538604571

In this paper, we revisit the LASSO sparse representation problem, which has been studied and used in a variety of different areas, ranging from signal processing and information theory to computer vision and machine learning. In the vision community, it found its way into many important applications, including face recognition, tracking, super resolution, image denoising, to name a few. Despite advances in efficient sparse algorithms, solving large-scale LASSO problems remains a challenge. To circumvent this difficulty, people tend to downsample and subsample the problem (e.g. via dimensionality reduction) to maintain a manageable sized LASSO, which usually comes at the cost of losing solution accuracy. This paper proposes a novel circulant reformulation of the LASSO that lifts the problem to a higher dimension, where ADMM can be efficiently applied to its dual form. Because of this lifting, all optimization variables are updated using only basic element-wise operations, the most computationally expensive of which is a 1D FFT. In this way, there is no need for a linear system solver nor matrix-vector multiplication. Since all operations in our FFTLasso method are element-wise, the subproblems are completely independent and can be trivially parallelized (e.g. on a GPU). The attractive computational properties of FFTLasso are verified by extensive experiments on synthetic and real data and on the face recognition task. They demonstrate that FFTLasso scales much more effectively than a state-of-the-art solver.

关键词： Dictionaries Face recognition Sparse matrices Convolutional codes Linear systems Graphics processing units Encoding

来源：评论

学校读者我要写书评

暂无评论

Performance of mid infrared spectroscopy in skin cancer cell type identification 15

Performance of mid infrared spectroscopy in skin cancer cell...

引用

conference on Optical Biopsy XV - Toward Real-Time Spectroscopic Imaging and Diagnosis

作者： Kastl, Lena Kemper, Bjoern Lloyd, Gavin R. Nallala, Jayakrupakar Stone, Nick Naranjo, Valery Penaranda, Francisco Schnekenburger, Jurgen Univ Munster Biomed Technol Ctr Mendelstr 17 D-48149 Munster Germany Gloucestershire Hosp NHS Fdn Trust Biophoton Res Unit Gloucester England Exeter Univ Dept Phys Exeter Devon England Univ Politecn Valencia Inst Int Invest Bioingn & Tecnol Orientada Ser Hu Valencia Spain

ISBN: (纸本)9781510605619;9781510605626

Marker free optical spectroscopy is a powerful tool for the rapid inspection of pathologically suspicious skin lesions and the non-invasive detection of early skin tumors. This goal can be reached by the combination of signal localization and the spectroscopical detection of chemical cell signatures. We here present the development and application of mid infrared spectroscopy (midIR) for the analysis of skin tumor cell types and three dimensional tissue phantoms towards the application of midIR spectroscopy for fast and reliable skin diagnostics. We developed standardized in vitro skin systems with increasing complexity, from single skin cell types as fibroblasts, keratinocytes and melanoma cells, to mixtures of these and finally three dimensional skin cancer phantoms. The cell systems were characterized with different systems in the midIR range up to 12 mu m. The analysis of the spectra by novel data processing algorithms demonstrated the clear separation of all cell types, especially melanoma cells. Special attention and algorithm training was required for closely related mesenchymal cell types as dedifferentiated melanoma cells and fibroblasts. Proof of concept experiments with mixtures of in vivo fluorescence labelled skin cell types allowed the test of the new algorithms performance for the identification of specific cell types. The intense training of the software systems with various samples resulted in a increased sensitivity and specificity of the combined midIR and software system. These data highlight the potential of midIR spectroscopy as sensitive and specific future optical biopsy technology.

关键词： mid infrared spectroscopy skin cancer three dimensional skin phantoms image analysis algorithms

来源：评论

学校读者我要写书评

暂无评论

A novel optical investigation technique for railroad track inspection and assessment 11

A novel optical investigation technique for railroad track i...

引用

conference on Nondestructive Characterization and Monitoring of Advanced Materials, Aerospace, and Civil Infrastructure

作者： Sabato, Alessandro Beale, Christopher H. Niezrecki, Christopher Univ Massachusetts Lowell 1 Univ Ave Lowell MA 01852 USA

ISBN: (纸本)9781510608238;9781510608245

Track failures due to cross tie degradation or loss in ballast support may result in a number of problems ranging from simple service interruptions to derailments. Structural Health Monitoring (SHM) of railway track is important for safety reasons and to reduce downtime and maintenance costs. For this reason, novel and cost-effective track inspection technologies for assessing tracks' health are currently insufficient and needed. Advancements achieved in recent years in cameras technology, optical sensors, and image-processing algorithms have made machine vision, Structure from Motion (SfM), and three-dimensional (3D) Digital image Correlation (DIC) systems extremely appealing techniques for extracting structural deformations and geometry profiles. Therefore, optically based, non-contact measurement techniques may be used for assessing surface defects, rail and tie deflection profiles, and ballast condition. In this study, the design of two camera-based measurement systems is proposed for crossties-ballast condition assessment and track examination purposes. The first one consists of four pairs of cameras installed on the underside of a rail car to detect the induced deformation and displacement on the whole length of the track's cross tie using 3D DIC measurement techniques. The second consists of another set of cameras using SfM techniques for obtaining a 3D rendering of the infrastructure from a series of two-dimensional (2D) images to evaluate the state of the track qualitatively. The feasibility of the proposed optical systems is evaluated through extensive laboratory tests, demonstrating their ability to measure parameters of interest (e.g. crosstie's full-field displacement, vertical deflection, shape, etc.) for assessment and SHM of railroad track.

关键词： Ballast crosstie Digital image Correlation railroad Structure from Motion Structural Health Monitoring

来源：评论

学校读者我要写书评

暂无评论

Local statistics estimation for rapid frequency selective extrapolation

Local statistics estimation for rapid frequency selective ex...

引用

International conference on systems, Signals and image processing, IWSSIP

作者： Nils Genser Jürgen Seiler André Kaup Multimedia Communications and Signal Processing Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) Erlangen Germany

ISBN: (纸本)9781509063451

Fast error concealment algorithms play an important role in image and video signal processing. Within this paper, a novel and highly accelerated Frequency Selective Extrapolation adaption is introduced for rapid image error concealment. The state-of-the-art complex-valued Frequency Selective Extrapolation runs with fixed parameters, e.g., a constant number of iterations, so far. In this paper, the algorithm is accelerated by determining local statistics of neighboring blocks. As a consequence, internal parameters can be estimated to match the processed image content dynamically instead of using predefined constants. Due to this, it is possible to achieve the same reconstruction quality as the state-of-the-art algorithm, while the execution time decreases on average by 41.30 % and at best by up to 57.08 %.

关键词： image reconstruction Estimation Extrapolation Standards Frequency estimation Signal processing algorithms Acceleration

来源：评论

学校读者我要写书评

暂无评论

Disparity map estimation uses block matching algorithm and bilateral filter 4

Disparity map estimation uses block matching algorithm and b...

引用

4th International conference on Information Technology systems and Innovation, ICITSI 2017

作者： Hamzah, Rostam Affendi Ghani, S. Fakhar Abd Kadmin, A. Fauzan Hamid, M. Saad Salam, S. Wook, T. M. Faisal T. Fakulti Teknologi Kejuruteraan Kampus Teknologi Universiti Teknikal Malaysia Melaka Hang Tuah Jaya Durian Tunggal Melaka76100 Malaysia

ISBN: (纸本)9781538631003

This paper proposes a new disparity map algorithm which uses block matching and edge-preserving filter. The Sum of Absolute Differences (SAD) algorithm uses block matching technique which produces accurate results on the disparity map for the textured regions. However, this algorithm is sensitive to low texture areas and high noise on images with high different brightness and contrast of images. To get over these problems, the proposed algorithm utilizes edge-preserving filter which is known as Bilateral Filter (BF). The BF kernel well-recovered low texture areas which is able to reduce noise and sharpen the images. Additionally, BF is strong against the distortions due to high brightness and contrast. The proposed work in this paper produces accurate results and performs much better compared to some established algorithms based on the quantitative and qualitative measurements using standard stereo benchmarking evaluation from the Middlebury. © 2017 IEEE.

关键词： Stereo image processing

来源：评论

学校读者我要写书评

暂无评论

Moving beyond sparsity by transferring expert knowledge

Moving beyond sparsity by transferring expert knowledge

引用

Annual conference on Information Sciences and systems (CISS)

作者： Arian Maleki Columbia University New York NY US

ISBN: (纸本)9781509026975

Summary form only given. In the last fifty years, signal processing and machine learning experts have developed a wide range of algorithms to address a diverse set of inference and information processing tasks. New algorithms are often developed based on new structures that experts discover in data. For instance, the JPEG compression uses the sparse representation of images in the discrete cosine domain. In this talk, all algorithms and structures that researchers have discovered are called expert knowledge. Ideally, this knowledge is expected to be readily transferable to new applications. The evolution of imaging has shown that this natural expectation has still not been met; researchers have worked for ten years on compressed imaging, and the “signal structures” that have been developed for this task are still simpler than those used for image denoising.

关键词： Inference algorithms Signal processing algorithms Machine learning algorithms Information processing image coding Imaging Signal processing

来源：评论

学校读者我要写书评

暂无评论

Impact of interactivity on the assessment of quality of experience for light field content 9

Impact of interactivity on the assessment of quality of expe...

引用

9th International conference on Quality of Multimedia Experience, QoMEX 2017

作者： Viola, Irene Řeřábek, Martin Ebrahimi, Touradj Multimedia Signal Processing Group EPFL Lausanne Switzerland

ISBN: (纸本)9781538640241

The recent advances in light field imaging are changing the way in which visual content is captured, processed and consumed. Storage and delivery systems for light field images rely on efficient compression algorithms. Such algorithms must additionally take into account the feature-rich rendering for light field content. Therefore, a proper evaluation of visual quality is essential to design and improve coding solutions for light field content. Consequently, the design of subjective tests should also reflect the light field rendering process. This paper aims at presenting and comparing two methodologies to assess the quality of experience in light field imaging. The first methodology uses an interactive approach, allowing subjects to engage with the light field content when assessing it. The second, on the other hand, is completely passive to ensure all the subjects will have the same experience. Advantages and drawbacks of each approach are compared by relying on statistical analysis of results and conclusions are drawn. The obtained results provide useful insights for future design of evaluation techniques for light field content. © 2017 IEEE.

关键词： image compression

来源：评论

学校读者我要写书评

暂无评论

Fault Tolerant Control of an Industrial Manufacturing Process Using image processing

Fault Tolerant Control of an Industrial Manufacturing Proces...

引用

IEEE International conference on Environment and Electrical Engineering

作者： Jamal Siam Ali Abdo Ahmed Abdou Ashraf Al-Rimawi Hakam Shehadeh Electrical and Computer Engineering Department Birzeit University Birzeit Palestine Electronic and Communication Engineering Department Al-Quds University Jerusalem Palestine

This paper proposes a fault-tolerant control technique, for industrial applications, based on image processing. It is known that, sensor faults disturb the normal sequences of manufacturing operations. This causes a serious delay in the manufacturing process, or even it terminated for maintenance. Cameras with ad-hoc video and image processing algorithms can be used to detect and track objects, thus they can play the role of sensors in industrial application. The proposed technique uses visual information, collected by cameras, to raise the fault-tolerant of the manufacturing system. The visual information is used to generate a redundant sensing signal to substitute the faulty-sensor signal in case of faults. The main advantages of the proposed technique are: the continuity throughout the manufacturing process as well as the detection and isolation of faulty-sensors. The proposed technique was applied to a real production line model to illustrate its performance.

关键词： image processing fault tolerant control manufacturing process fault detection

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：