Pub Date : 2017-07-01DOI: 10.23919/ICIF.2017.8009692
Honglei Lin, Shu-Li Sun
This paper is concerned with the distributed fusion estimation problem for a class of multi-sensor non-uniform sampling systems with correlated noises and fading measurements. The state is updated uniformly and the sensors sample measurement data randomly. The process noise and different measurement noises are correlated at the same instant. Moreover, the fading measurement phenomena may occur in different sensor channels. The independent random variables obeying different certain probability distributions over different known intervals are employed to describe the phenomena. Based on the measurement augmentation method, the state space model is reconstructed in which the asynchronous sampling estimation problem is transformed to the synchronous one. Afterwards, local optimal filters are designed by using an innovation analysis approach. Then, the filtering error cross-covariance matrices between any two local filters are derived. At last, the optimal matrix-weighted distributed fusion filter is given in the linear unbiased minimum variance sense. Simulation results show the effectiveness of the proposed algorithms.
{"title":"Distributed fusion estimation for multi-sensor non-uniform sampling systems with correlated noises and fading measurements","authors":"Honglei Lin, Shu-Li Sun","doi":"10.23919/ICIF.2017.8009692","DOIUrl":"https://doi.org/10.23919/ICIF.2017.8009692","url":null,"abstract":"This paper is concerned with the distributed fusion estimation problem for a class of multi-sensor non-uniform sampling systems with correlated noises and fading measurements. The state is updated uniformly and the sensors sample measurement data randomly. The process noise and different measurement noises are correlated at the same instant. Moreover, the fading measurement phenomena may occur in different sensor channels. The independent random variables obeying different certain probability distributions over different known intervals are employed to describe the phenomena. Based on the measurement augmentation method, the state space model is reconstructed in which the asynchronous sampling estimation problem is transformed to the synchronous one. Afterwards, local optimal filters are designed by using an innovation analysis approach. Then, the filtering error cross-covariance matrices between any two local filters are derived. At last, the optimal matrix-weighted distributed fusion filter is given in the linear unbiased minimum variance sense. Simulation results show the effectiveness of the proposed algorithms.","PeriodicalId":148407,"journal":{"name":"2017 20th International Conference on Information Fusion (Fusion)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116716102","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The ability to recognize physical activity, such as sedentary, driving, riding, daily activities and effective training, is useful for health conscious users to catalogue their daily activities and to develop good exercise routines. Conventional activity recognition algorithms require complex calculations, which are not suitable for wearable devices developed on low-cost, low-power hardware platforms. In this paper, inspired by the text mining related work, we design a novel activity recognition algorithm, which is named “Motionword”. In the wearable device proper, a lightweight recognition algorithm is adopted to compute in real-time predefined atomic events, and count the frequency that these events occur, resulting in a data summary, and then the data summary is transmitted to the platform. On the platform, intelligent method is used to identify and categorize the user's main activity into 5 classes. The test results on a dataset composed of 110 user∗day real world data, contributed by 10 users, show that the recognition accuracy is 95.52%. The Motionword algorithm is capable of achieving accurate activity recognition results without additional hardware cost or power consumption.
{"title":"Motionword: An activity recognition algorithm based on intelligent terminal and cloud","authors":"Zhen-Jie Yao, Zhi-Peng Zhang, Junyan Wang, Li-Qun Xu","doi":"10.23919/ICIF.2017.8009780","DOIUrl":"https://doi.org/10.23919/ICIF.2017.8009780","url":null,"abstract":"The ability to recognize physical activity, such as sedentary, driving, riding, daily activities and effective training, is useful for health conscious users to catalogue their daily activities and to develop good exercise routines. Conventional activity recognition algorithms require complex calculations, which are not suitable for wearable devices developed on low-cost, low-power hardware platforms. In this paper, inspired by the text mining related work, we design a novel activity recognition algorithm, which is named “Motionword”. In the wearable device proper, a lightweight recognition algorithm is adopted to compute in real-time predefined atomic events, and count the frequency that these events occur, resulting in a data summary, and then the data summary is transmitted to the platform. On the platform, intelligent method is used to identify and categorize the user's main activity into 5 classes. The test results on a dataset composed of 110 user∗day real world data, contributed by 10 users, show that the recognition accuracy is 95.52%. The Motionword algorithm is capable of achieving accurate activity recognition results without additional hardware cost or power consumption.","PeriodicalId":148407,"journal":{"name":"2017 20th International Conference on Information Fusion (Fusion)","volume":"20 19-20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116721236","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-07-01DOI: 10.23919/ICIF.2017.8009818
Xiaolei Bian, X. Li
This paper deals with the problem of estimating the state of a discrete-time stochastic linear system based on data collected from multiple sensors with limited communication resources. For the cases of transmitting measurements and local state estimates, respectively, we design data-driven communication schemes based on a normalized innovation vector and corresponding fusion rules in the (approximate) minimum mean square error (MMSE) sense. These communication schemes can achieve a trade-off between communication costs and estimation performance. These fusion rules can allow the estimator to improve its estimate based on the fact that no transmission of data indicates a small innovation. A simulation example is provided to confirm the effectiveness of the proposed strategies.
{"title":"Estimation fusion with data-driven communication","authors":"Xiaolei Bian, X. Li","doi":"10.23919/ICIF.2017.8009818","DOIUrl":"https://doi.org/10.23919/ICIF.2017.8009818","url":null,"abstract":"This paper deals with the problem of estimating the state of a discrete-time stochastic linear system based on data collected from multiple sensors with limited communication resources. For the cases of transmitting measurements and local state estimates, respectively, we design data-driven communication schemes based on a normalized innovation vector and corresponding fusion rules in the (approximate) minimum mean square error (MMSE) sense. These communication schemes can achieve a trade-off between communication costs and estimation performance. These fusion rules can allow the estimator to improve its estimate based on the fact that no transmission of data indicates a small innovation. A simulation example is provided to confirm the effectiveness of the proposed strategies.","PeriodicalId":148407,"journal":{"name":"2017 20th International Conference on Information Fusion (Fusion)","volume":"313 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131518306","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-07-01DOI: 10.23919/ICIF.2017.8009845
Yu Liu, Jun Liu, Cong'an Xu, Lin Qi, Shun Sun, Ziran Ding
Considering the convergence rate is a very important issue as distributed sensors networks usually consist of low-powered wireless devices and speeding up the consensus convergence rate is also important to reduce the number of messages exchanged among neighbors, a new adaptive method for weight assignment of communication links between sensor nodes is proposed based on the dynamic network topology. Based on the adaptive weight assignment method, an improved Kalman consensus filter (KCF) named IKCF is tailored in this letter for distributed state estimation in sensor networks with cluster structure. Furthermore, the experiments demonstrate the adaptive weight assignment method is effective for distributed state estimation when the sensor network is sparsely deployed. In addition, the simulation results also validate the superior performance of the new algorithm and show that IKCF is an excellent algorithm for multi-clusters sensor networks. And there is no additional communication overhead in IKCF because only some local knowledge is used to autonomously calculate the adaptive consensus rate parameter for each node.
{"title":"Consensus algorithm for distributed state estimation in multi-clusters sensor network","authors":"Yu Liu, Jun Liu, Cong'an Xu, Lin Qi, Shun Sun, Ziran Ding","doi":"10.23919/ICIF.2017.8009845","DOIUrl":"https://doi.org/10.23919/ICIF.2017.8009845","url":null,"abstract":"Considering the convergence rate is a very important issue as distributed sensors networks usually consist of low-powered wireless devices and speeding up the consensus convergence rate is also important to reduce the number of messages exchanged among neighbors, a new adaptive method for weight assignment of communication links between sensor nodes is proposed based on the dynamic network topology. Based on the adaptive weight assignment method, an improved Kalman consensus filter (KCF) named IKCF is tailored in this letter for distributed state estimation in sensor networks with cluster structure. Furthermore, the experiments demonstrate the adaptive weight assignment method is effective for distributed state estimation when the sensor network is sparsely deployed. In addition, the simulation results also validate the superior performance of the new algorithm and show that IKCF is an excellent algorithm for multi-clusters sensor networks. And there is no additional communication overhead in IKCF because only some local knowledge is used to autonomously calculate the adaptive consensus rate parameter for each node.","PeriodicalId":148407,"journal":{"name":"2017 20th International Conference on Information Fusion (Fusion)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132289589","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-07-01DOI: 10.23919/ICIF.2017.8009863
Nils Rexin, Dominik Nuss, Stephan Reuter, K. Dietmayer
The dynamic grid map illustrates the environment of robots with moving and static obstacles. Nuss et al. describe in [1] an implementation of this grid map, in which the state of the grid cells is to be modeled as a random finite set (RFS) based on a stochastic measurement system. For a real-time implementation this approach was approximated with Dempster-Shafer (DS). For this Nuss et al. design the areas without information (unknown areas) so, that no probabilistic calculations are executed. Only in the field of view, hypotheses represent the dynamic behavior of objects. This hypotheses are generated with particles. Therefore, in [1] it was proposed to extend this modeling. In this paper a pure Bayes approach is presented, which calculates all areas of the dynamic grid map probabilistic. Now, the resulting modeling generates hypotheses, which represent the dynamic behavior of unobservable objects. Thus, objects moving out of unknown areas can be detected more quickly. This leads to a more intuitive understanding as well as representation of the environment.
{"title":"Modeling occluded areas in dynamic grid maps","authors":"Nils Rexin, Dominik Nuss, Stephan Reuter, K. Dietmayer","doi":"10.23919/ICIF.2017.8009863","DOIUrl":"https://doi.org/10.23919/ICIF.2017.8009863","url":null,"abstract":"The dynamic grid map illustrates the environment of robots with moving and static obstacles. Nuss et al. describe in [1] an implementation of this grid map, in which the state of the grid cells is to be modeled as a random finite set (RFS) based on a stochastic measurement system. For a real-time implementation this approach was approximated with Dempster-Shafer (DS). For this Nuss et al. design the areas without information (unknown areas) so, that no probabilistic calculations are executed. Only in the field of view, hypotheses represent the dynamic behavior of objects. This hypotheses are generated with particles. Therefore, in [1] it was proposed to extend this modeling. In this paper a pure Bayes approach is presented, which calculates all areas of the dynamic grid map probabilistic. Now, the resulting modeling generates hypotheses, which represent the dynamic behavior of unobservable objects. Thus, objects moving out of unknown areas can be detected more quickly. This leads to a more intuitive understanding as well as representation of the environment.","PeriodicalId":148407,"journal":{"name":"2017 20th International Conference on Information Fusion (Fusion)","volume":"136 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"113991612","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this paper, the maximum correntropy (MC) criterion is used as the cost function in the online sequential extreme learning machine (OS-ELM) algorithm and constraint OS-ELM (COS-ELM) algorithm, generating the proposed OS-ELM based on maximum correntropy (OS-ELM-MC) and COS-ELM based on maximum correntropy (COS-ELM-MC). In comparison with OS-ELM and COS-ELM, the proposed OS-ELM-MC and COS-ELM-MC present superior performance in non-Gaussian noise environments and almost the same performance in Gaussian noise environments. As an important parameter, the hidden node number is also discussed by simulations in this paper. Simulations on the examples of Mackey-Glass (MG) chaotic time series prediction and nonlinear regression validate the efficiency of the proposed OS-ELM-MC and COS-ELM-MC.
{"title":"Online sequential extreme learning machine algorithms based on maximum correntropy citerion","authors":"Wenyue Wang, Chunfen Shi, Wanli Wang, Lujuan Dang, Shiyuan Wang, Shukai Duan","doi":"10.23919/ICIF.2017.8009772","DOIUrl":"https://doi.org/10.23919/ICIF.2017.8009772","url":null,"abstract":"In this paper, the maximum correntropy (MC) criterion is used as the cost function in the online sequential extreme learning machine (OS-ELM) algorithm and constraint OS-ELM (COS-ELM) algorithm, generating the proposed OS-ELM based on maximum correntropy (OS-ELM-MC) and COS-ELM based on maximum correntropy (COS-ELM-MC). In comparison with OS-ELM and COS-ELM, the proposed OS-ELM-MC and COS-ELM-MC present superior performance in non-Gaussian noise environments and almost the same performance in Gaussian noise environments. As an important parameter, the hidden node number is also discussed by simulations in this paper. Simulations on the examples of Mackey-Glass (MG) chaotic time series prediction and nonlinear regression validate the efficiency of the proposed OS-ELM-MC and COS-ELM-MC.","PeriodicalId":148407,"journal":{"name":"2017 20th International Conference on Information Fusion (Fusion)","volume":"115 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115568264","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-07-01DOI: 10.23919/ICIF.2017.8009734
Yifan Xie, Hyoung-Won Kim, H. Kim, T. Song
The conventional multi-target tracking (MTT) algorithms usually suffer from computational intractability problem. The appearance of Iterative Joint Integrated Probabilistic Data Association (iJIPDA) filter solves this problem by providing a tradeoff between the tracking performance and computational cost for computational resource management of sensor systems. However, the iJIPDA filter essentially involves repetitive computation which makes it impractical to perform at high levels. Thus we provide an improved iJIPDA filter which prevents repetitive computations and increases the computational efficiency such that better performances can be obtained within limited time.
{"title":"Reduction of computational load for implementing iJIPDA filter","authors":"Yifan Xie, Hyoung-Won Kim, H. Kim, T. Song","doi":"10.23919/ICIF.2017.8009734","DOIUrl":"https://doi.org/10.23919/ICIF.2017.8009734","url":null,"abstract":"The conventional multi-target tracking (MTT) algorithms usually suffer from computational intractability problem. The appearance of Iterative Joint Integrated Probabilistic Data Association (iJIPDA) filter solves this problem by providing a tradeoff between the tracking performance and computational cost for computational resource management of sensor systems. However, the iJIPDA filter essentially involves repetitive computation which makes it impractical to perform at high levels. Thus we provide an improved iJIPDA filter which prevents repetitive computations and increases the computational efficiency such that better performances can be obtained within limited time.","PeriodicalId":148407,"journal":{"name":"2017 20th International Conference on Information Fusion (Fusion)","volume":"120 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116586964","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-07-01DOI: 10.23919/ICIF.2017.8009621
J. Liu, XiaoChao Li
We present a novel joint detection and tracking algorithm using raw measurements, in a compressed sensing framework. The sparse vector representing the state space is directly reconstructed, which transforms the nonlinear estimation problem into a linear one through sparse representation. A number of significant grids are obtained based on the sparse vector, indicating the positions of multiple potential targets in the state space. Therefore, the multi-model posterior distribution of the state can be sparsely represented by a number of modes centering around the significant grids at each scan. Consequently, a novel algorithm named sparse mixture particle filter is proposed in this work, which provides a sparse representation of the multi-model posterior distribution by identifying the significant grids. Furthermore, a novel adaptive sparse mixture particle filter algorithm is proposed to tackle the high coherence and high computation burden problems, by constructing a compact dictionary based on the state space with low resolution. The simulation results show that the proposed adaptive sparse mixture particle filter based joint detection and tracking algorithm can successfully detect and track multiple targets, which appear and disappear at different times, as well as track closely spaced targets with similar dynamic model.
{"title":"Adaptive sparse mixture particle filter","authors":"J. Liu, XiaoChao Li","doi":"10.23919/ICIF.2017.8009621","DOIUrl":"https://doi.org/10.23919/ICIF.2017.8009621","url":null,"abstract":"We present a novel joint detection and tracking algorithm using raw measurements, in a compressed sensing framework. The sparse vector representing the state space is directly reconstructed, which transforms the nonlinear estimation problem into a linear one through sparse representation. A number of significant grids are obtained based on the sparse vector, indicating the positions of multiple potential targets in the state space. Therefore, the multi-model posterior distribution of the state can be sparsely represented by a number of modes centering around the significant grids at each scan. Consequently, a novel algorithm named sparse mixture particle filter is proposed in this work, which provides a sparse representation of the multi-model posterior distribution by identifying the significant grids. Furthermore, a novel adaptive sparse mixture particle filter algorithm is proposed to tackle the high coherence and high computation burden problems, by constructing a compact dictionary based on the state space with low resolution. The simulation results show that the proposed adaptive sparse mixture particle filter based joint detection and tracking algorithm can successfully detect and track multiple targets, which appear and disappear at different times, as well as track closely spaced targets with similar dynamic model.","PeriodicalId":148407,"journal":{"name":"2017 20th International Conference on Information Fusion (Fusion)","volume":"80 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116224104","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-07-01DOI: 10.23919/ICIF.2017.8009873
Chao Wan, Yongxin Gao, X. Li
This paper studies and formulates the problem of distributed filtering with a diffusion strategy for state estimation of a dynamic system by using observations from sensors in a network. The sensor-nodes have estimation ability and work in a collaborative manner. The information transmission across the network abides by the diffusion strategy that each node communicates only with its neighbors. First, we propose a cost function for a trade-off between accuracy and consensus. Then, we derive our algorithm based on this cost and analyze its mean-square performance. Illustrative numerical examples are provided to verify the good performance of our method.
{"title":"Distributed filtering over networks based on diffusion strategy","authors":"Chao Wan, Yongxin Gao, X. Li","doi":"10.23919/ICIF.2017.8009873","DOIUrl":"https://doi.org/10.23919/ICIF.2017.8009873","url":null,"abstract":"This paper studies and formulates the problem of distributed filtering with a diffusion strategy for state estimation of a dynamic system by using observations from sensors in a network. The sensor-nodes have estimation ability and work in a collaborative manner. The information transmission across the network abides by the diffusion strategy that each node communicates only with its neighbors. First, we propose a cost function for a trade-off between accuracy and consensus. Then, we derive our algorithm based on this cost and analyze its mean-square performance. Illustrative numerical examples are provided to verify the good performance of our method.","PeriodicalId":148407,"journal":{"name":"2017 20th International Conference on Information Fusion (Fusion)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124957397","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Several models based on deep neural networks have applied to single image super-resolution and obtained great improvements in terms of both reconstruction accuracy and computational performance. All these methods focus either on performing the super-resolution (SR) reconstruction operation in the high resolution (HR) space after upscaling with a single filter, usually bicubic interpolation, or optimizing parts of the reconstruction pipeline. Then the studies of network-based model advance to attempting to shrink the feature dimension of the nonlinear mapping considering the tradeoff between accuracy and time cost. In this paper, we present an improved convolutional neural network (CNN) architecture based on channels combination, which benefits from both quick training and accuracy gain. In addition, we propose that the feature maps can be extracted in the LR space and an efficient multi-channel convolution layer which learns an array of upscaling filters, specifically trained for each feature map, to upscale the final HR feature maps into the HR output. We explore different settings and evaluate the proposed approach using images from publicly available datasets and show that it performs significantly better (about + 0.3 dB margin on term of PSNR and + 0.03 on term of SSIM than previous works) with better visual appearance.
{"title":"A novel convolutional neural network architecture for image super-resolution based on channels combination","authors":"Cun-Gen Liu, Yuanxiang Li, Jianhua Luo, Yongjun Zhou","doi":"10.23919/ICIF.2017.8009771","DOIUrl":"https://doi.org/10.23919/ICIF.2017.8009771","url":null,"abstract":"Several models based on deep neural networks have applied to single image super-resolution and obtained great improvements in terms of both reconstruction accuracy and computational performance. All these methods focus either on performing the super-resolution (SR) reconstruction operation in the high resolution (HR) space after upscaling with a single filter, usually bicubic interpolation, or optimizing parts of the reconstruction pipeline. Then the studies of network-based model advance to attempting to shrink the feature dimension of the nonlinear mapping considering the tradeoff between accuracy and time cost. In this paper, we present an improved convolutional neural network (CNN) architecture based on channels combination, which benefits from both quick training and accuracy gain. In addition, we propose that the feature maps can be extracted in the LR space and an efficient multi-channel convolution layer which learns an array of upscaling filters, specifically trained for each feature map, to upscale the final HR feature maps into the HR output. We explore different settings and evaluate the proposed approach using images from publicly available datasets and show that it performs significantly better (about + 0.3 dB margin on term of PSNR and + 0.03 on term of SSIM than previous works) with better visual appearance.","PeriodicalId":148407,"journal":{"name":"2017 20th International Conference on Information Fusion (Fusion)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121899168","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：