Pub Date : 2017-05-01DOI: 10.1109/IWSSIP.2017.7965583
Tianyu Geng, Guiling Sun, Yi Xu, Zhouzhou Li
Group sparse representation (GSR) has shown great potential in image Compressive Sensing (CS) recovery, which can be considered as a low rank matrix approximation problem. The nuclear norm minimization can only minimize all the singular values simultaneously. Recent advances have suggested the truncated nuclear norm minimization (TNNM) to better approximate the matrix rank. In this paper, we connect group sparse representation with truncated nuclear norm minimization for CS image recovery. Then, an implementation of fast convergence via the alternating direction method of multipliers (ADMM) is developed to solve the proposed problem. Moreover, an effective dictionary for each group is learned from the recovery image itself rather than a large number of natural image dataset. Experimental results demonstrate that the proposed GSR-TNNM method achieves a good convergence performance and is able to improve image CS recovery quality significantly compared with the state-of-the-art methods.
{"title":"Image compressive sensing using group sparse representation via truncated nuclear norm minimization","authors":"Tianyu Geng, Guiling Sun, Yi Xu, Zhouzhou Li","doi":"10.1109/IWSSIP.2017.7965583","DOIUrl":"https://doi.org/10.1109/IWSSIP.2017.7965583","url":null,"abstract":"Group sparse representation (GSR) has shown great potential in image Compressive Sensing (CS) recovery, which can be considered as a low rank matrix approximation problem. The nuclear norm minimization can only minimize all the singular values simultaneously. Recent advances have suggested the truncated nuclear norm minimization (TNNM) to better approximate the matrix rank. In this paper, we connect group sparse representation with truncated nuclear norm minimization for CS image recovery. Then, an implementation of fast convergence via the alternating direction method of multipliers (ADMM) is developed to solve the proposed problem. Moreover, an effective dictionary for each group is learned from the recovery image itself rather than a large number of natural image dataset. Experimental results demonstrate that the proposed GSR-TNNM method achieves a good convergence performance and is able to improve image CS recovery quality significantly compared with the state-of-the-art methods.","PeriodicalId":302860,"journal":{"name":"2017 International Conference on Systems, Signals and Image Processing (IWSSIP)","volume":"247 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122253289","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 : 2017-05-01DOI: 10.1109/IWSSIP.2017.7965604
M. Kociński, A. Materka, A. Deistung, J. Reichenbach, A. Lundervold
A technique is proposed for personalized modeling of cerebral brain vasculature based on three-dimensional magnetic resonance images. High resolution ToF, QSM MR images were used to build 3D geometric models of arteries and veins. To make a next step towards modeling of the whole vascular system, a surface of gray matter was extracted from T1 weighted image. Then, within selected part of the cortex, a computer-synthesized blood vessels originating from nearby artery were built as mesoscopic part of the cerebral blood system. Limitations of the ToF and QSM-based approach to development of such a comprehensive model are pointed out and discussed.
{"title":"Towards multi-scale personalized modeling of brain vasculature based on magnetic resonance image processing","authors":"M. Kociński, A. Materka, A. Deistung, J. Reichenbach, A. Lundervold","doi":"10.1109/IWSSIP.2017.7965604","DOIUrl":"https://doi.org/10.1109/IWSSIP.2017.7965604","url":null,"abstract":"A technique is proposed for personalized modeling of cerebral brain vasculature based on three-dimensional magnetic resonance images. High resolution ToF, QSM MR images were used to build 3D geometric models of arteries and veins. To make a next step towards modeling of the whole vascular system, a surface of gray matter was extracted from T1 weighted image. Then, within selected part of the cortex, a computer-synthesized blood vessels originating from nearby artery were built as mesoscopic part of the cerebral blood system. Limitations of the ToF and QSM-based approach to development of such a comprehensive model are pointed out and discussed.","PeriodicalId":302860,"journal":{"name":"2017 International Conference on Systems, Signals and Image Processing (IWSSIP)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127485173","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 : 2017-05-01DOI: 10.1109/IWSSIP.2017.7965601
M. Kociński, A. Materka, M. Elgalal, A. Majos
Possibilities of constructing an anatomically correct and accurate geometric model of brain blood vessels basing on clinical 1.5T magnetic resonance images are explored. A high-resolution ToF MR image (0.49 mm3 voxel) was used to build a reference geometric model of selected real-brain arteries. This model was STL-described and 3D printed using a photopolymer material. The printed phantom was submerged in water and scanned using a low-resolution clinical MR system (0.33×0.33×2.2 mm). Level-set segmentation of the obtained T2 images showed significant staircase effect. After T2 image resampling to 0.33mm3 voxel size, the model walls become smoother, but thin branches were still missing. A Frangi filtering-based, smooth centerline-radius vessel branches description was then developed to achieve their correct reconstruction with subvoxel accuracy. Challenges of MRI acquisition of 3D printed models are discussed.
{"title":"On accuracy of personalized 3D-printed MRI-based models of brain arteries","authors":"M. Kociński, A. Materka, M. Elgalal, A. Majos","doi":"10.1109/IWSSIP.2017.7965601","DOIUrl":"https://doi.org/10.1109/IWSSIP.2017.7965601","url":null,"abstract":"Possibilities of constructing an anatomically correct and accurate geometric model of brain blood vessels basing on clinical 1.5T magnetic resonance images are explored. A high-resolution ToF MR image (0.49 mm3 voxel) was used to build a reference geometric model of selected real-brain arteries. This model was STL-described and 3D printed using a photopolymer material. The printed phantom was submerged in water and scanned using a low-resolution clinical MR system (0.33×0.33×2.2 mm). Level-set segmentation of the obtained T2 images showed significant staircase effect. After T2 image resampling to 0.33mm3 voxel size, the model walls become smoother, but thin branches were still missing. A Frangi filtering-based, smooth centerline-radius vessel branches description was then developed to achieve their correct reconstruction with subvoxel accuracy. Challenges of MRI acquisition of 3D printed models are discussed.","PeriodicalId":302860,"journal":{"name":"2017 International Conference on Systems, Signals and Image Processing (IWSSIP)","volume":"93 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127730579","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 : 2017-05-01DOI: 10.1109/IWSSIP.2017.7965613
C. Beumier, X. Neyt
In the field of aerial surveillance, tracking targets in images is complicated by the possible motion of the camera, especially if frame differencing is used to detect moving objects. We propose in this paper to exploit the high similarity in sequences acquired from a nearly static camera. In this case distance maps grown from image edge points share many similarities and T-junctions of distance map skeletons appear to offer precisely located reference points. Each T-junction is attributed seven features: the value in the distance map, three orientations of the junction branches and R, G, B image intensities. Registering images is carried out on a division of the images into tiles, looking for the dominant translation per tile of matching T-junction points. The obtained displacement field allow for the compensation of small camera motion. This was tested on image sequences captured by a smartphone held in hand while targeting a given static scene with a few moving vehicles and pedestrians.
{"title":"Camera motion compensation from T-junctions in distance map skeleton","authors":"C. Beumier, X. Neyt","doi":"10.1109/IWSSIP.2017.7965613","DOIUrl":"https://doi.org/10.1109/IWSSIP.2017.7965613","url":null,"abstract":"In the field of aerial surveillance, tracking targets in images is complicated by the possible motion of the camera, especially if frame differencing is used to detect moving objects. We propose in this paper to exploit the high similarity in sequences acquired from a nearly static camera. In this case distance maps grown from image edge points share many similarities and T-junctions of distance map skeletons appear to offer precisely located reference points. Each T-junction is attributed seven features: the value in the distance map, three orientations of the junction branches and R, G, B image intensities. Registering images is carried out on a division of the images into tiles, looking for the dominant translation per tile of matching T-junction points. The obtained displacement field allow for the compensation of small camera motion. This was tested on image sequences captured by a smartphone held in hand while targeting a given static scene with a few moving vehicles and pedestrians.","PeriodicalId":302860,"journal":{"name":"2017 International Conference on Systems, Signals and Image Processing (IWSSIP)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129625214","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 : 1900-01-01DOI: 10.1109/IWSSIP.2017.7965570
Sandy Ardianto, Chih-Jung Chen, H. Hang
Traffic Sign Recognition (TSR) that can automatically notify and warn a vehicle driver is an essential element in the Advanced Driver Assistance System. In this study, we design and implement a real time traffic sign recognition system implemented on Advantech ARK-2121, a small computer mounted on car. The entire process is divided into two parts, the detection step and the classification step. In the detection step, we adopt color filtering, Laplacian and Gaussian filter to enhance an acquired image. Then, we detect the sign based on the contours. The recognition algorithm is accelerated by dividing an input frame into multiple blocks and process them in parallel. We improve the detection accuracy by enhancing input image before the recognition step. The SVM and HOG features are the major techniques in the recognition step. Our detection accuracy is around 91% and the classification accuracy is higher than 98% on the average.
{"title":"Real-time traffic sign recognition using color segmentation and SVM","authors":"Sandy Ardianto, Chih-Jung Chen, H. Hang","doi":"10.1109/IWSSIP.2017.7965570","DOIUrl":"https://doi.org/10.1109/IWSSIP.2017.7965570","url":null,"abstract":"Traffic Sign Recognition (TSR) that can automatically notify and warn a vehicle driver is an essential element in the Advanced Driver Assistance System. In this study, we design and implement a real time traffic sign recognition system implemented on Advantech ARK-2121, a small computer mounted on car. The entire process is divided into two parts, the detection step and the classification step. In the detection step, we adopt color filtering, Laplacian and Gaussian filter to enhance an acquired image. Then, we detect the sign based on the contours. The recognition algorithm is accelerated by dividing an input frame into multiple blocks and process them in parallel. We improve the detection accuracy by enhancing input image before the recognition step. The SVM and HOG features are the major techniques in the recognition step. Our detection accuracy is around 91% and the classification accuracy is higher than 98% on the average.","PeriodicalId":302860,"journal":{"name":"2017 International Conference on Systems, Signals and Image Processing (IWSSIP)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131573220","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：