Pub Date : 2014-11-01DOI: 10.1109/DICTA.2014.7008096
M. Awrangjeb, Guojun Lu
This paper presents a segmentation of LIDAR point cloud data for automatic extraction of building footprint. Using the ground height information from a DEM (Digital Elevation Model), the non-ground points (mainly buildings and trees) are separated from the ground points. Points on walls are removed from the set of non-ground points. The remaining non-ground points are then divided into clusters based on height and local neighbourhood. Planar roof segments are extracted from each cluster of points following a region-growing technique. Planes are initialised using coplanar points as seed points and then grown using plane compatibility tests. Once all the planar segments are extracted, a rule-based procedure is applied to remove tree planes which are small in size and randomly oriented. The neighbouring planes are then merged to obtain individual building boundaries, which are regularised based on a new feature-based technique. Corners and line-segments are extracted from each boundary and adjusted using the assumption that each short building side is parallel or perpendicular to one or more neighbouring long building sides. Experimental results on five Australian data sets show that the proposed method offers higher correctness rate in building footprint extraction than a state-of-the-art method.
{"title":"Automatic Building Footprint Extraction and Regularisation from LIDAR Point Cloud Data","authors":"M. Awrangjeb, Guojun Lu","doi":"10.1109/DICTA.2014.7008096","DOIUrl":"https://doi.org/10.1109/DICTA.2014.7008096","url":null,"abstract":"This paper presents a segmentation of LIDAR point cloud data for automatic extraction of building footprint. Using the ground height information from a DEM (Digital Elevation Model), the non-ground points (mainly buildings and trees) are separated from the ground points. Points on walls are removed from the set of non-ground points. The remaining non-ground points are then divided into clusters based on height and local neighbourhood. Planar roof segments are extracted from each cluster of points following a region-growing technique. Planes are initialised using coplanar points as seed points and then grown using plane compatibility tests. Once all the planar segments are extracted, a rule-based procedure is applied to remove tree planes which are small in size and randomly oriented. The neighbouring planes are then merged to obtain individual building boundaries, which are regularised based on a new feature-based technique. Corners and line-segments are extracted from each boundary and adjusted using the assumption that each short building side is parallel or perpendicular to one or more neighbouring long building sides. Experimental results on five Australian data sets show that the proposed method offers higher correctness rate in building footprint extraction than a state-of-the-art method.","PeriodicalId":146695,"journal":{"name":"2014 International Conference on Digital Image Computing: Techniques and Applications (DICTA)","volume":"110 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116881208","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 : 2014-11-01DOI: 10.1109/DICTA.2014.7008119
Sina Aminmansour, F. Maire, C. Wullems
Recent modelling of socio-economic costs by the Australian railway industry in 2010 has estimated the cost of level crossing accidents to exceed AU$116 million annually. To better understand the causal factors of these accidents, a video analytics application is being developed to automatically detect near- miss incidents using forward facing videos from trains. As near-miss events occur more frequently than collisions, by detecting these occurrences there will be more safety data available for analysis. The application that is being developed will improve the objectivity of near- miss reporting by providing quantitative data about the position of vehicles at level crossings through the automatic analysis of video footage. In this paper we present a novel method for detecting near-miss occurrences at railway level crossings from video data of trains. Our system detects and localizes vehicles at railway level crossings. It also detects the position of railways to calculate the distance of the detected vehicles to the railway centerline. The system logs the information about the position of the vehicles and railway centerline into a database for further analysis by the safety data recording and analysis system, to determine whether or not the event is a near-miss. We present preliminary results of our system on a dataset of videos taken from a train that passed through 14 railway level crossings. We demonstrate the robustness of our system by showing the results of our system on day and night videos.
{"title":"Near-Miss Event Detection at Railway Level Crossings","authors":"Sina Aminmansour, F. Maire, C. Wullems","doi":"10.1109/DICTA.2014.7008119","DOIUrl":"https://doi.org/10.1109/DICTA.2014.7008119","url":null,"abstract":"Recent modelling of socio-economic costs by the Australian railway industry in 2010 has estimated the cost of level crossing accidents to exceed AU$116 million annually. To better understand the causal factors of these accidents, a video analytics application is being developed to automatically detect near- miss incidents using forward facing videos from trains. As near-miss events occur more frequently than collisions, by detecting these occurrences there will be more safety data available for analysis. The application that is being developed will improve the objectivity of near- miss reporting by providing quantitative data about the position of vehicles at level crossings through the automatic analysis of video footage. In this paper we present a novel method for detecting near-miss occurrences at railway level crossings from video data of trains. Our system detects and localizes vehicles at railway level crossings. It also detects the position of railways to calculate the distance of the detected vehicles to the railway centerline. The system logs the information about the position of the vehicles and railway centerline into a database for further analysis by the safety data recording and analysis system, to determine whether or not the event is a near-miss. We present preliminary results of our system on a dataset of videos taken from a train that passed through 14 railway level crossings. We demonstrate the robustness of our system by showing the results of our system on day and night videos.","PeriodicalId":146695,"journal":{"name":"2014 International Conference on Digital Image Computing: Techniques and Applications (DICTA)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120998805","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 : 2014-11-01DOI: 10.1109/DICTA.2014.7008127
Di Yang, Srimal Jayawardena, Stephen Gould, Marcus Hutter
Computer vision techniques such as Structurefrom- Motion (SfM) and object recognition tend to fail on scenes with highly reflective objects because the reflections behave differently to the true geometry of the scene. Such image sequences may be treated as two layers superimposed over each other - the nonreflection scene source layer and the reflection layer. However, decomposing the two layers is a very challenging task as it is ill-posed and common methods rely on prior information. This work presents an automated technique for detecting reflective features with a comprehensive analysis of the intrinsic, spatial, and temporal properties of feature points. A support vector machine (SVM) is proposed to learn reflection feature points. Predicted reflection feature points are used as priors to guide the reflection layer separation. This gives more robust and reliable results than what is achieved by performing layer separation alone.
{"title":"Reflective Features Detection and Hierarchical Reflections Separation in Image Sequences","authors":"Di Yang, Srimal Jayawardena, Stephen Gould, Marcus Hutter","doi":"10.1109/DICTA.2014.7008127","DOIUrl":"https://doi.org/10.1109/DICTA.2014.7008127","url":null,"abstract":"Computer vision techniques such as Structurefrom- Motion (SfM) and object recognition tend to fail on scenes with highly reflective objects because the reflections behave differently to the true geometry of the scene. Such image sequences may be treated as two layers superimposed over each other - the nonreflection scene source layer and the reflection layer. However, decomposing the two layers is a very challenging task as it is ill-posed and common methods rely on prior information. This work presents an automated technique for detecting reflective features with a comprehensive analysis of the intrinsic, spatial, and temporal properties of feature points. A support vector machine (SVM) is proposed to learn reflection feature points. Predicted reflection feature points are used as priors to guide the reflection layer separation. This gives more robust and reliable results than what is achieved by performing layer separation alone.","PeriodicalId":146695,"journal":{"name":"2014 International Conference on Digital Image Computing: Techniques and Applications (DICTA)","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125577859","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 : 2014-11-01DOI: 10.1109/DICTA.2014.7008086
F. Westling, Changming Sun, Dadong Wang
An approach is suggested for automating fish identification and measurement using stereo Baited Remote Underwater Video footage. Simple methods for identifying fish are not sufficient for measurement, since the snout and tail points must be found, and the stereo data should be incorporated to find a true measurement. We present a modular framework that ties together various approaches in order to develop a generalised system for automated fish detection. A method is also suggested for using machine learning to improve identification. Experimental results indicate the suitability of our approach.
{"title":"A Modular Learning Approach for Fish Counting and Measurement Using Stereo Baited Remote Underwater Video","authors":"F. Westling, Changming Sun, Dadong Wang","doi":"10.1109/DICTA.2014.7008086","DOIUrl":"https://doi.org/10.1109/DICTA.2014.7008086","url":null,"abstract":"An approach is suggested for automating fish identification and measurement using stereo Baited Remote Underwater Video footage. Simple methods for identifying fish are not sufficient for measurement, since the snout and tail points must be found, and the stereo data should be incorporated to find a true measurement. We present a modular framework that ties together various approaches in order to develop a generalised system for automated fish detection. A method is also suggested for using machine learning to improve identification. Experimental results indicate the suitability of our approach.","PeriodicalId":146695,"journal":{"name":"2014 International Conference on Digital Image Computing: Techniques and Applications (DICTA)","volume":"429 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122869135","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 : 2014-11-01DOI: 10.1109/DICTA.2014.7008123
S. A. A. Shah, Bennamoun, F. Boussaïd
Despite the advent and popularity of low-cost commercial sensors (e.g., Microsoft Kinect), research in 3D vision still primarily focuses on the development of advanced algorithms geared towards high resolution data. This paper presents a comparative performance evaluation of renowned state-of-the-art 3D local surface descriptors for the task of registration of both high and low resolution range image data. The datasets used in these experiments are the renowned high resolution Stanford 3D models dataset and challenging low resolution Washington RGB-D object dataset. Experimental results show that the performance of certain local surface descriptors is significantly affected by low resolution data.
{"title":"Performance Evaluation of 3D Local Surface Descriptors for Low and High Resolution Range Image Registration","authors":"S. A. A. Shah, Bennamoun, F. Boussaïd","doi":"10.1109/DICTA.2014.7008123","DOIUrl":"https://doi.org/10.1109/DICTA.2014.7008123","url":null,"abstract":"Despite the advent and popularity of low-cost commercial sensors (e.g., Microsoft Kinect), research in 3D vision still primarily focuses on the development of advanced algorithms geared towards high resolution data. This paper presents a comparative performance evaluation of renowned state-of-the-art 3D local surface descriptors for the task of registration of both high and low resolution range image data. The datasets used in these experiments are the renowned high resolution Stanford 3D models dataset and challenging low resolution Washington RGB-D object dataset. Experimental results show that the performance of certain local surface descriptors is significantly affected by low resolution data.","PeriodicalId":146695,"journal":{"name":"2014 International Conference on Digital Image Computing: Techniques and Applications (DICTA)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130117862","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 : 2014-11-01DOI: 10.1109/DICTA.2014.7008120
Sheikh Md. Rabiul Islam, Xu Huang, K. Le
Indexes used for image quality evaluation are provided as computational models to measure the quality of images in a perceptually consistent manner. This paper presents a novel evaluation index for assessing image qualities. The index is a modification of the existing traditional Structural Similarity Index Measure (SSIM) by adding another factor to reflect the shape of the brightness histogram of the assessed image. The proposed index therefore is a combination of four major factors luminance, contrast, structure and shape of histogram. This index is mathematically simple and applicable in various image processing. For demonstration a new image de-noising approach using an adaptive shrinkage threshold in the shearlet domain is used. Experimental results show that the new image quality indexes give better prediction accuracy, better prediction monotonicity than PSNR, HQI, UIQI and SSIM.
{"title":"Novel Evaluation Index for Image Quality","authors":"Sheikh Md. Rabiul Islam, Xu Huang, K. Le","doi":"10.1109/DICTA.2014.7008120","DOIUrl":"https://doi.org/10.1109/DICTA.2014.7008120","url":null,"abstract":"Indexes used for image quality evaluation are provided as computational models to measure the quality of images in a perceptually consistent manner. This paper presents a novel evaluation index for assessing image qualities. The index is a modification of the existing traditional Structural Similarity Index Measure (SSIM) by adding another factor to reflect the shape of the brightness histogram of the assessed image. The proposed index therefore is a combination of four major factors luminance, contrast, structure and shape of histogram. This index is mathematically simple and applicable in various image processing. For demonstration a new image de-noising approach using an adaptive shrinkage threshold in the shearlet domain is used. Experimental results show that the new image quality indexes give better prediction accuracy, better prediction monotonicity than PSNR, HQI, UIQI and SSIM.","PeriodicalId":146695,"journal":{"name":"2014 International Conference on Digital Image Computing: Techniques and Applications (DICTA)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134243159","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 : 2014-11-01DOI: 10.1109/DICTA.2014.7008122
M. Le, S. L. Phung, A. Bouzerdoum
Automatically finding paths is a crucial and challenging task in autonomous navigation systems. The task becomes more difficult in unstructured environments such as indoor or outdoor scenes with unmarked pedestrian lanes under severe illumination conditions, complex lane surface structures, and occlusion. This paper proposes a robust method for pedestrian lane detection in such unstructured environments. The proposed method detects the walking lane in a probabilistic framework integrating both appearance of the lane region and characteristics of the lane borders. The vanishing point is employed to identify the lane borders. We propose an improved vanishing point estimation method based on orientation of color edges, and use pedestrian detection for occlusion handling. The proposed pedestrian lane detection method is evaluated on a new data set of 2000 images collected from various indoor and outdoor scenes with different types of unmarked lanes. Experimental results and comparisons with other existing methods on the new data set have shown the efficiency and robustness of the proposed method.
{"title":"Pedestrian Lane Detection in Unstructured Environments for Assistive Navigation","authors":"M. Le, S. L. Phung, A. Bouzerdoum","doi":"10.1109/DICTA.2014.7008122","DOIUrl":"https://doi.org/10.1109/DICTA.2014.7008122","url":null,"abstract":"Automatically finding paths is a crucial and challenging task in autonomous navigation systems. The task becomes more difficult in unstructured environments such as indoor or outdoor scenes with unmarked pedestrian lanes under severe illumination conditions, complex lane surface structures, and occlusion. This paper proposes a robust method for pedestrian lane detection in such unstructured environments. The proposed method detects the walking lane in a probabilistic framework integrating both appearance of the lane region and characteristics of the lane borders. The vanishing point is employed to identify the lane borders. We propose an improved vanishing point estimation method based on orientation of color edges, and use pedestrian detection for occlusion handling. The proposed pedestrian lane detection method is evaluated on a new data set of 2000 images collected from various indoor and outdoor scenes with different types of unmarked lanes. Experimental results and comparisons with other existing methods on the new data set have shown the efficiency and robustness of the proposed method.","PeriodicalId":146695,"journal":{"name":"2014 International Conference on Digital Image Computing: Techniques and Applications (DICTA)","volume":"115 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123987905","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 : 2014-11-01DOI: 10.1109/DICTA.2014.7008133
Abass A. Olaode, G. Naghdy, Catherine A. Todd
Image annotation has been identified to be a suitable means by which the semantic gap which has made the accuracy of Content-based image retrieval unsatisfactory be eliminated. However existing methods of automatic annotation of images depends on supervised learning, which can be difficult to implement due to the need for manually annotated training samples which are not always readily available. This paper argues that the unsupervised learning via Probabilistic Latent Semantic Analysis provides a more suitable machine learning approach for image annotation especially due to its potential to based categorisation on the latent semantic content of the image samples, which can bridge the semantic gap present in Content Based Image Retrieval. This paper therefore proposes an unsupervised image categorisation model in which the semantic content of images are discovered using Probabilistic Latent Semantic Analysis, after which they are clustered into unique groups based on semantic content similarities using K-means algorithm, thereby providing suitable annotation exemplars. A common problem with categorisation algorithms based on Bag-of-Visual Words modelling is the loss of accuracy due to spatial incoherency of the Bag-of-Visual Word modelling, this paper also examines the effectiveness of Spatial pyramid as a means of eliminating spatial incoherency in Probabilistic Latent Semantic Analysis classification.
{"title":"Unsupervised Image Classification by Probabilistic Latent Semantic Analysis for the Annotation of Images","authors":"Abass A. Olaode, G. Naghdy, Catherine A. Todd","doi":"10.1109/DICTA.2014.7008133","DOIUrl":"https://doi.org/10.1109/DICTA.2014.7008133","url":null,"abstract":"Image annotation has been identified to be a suitable means by which the semantic gap which has made the accuracy of Content-based image retrieval unsatisfactory be eliminated. However existing methods of automatic annotation of images depends on supervised learning, which can be difficult to implement due to the need for manually annotated training samples which are not always readily available. This paper argues that the unsupervised learning via Probabilistic Latent Semantic Analysis provides a more suitable machine learning approach for image annotation especially due to its potential to based categorisation on the latent semantic content of the image samples, which can bridge the semantic gap present in Content Based Image Retrieval. This paper therefore proposes an unsupervised image categorisation model in which the semantic content of images are discovered using Probabilistic Latent Semantic Analysis, after which they are clustered into unique groups based on semantic content similarities using K-means algorithm, thereby providing suitable annotation exemplars. A common problem with categorisation algorithms based on Bag-of-Visual Words modelling is the loss of accuracy due to spatial incoherency of the Bag-of-Visual Word modelling, this paper also examines the effectiveness of Spatial pyramid as a means of eliminating spatial incoherency in Probabilistic Latent Semantic Analysis classification.","PeriodicalId":146695,"journal":{"name":"2014 International Conference on Digital Image Computing: Techniques and Applications (DICTA)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116015302","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 : 2014-11-01DOI: 10.1109/DICTA.2014.7008128
M. Uzair, A. Mian
Image set based face recognition provides more opportunities compared to single mug-shot face recognition. However, modelling the variations in an image set is a challenging task. We propose a computationally efficient and accurate image set modelling technique. The idea is to reconstruct each image set sample with an unlabeled dictionary using the computationally efficient regularized least squares. The reconstruction coefficients form a latent representation of an image set and efficiently model its underlying structure. We propose max and sum pooling to aggregate the latent representations into a single compact feature vector representation per set. We then perform Linear Discriminant Analysis on the pooled reconstruction coefficients to increase the discrimination and reduce the dimensionality of the proposed features. The proposed algorithm is extensively evaluated for the task of image set based face recognition on the Honda/UCSD, CMU Mobo and YouTube celebrities datasets. Experimental results show that the proposed algorithm outperforms current state-of-the-art image set classification algorithms in terms of both accuracy and execution time.
{"title":"Regularized Least-Squares Coding with Unlabeled Dictionary for Image-Set Based Face Recognition","authors":"M. Uzair, A. Mian","doi":"10.1109/DICTA.2014.7008128","DOIUrl":"https://doi.org/10.1109/DICTA.2014.7008128","url":null,"abstract":"Image set based face recognition provides more opportunities compared to single mug-shot face recognition. However, modelling the variations in an image set is a challenging task. We propose a computationally efficient and accurate image set modelling technique. The idea is to reconstruct each image set sample with an unlabeled dictionary using the computationally efficient regularized least squares. The reconstruction coefficients form a latent representation of an image set and efficiently model its underlying structure. We propose max and sum pooling to aggregate the latent representations into a single compact feature vector representation per set. We then perform Linear Discriminant Analysis on the pooled reconstruction coefficients to increase the discrimination and reduce the dimensionality of the proposed features. The proposed algorithm is extensively evaluated for the task of image set based face recognition on the Honda/UCSD, CMU Mobo and YouTube celebrities datasets. Experimental results show that the proposed algorithm outperforms current state-of-the-art image set classification algorithms in terms of both accuracy and execution time.","PeriodicalId":146695,"journal":{"name":"2014 International Conference on Digital Image Computing: Techniques and Applications (DICTA)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116287696","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 : 2014-11-01DOI: 10.1109/DICTA.2014.7008126
Yuchao Dai, Hongdong Li
Low rank method or rank-minimization has received considerable attention from recent computer vision community. Due to the inherent computational complexity of rank problems, the non-convex rank function is often relaxed to its convex relaxation, i.e. the nuclear norm. Thanks to recent progress made in the filed of compressive sensing (CS), vision researchers who are practicing CS are fully aware, and conscious, of the convex relaxation gap, as well as under which condition (e.g. Restricted Isometry Property) the relaxation is tight (i.e. with nil gap). In this paper, we however wish to alert the potential users of the low-rank method that: focusing too much on the issue of relaxation gap and optimization may possibly adversely obscure the "big picture'' of the original vision problem. In particular, this paper shows that for many commonly cited low-rank problems, nuclear norm minimization formulation of the original rank-minimization problem do not necessarily lead to the desired solution. Degenerate solutions and multiplicity seem often or always exist. Even if a certain nuclear-norm minimization solution is a provably tight relaxation, this solution can possibly be meaningless in its particular context. We therefore advocate that, in solving vision problems via nuclear norm minimization, special care must be given, and domain-dependent prior knowledge must be taken into account. This paper summarizes recent relevant theoretical results, provides original analysis, uses real examples to demonstrate the practical implications.
{"title":"Rank Minimization or Nuclear-Norm Minimization: Are We Solving the Right Problem?","authors":"Yuchao Dai, Hongdong Li","doi":"10.1109/DICTA.2014.7008126","DOIUrl":"https://doi.org/10.1109/DICTA.2014.7008126","url":null,"abstract":"Low rank method or rank-minimization has received considerable attention from recent computer vision community. Due to the inherent computational complexity of rank problems, the non-convex rank function is often relaxed to its convex relaxation, i.e. the nuclear norm. Thanks to recent progress made in the filed of compressive sensing (CS), vision researchers who are practicing CS are fully aware, and conscious, of the convex relaxation gap, as well as under which condition (e.g. Restricted Isometry Property) the relaxation is tight (i.e. with nil gap). In this paper, we however wish to alert the potential users of the low-rank method that: focusing too much on the issue of relaxation gap and optimization may possibly adversely obscure the \"big picture'' of the original vision problem. In particular, this paper shows that for many commonly cited low-rank problems, nuclear norm minimization formulation of the original rank-minimization problem do not necessarily lead to the desired solution. Degenerate solutions and multiplicity seem often or always exist. Even if a certain nuclear-norm minimization solution is a provably tight relaxation, this solution can possibly be meaningless in its particular context. We therefore advocate that, in solving vision problems via nuclear norm minimization, special care must be given, and domain-dependent prior knowledge must be taken into account. This paper summarizes recent relevant theoretical results, provides original analysis, uses real examples to demonstrate the practical implications.","PeriodicalId":146695,"journal":{"name":"2014 International Conference on Digital Image Computing: Techniques and Applications (DICTA)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128281582","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：