Data-driven techniques can reliably build semantic correspondence among images. In this paper, we present a new regularization model for stereo or flow through transferring the shape information of the disparity or flow from semantically matched patches in the training database. Compared to previous regularization models based on image appearance alone, we can better resolve local ambiguity of the disparity or flow by considering the semantic information without explicit object modeling. We incorporate this data-driven regularization model into a standard Markov Random Field (MRF) model, inferred with a gradient descent algorithm and learned with a discriminative learning approach. Compared to prior state-of-the-art methods, our full model achieves comparable or better results on the KITTI stereo and flow datasets, and improves results on the Sintel Flow dataset under an online estimation setting.
{"title":"A Data-Driven Regularization Model for Stereo and Flow","authors":"D. Wei, Ce Liu, W. Freeman","doi":"10.1109/3DV.2014.97","DOIUrl":"https://doi.org/10.1109/3DV.2014.97","url":null,"abstract":"Data-driven techniques can reliably build semantic correspondence among images. In this paper, we present a new regularization model for stereo or flow through transferring the shape information of the disparity or flow from semantically matched patches in the training database. Compared to previous regularization models based on image appearance alone, we can better resolve local ambiguity of the disparity or flow by considering the semantic information without explicit object modeling. We incorporate this data-driven regularization model into a standard Markov Random Field (MRF) model, inferred with a gradient descent algorithm and learned with a discriminative learning approach. Compared to prior state-of-the-art methods, our full model achieves comparable or better results on the KITTI stereo and flow datasets, and improves results on the Sintel Flow dataset under an online estimation setting.","PeriodicalId":275516,"journal":{"name":"2014 2nd International Conference on 3D Vision","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115740362","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}
Unprecedented amounts of 3D data can be acquired in urban environments, but their use for scene understanding is challenging due to varying data resolution and variability of objects in the same class. An additional challenge is due to the nature of the point clouds themselves, since they lack detailed geometric or semantic information that would aid scene understanding. In this paper we present a general algorithm for segmenting and jointly classifying object parts and the object itself. Our pipeline consists of local feature extraction, robust RANSAC part segmentation, part-level feature extraction, a structured model for parts in objects, and classification using state-of-the-art classifiers. We have tested this pipeline in a very challenging dataset that consists of real world scans of vehicles. Our contributions include the development of a segmentation and classification pipeline for objects and their parts, and a method for segmentation that is robust to the complexity of unstructured 3D points clouds, as well as a part ordering strategy for the sequential structured model and a joint feature representation between object parts.
{"title":"Classification of Vehicle Parts in Unstructured 3D Point Clouds","authors":"Allan Zelener, Philippos Mordohai, I. Stamos","doi":"10.1109/3DV.2014.58","DOIUrl":"https://doi.org/10.1109/3DV.2014.58","url":null,"abstract":"Unprecedented amounts of 3D data can be acquired in urban environments, but their use for scene understanding is challenging due to varying data resolution and variability of objects in the same class. An additional challenge is due to the nature of the point clouds themselves, since they lack detailed geometric or semantic information that would aid scene understanding. In this paper we present a general algorithm for segmenting and jointly classifying object parts and the object itself. Our pipeline consists of local feature extraction, robust RANSAC part segmentation, part-level feature extraction, a structured model for parts in objects, and classification using state-of-the-art classifiers. We have tested this pipeline in a very challenging dataset that consists of real world scans of vehicles. Our contributions include the development of a segmentation and classification pipeline for objects and their parts, and a method for segmentation that is robust to the complexity of unstructured 3D points clouds, as well as a part ordering strategy for the sequential structured model and a joint feature representation between object parts.","PeriodicalId":275516,"journal":{"name":"2014 2nd International Conference on 3D Vision","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125423148","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}
Weifeng Chen, Guofeng Zhang, Xiaojun Xiang, Jiaya Jia, H. Bao
Although multi-view stereo has been extensively studied during the past decades, automatically computing high-quality dense depth information from captured images/videos is still quite difficult. Many factors, such as serious occlusion, large texture less regions and strong reflection, easily cause erroneous depth recovery. In this paper, we present a novel semi-automatic multi-view stereo system, which can quickly create and repair depth from a monocular sequence taken by a freely moving camera. One of our main contributions is that we propose a novel multi-view stereo model incorporating prior constraints indicated by user interaction, which makes it possible to even handle Non-Lambertian surface that surely violates the photo-consistency constraint. Users only need to provide a coarse segmentation and a few user interactions, our system can automatically correct depth and refine boundary. With other priors and occlusion handling, the erroneous depth can be effectively corrected even for very challenging examples that are difficult for state-of-the-art methods.
{"title":"High-Quality Depth Recovery via Interactive Multi-view Stereo","authors":"Weifeng Chen, Guofeng Zhang, Xiaojun Xiang, Jiaya Jia, H. Bao","doi":"10.1109/3DV.2014.55","DOIUrl":"https://doi.org/10.1109/3DV.2014.55","url":null,"abstract":"Although multi-view stereo has been extensively studied during the past decades, automatically computing high-quality dense depth information from captured images/videos is still quite difficult. Many factors, such as serious occlusion, large texture less regions and strong reflection, easily cause erroneous depth recovery. In this paper, we present a novel semi-automatic multi-view stereo system, which can quickly create and repair depth from a monocular sequence taken by a freely moving camera. One of our main contributions is that we propose a novel multi-view stereo model incorporating prior constraints indicated by user interaction, which makes it possible to even handle Non-Lambertian surface that surely violates the photo-consistency constraint. Users only need to provide a coarse segmentation and a few user interactions, our system can automatically correct depth and refine boundary. With other priors and occlusion handling, the erroneous depth can be effectively corrected even for very challenging examples that are difficult for state-of-the-art methods.","PeriodicalId":275516,"journal":{"name":"2014 2nd International Conference on 3D Vision","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114393387","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}
This paper introduces a method for automatically transforming a point cloud from a laser scanner into a volumetric 3D building model based on the new concept of enclosure reasoning. Rather than simply classifying and modeling building surfaces independently or with pair wise contextual relationships, this work introduces room, floor and building level reasoning. Enclosure reasoning premises that rooms are cycles of walls enclosing free interior space. These cycles should be of minimum description length (MDL) and obey the statistical priors expected for rooms. Floors and buildings then contain the best coverage of the mostly likely rooms. This allows the pipeline to generate higher fidelity models by performing modeling and recognition jointly over the entire building at once. The complete pipeline takes raw, registered laser scan surveys of a single building. It extracts the most likely smooth architectural surfaces, locates the building, and generates wall hypotheses. The algorithm then optimizes the model by growing, merging, and pruning these hypotheses to generate the most likely rooms, floors, and building in the presence of significant clutter.
{"title":"Building Modeling through Enclosure Reasoning","authors":"Adam Stambler, Daniel F. Huber","doi":"10.1109/3DV.2014.65","DOIUrl":"https://doi.org/10.1109/3DV.2014.65","url":null,"abstract":"This paper introduces a method for automatically transforming a point cloud from a laser scanner into a volumetric 3D building model based on the new concept of enclosure reasoning. Rather than simply classifying and modeling building surfaces independently or with pair wise contextual relationships, this work introduces room, floor and building level reasoning. Enclosure reasoning premises that rooms are cycles of walls enclosing free interior space. These cycles should be of minimum description length (MDL) and obey the statistical priors expected for rooms. Floors and buildings then contain the best coverage of the mostly likely rooms. This allows the pipeline to generate higher fidelity models by performing modeling and recognition jointly over the entire building at once. The complete pipeline takes raw, registered laser scan surveys of a single building. It extracts the most likely smooth architectural surfaces, locates the building, and generates wall hypotheses. The algorithm then optimizes the model by growing, merging, and pruning these hypotheses to generate the most likely rooms, floors, and building in the presence of significant clutter.","PeriodicalId":275516,"journal":{"name":"2014 2nd International Conference on 3D Vision","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129413942","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}
This paper is aimed at proposing a new high quality free-viewpoint rendering algorithm of 3D video. The main challenge on visualizing 3D video is how to utilize the original multi-view images used to estimate the 3D surface, and how to manage the mismatches between them due to calibration and reconstruction errors. The key idea to solve this problem is to optimize the 3D shape on a per-viewpoint basis on the fly. Given a virtual viewpoint for visualization, our algorithm optimizes the 3D shape so as to maximize the photo-consistency over the surface visible from the virtual viewpoint. An evaluation demonstrates that our method outperforms the state-of-the-art rendering qualitatively and quantitatively.
{"title":"A Real-Time View-Dependent Shape Optimization for High Quality Free-Viewpoint Rendering of 3D Video","authors":"S. Nobuhara, Wei Ning, T. Matsuyama","doi":"10.1109/3DV.2014.28","DOIUrl":"https://doi.org/10.1109/3DV.2014.28","url":null,"abstract":"This paper is aimed at proposing a new high quality free-viewpoint rendering algorithm of 3D video. The main challenge on visualizing 3D video is how to utilize the original multi-view images used to estimate the 3D surface, and how to manage the mismatches between them due to calibration and reconstruction errors. The key idea to solve this problem is to optimize the 3D shape on a per-viewpoint basis on the fly. Given a virtual viewpoint for visualization, our algorithm optimizes the 3D shape so as to maximize the photo-consistency over the surface visible from the virtual viewpoint. An evaluation demonstrates that our method outperforms the state-of-the-art rendering qualitatively and quantitatively.","PeriodicalId":275516,"journal":{"name":"2014 2nd International Conference on 3D Vision","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124577817","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}
Reasonable computation and accurate camera calibration require matching many interest points over long baselines. This is a difficult problem requiring better solutions than presently exist for urban scenes involving large buildings containing many windows since windows in a facade all have the same texture and, therefore, cannot be distinguished from one another based solely on appearance. Hence, the usual approach to feature detection and matching, such as use of SIFT, does not work in these scenes. A novel algorithm is introduced to provide correspondences for multiple repeating feature patterns seen under significant viewpoint changes. Most existing appearance-based algorithms cannot handle highly repetitive textures due to the match location ambiguity. However, the target structure provides a rich set of repeating features to be matched and tracked across multiple views, thus potentially improving camera estimation accuracy. The proposed method also exploits the geometric structure of regular grids of repeating features on planar surfaces.
{"title":"Matching Many Identical Features of Planar Urban Facades Using Global Regularity","authors":"Eduardo B. Almeida, D. Cooper","doi":"10.1109/3DV.2014.107","DOIUrl":"https://doi.org/10.1109/3DV.2014.107","url":null,"abstract":"Reasonable computation and accurate camera calibration require matching many interest points over long baselines. This is a difficult problem requiring better solutions than presently exist for urban scenes involving large buildings containing many windows since windows in a facade all have the same texture and, therefore, cannot be distinguished from one another based solely on appearance. Hence, the usual approach to feature detection and matching, such as use of SIFT, does not work in these scenes. A novel algorithm is introduced to provide correspondences for multiple repeating feature patterns seen under significant viewpoint changes. Most existing appearance-based algorithms cannot handle highly repetitive textures due to the match location ambiguity. However, the target structure provides a rich set of repeating features to be matched and tracked across multiple views, thus potentially improving camera estimation accuracy. The proposed method also exploits the geometric structure of regular grids of repeating features on planar surfaces.","PeriodicalId":275516,"journal":{"name":"2014 2nd International Conference on 3D Vision","volume":"112 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130377306","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}
Kent Fujiwara, Hiroshi Kawasaki, R. Sagawa, K. Ogawara, K. Ikeuchi
We propose a non-rigid registration method for completion of dynamic shapes with occlusion. Our method is based on the idea that an occluded region in a certain frame should be visible in another frame and that local regions should be moving rigidly when the motion is small. We achieve this with a novel reliable distance field (DF) for non-rigid registration with missing regions. We first fit a pseudo-surface onto the input shape using a surface reconstruction method. We then calculate the difference between the DF of the input shape and the pseudo-surface. We define the areas with large difference as unreliable, as these areas indicate that the original shape cannot be found nearby. We then conduct non-rigid registration using local rigid transformations to match the source and target data at visible regions and maintain the original shape as much as possible in occluded regions. The experimental results demonstrate that our method is capable of accurately filling in the missing regions using the shape information from prior or posterior frames. By sequentially processing the data, our method is also capable of completing an entire sequence with missing regions.
{"title":"Non-rigid registration with reliable distance field for dynamic shape completion","authors":"Kent Fujiwara, Hiroshi Kawasaki, R. Sagawa, K. Ogawara, K. Ikeuchi","doi":"10.1109/3DV.2014.111","DOIUrl":"https://doi.org/10.1109/3DV.2014.111","url":null,"abstract":"We propose a non-rigid registration method for completion of dynamic shapes with occlusion. Our method is based on the idea that an occluded region in a certain frame should be visible in another frame and that local regions should be moving rigidly when the motion is small. We achieve this with a novel reliable distance field (DF) for non-rigid registration with missing regions. We first fit a pseudo-surface onto the input shape using a surface reconstruction method. We then calculate the difference between the DF of the input shape and the pseudo-surface. We define the areas with large difference as unreliable, as these areas indicate that the original shape cannot be found nearby. We then conduct non-rigid registration using local rigid transformations to match the source and target data at visible regions and maintain the original shape as much as possible in occluded regions. The experimental results demonstrate that our method is capable of accurately filling in the missing regions using the shape information from prior or posterior frames. By sequentially processing the data, our method is also capable of completing an entire sequence with missing regions.","PeriodicalId":275516,"journal":{"name":"2014 2nd International Conference on 3D Vision","volume":"140 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121158327","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}
Traditional Structure-from-Motion (SfM) approaches work well for richly textured scenes with a high number of distinctive feature points. Since man-made environments often contain texture less objects, the resulting point cloud suffers from a low density in corresponding scene parts. The missing 3D information heavily affects all kinds of subsequent post-processing tasks (e.g. Meshing), and significantly decreases the visual appearance of the resulting 3D model. We propose a novel 3D reconstruction approach, which uses the output of conventional SfM pipelines to generate additional complementary 3D information, by exploiting line segments. We use appearance-less epipolar guided line matching to create a potentially large set of 3D line hypotheses, which are then verified using a global graph clustering procedure. We show that our proposed method outperforms the current state-of-the-art in terms of runtime and accuracy, as well as visual appearance of the resulting reconstructions.
{"title":"Improving Sparse 3D Models for Man-Made Environments Using Line-Based 3D Reconstruction","authors":"Manuel Hofer, Michael Maurer, H. Bischof","doi":"10.1109/3DV.2014.14","DOIUrl":"https://doi.org/10.1109/3DV.2014.14","url":null,"abstract":"Traditional Structure-from-Motion (SfM) approaches work well for richly textured scenes with a high number of distinctive feature points. Since man-made environments often contain texture less objects, the resulting point cloud suffers from a low density in corresponding scene parts. The missing 3D information heavily affects all kinds of subsequent post-processing tasks (e.g. Meshing), and significantly decreases the visual appearance of the resulting 3D model. We propose a novel 3D reconstruction approach, which uses the output of conventional SfM pipelines to generate additional complementary 3D information, by exploiting line segments. We use appearance-less epipolar guided line matching to create a potentially large set of 3D line hypotheses, which are then verified using a global graph clustering procedure. We show that our proposed method outperforms the current state-of-the-art in terms of runtime and accuracy, as well as visual appearance of the resulting reconstructions.","PeriodicalId":275516,"journal":{"name":"2014 2nd International Conference on 3D Vision","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114692449","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}
Adrián Peñate Sánchez, F. Moreno-Noguer, J. Andrade-Cetto, F. Fleuret
We introduce LETHA (Learning on Easy data, Test on Hard), a new learning paradigm consisting of building strong priors from high quality training data, and combining them with discriminative machine learning to deal with low-quality test data. Our main contribution is an implementation of that concept for pose estimation. We first automatically build a 3D model of the object of interest from high-definition images, and devise from it a pose-indexed feature extraction scheme. We then train a single classifier to process these feature vectors. Given a low quality test image, we visit many hypothetical poses, extract features consistently and evaluate the response of the classifier. Since this process uses locations recorded during learning, it does not require matching points anymore. We use a boosting procedure to train this classifier common to all poses, which is able to deal with missing features, due in this context to self-occlusion. Our results demonstrate that the method combines the strengths of global image representations, discriminative even for very tiny images, and the robustness to occlusions of approaches based on local feature point descriptors.
我们介绍了LETHA (Learning on Easy data, Test on Hard)，这是一种新的学习范式，它包括从高质量的训练数据中构建强先验，并将它们与判别机器学习相结合，以处理低质量的测试数据。我们的主要贡献是实现了姿态估计的概念。我们首先从高清图像中自动建立感兴趣对象的三维模型，并设计了一个基于姿态索引的特征提取方案。然后我们训练一个分类器来处理这些特征向量。给定低质量的测试图像，我们访问许多假设的姿势，一致地提取特征并评估分类器的响应。由于这个过程使用的是在学习过程中记录的位置，因此不再需要匹配点。我们使用一个增强过程来训练这个对所有姿势都通用的分类器，它能够处理由于这种情况下的自遮挡而缺失的特征。我们的研究结果表明，该方法结合了全局图像表示的优势，即使对非常微小的图像也具有判别性，以及基于局部特征点描述符的方法对遮挡的鲁棒性。
{"title":"LETHA: Learning from High Quality Inputs for 3D Pose Estimation in Low Quality Images","authors":"Adrián Peñate Sánchez, F. Moreno-Noguer, J. Andrade-Cetto, F. Fleuret","doi":"10.1109/3DV.2014.18","DOIUrl":"https://doi.org/10.1109/3DV.2014.18","url":null,"abstract":"We introduce LETHA (Learning on Easy data, Test on Hard), a new learning paradigm consisting of building strong priors from high quality training data, and combining them with discriminative machine learning to deal with low-quality test data. Our main contribution is an implementation of that concept for pose estimation. We first automatically build a 3D model of the object of interest from high-definition images, and devise from it a pose-indexed feature extraction scheme. We then train a single classifier to process these feature vectors. Given a low quality test image, we visit many hypothetical poses, extract features consistently and evaluate the response of the classifier. Since this process uses locations recorded during learning, it does not require matching points anymore. We use a boosting procedure to train this classifier common to all poses, which is able to deal with missing features, due in this context to self-occlusion. Our results demonstrate that the method combines the strengths of global image representations, discriminative even for very tiny images, and the robustness to occlusions of approaches based on local feature point descriptors.","PeriodicalId":275516,"journal":{"name":"2014 2nd International Conference on 3D Vision","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114833464","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}
Key point matching in images of indoor scenes traditionally employs features like SIFT, GIST and HOG. While those features work very well for two images related to each other by small camera transformations, we commonly observe a drop in performance for patches representing scene elements visualized from a very different perspective. Since increasing the space of considered local transformations for feature matching decreases their discriminative abilities, we propose a more global approach inspired by the recent success of monocular scene understanding. In particular we propose to reconstruct a box-like model of the scene from every single image and use it to rectify images before matching. We show that a monocular scene model reconstruction and rectification preceding standard feature matching significantly improves key point matching and dramatically improves reconstruction of difficult indoor scenes.
{"title":"Match Box: Indoor Image Matching via Box-Like Scene Estimation","authors":"F. Srajer, A. Schwing, M. Pollefeys, T. Pajdla","doi":"10.1109/3DV.2014.56","DOIUrl":"https://doi.org/10.1109/3DV.2014.56","url":null,"abstract":"Key point matching in images of indoor scenes traditionally employs features like SIFT, GIST and HOG. While those features work very well for two images related to each other by small camera transformations, we commonly observe a drop in performance for patches representing scene elements visualized from a very different perspective. Since increasing the space of considered local transformations for feature matching decreases their discriminative abilities, we propose a more global approach inspired by the recent success of monocular scene understanding. In particular we propose to reconstruct a box-like model of the scene from every single image and use it to rectify images before matching. We show that a monocular scene model reconstruction and rectification preceding standard feature matching significantly improves key point matching and dramatically improves reconstruction of difficult indoor scenes.","PeriodicalId":275516,"journal":{"name":"2014 2nd International Conference on 3D Vision","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115869832","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：