Pub Date : 2018-10-01DOI: 10.1109/SIBGRAPI.2018.00061
A. L. L. Maia, Frank D. Julca-Aguilar, N. Hirata
We propose a new approach for segmenting a document image into its page components (e.g. text, graphics and tables). Our approach consists of two main steps. In the first step, a set of scores corresponding to the output of a convolutional neural network, one for each of the possible page component categories, is assigned to each connected component in the document. The labeled connected components define a fuzzy over-segmentation of the page. In the second step, spatially close connected components that are likely to belong to a same page component are grouped together. This is done by building an attributed region adjacency graph of the connected components and modeling the problem as an edge removal problem. Edges are then kept or removed based on a pre-trained classifier. The resulting groups, defined by the connected subgraphs, correspond to the detected page components. We evaluate our method on the ICDAR2009 dataset. Results show that our method effectively segments pages, being able to detect the nine types of page components. Furthermore, as our approach is based on simple machine learning models and graph-based techniques, it should be easily adapted to the segmentation of a variety of document types.
{"title":"A Machine Learning Approach for Graph-Based Page Segmentation","authors":"A. L. L. Maia, Frank D. Julca-Aguilar, N. Hirata","doi":"10.1109/SIBGRAPI.2018.00061","DOIUrl":"https://doi.org/10.1109/SIBGRAPI.2018.00061","url":null,"abstract":"We propose a new approach for segmenting a document image into its page components (e.g. text, graphics and tables). Our approach consists of two main steps. In the first step, a set of scores corresponding to the output of a convolutional neural network, one for each of the possible page component categories, is assigned to each connected component in the document. The labeled connected components define a fuzzy over-segmentation of the page. In the second step, spatially close connected components that are likely to belong to a same page component are grouped together. This is done by building an attributed region adjacency graph of the connected components and modeling the problem as an edge removal problem. Edges are then kept or removed based on a pre-trained classifier. The resulting groups, defined by the connected subgraphs, correspond to the detected page components. We evaluate our method on the ICDAR2009 dataset. Results show that our method effectively segments pages, being able to detect the nine types of page components. Furthermore, as our approach is based on simple machine learning models and graph-based techniques, it should be easily adapted to the segmentation of a variety of document types.","PeriodicalId":208985,"journal":{"name":"2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)","volume":"85 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133582686","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 : 2018-10-01DOI: 10.1109/SIBGRAPI.2018.00026
Rodrigo Nunes Moni da Silva, A. Spritzer, C. Freitas
In this paper, we propose a web-based application where the user can instantiate multiple, coordinated panels for exploring data concerning the votes of representatives in Brazil's lower legislative house (the Chamber of Deputies). Open data about roll calls made available by the Chamber allowed us to build a set of interactive visualizations to let users explore deputies' votes and build an understanding of their political profiles. Based on the set of roll call voting results from 1991 to 2016, our application displays the political behaviour of parties in a timeline from which users can select periods and instantiate panels showing the political spectrum of deputies using different methods of dimensionality reduction. Deputies can be separated in clusters based on their position in the political spectrum, and other panels can be instantiated showing details about each cluster. Users can select parts of the timeline and simultaneously analyze the behavior of parties and one or more deputies. Roll calls are represented as a combination of heatmaps and histograms. We illustrate the use of the different visualization techniques in a case study on party cohesiveness over time.
{"title":"Visualization of Roll Call Data for Supporting Analyses of Political Profiles","authors":"Rodrigo Nunes Moni da Silva, A. Spritzer, C. Freitas","doi":"10.1109/SIBGRAPI.2018.00026","DOIUrl":"https://doi.org/10.1109/SIBGRAPI.2018.00026","url":null,"abstract":"In this paper, we propose a web-based application where the user can instantiate multiple, coordinated panels for exploring data concerning the votes of representatives in Brazil's lower legislative house (the Chamber of Deputies). Open data about roll calls made available by the Chamber allowed us to build a set of interactive visualizations to let users explore deputies' votes and build an understanding of their political profiles. Based on the set of roll call voting results from 1991 to 2016, our application displays the political behaviour of parties in a timeline from which users can select periods and instantiate panels showing the political spectrum of deputies using different methods of dimensionality reduction. Deputies can be separated in clusters based on their position in the political spectrum, and other panels can be instantiated showing details about each cluster. Users can select parts of the timeline and simultaneously analyze the behavior of parties and one or more deputies. Roll calls are represented as a combination of heatmaps and histograms. We illustrate the use of the different visualization techniques in a case study on party cohesiveness over time.","PeriodicalId":208985,"journal":{"name":"2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132994639","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 : 2018-10-01DOI: 10.1109/SIBGRAPI.2018.00033
H. Oliveira, J. A. D. Santos
Segmentation of anatomical structures in Chest Posterior-Anterior Radiographs is a classical task on biomedical image analysis. Deep Learning has been widely used for detection and diagnosis of illnesses in several medical image modalities over the last years, but the portability of deep methods is still limited, hampering the reusability of pre-trained models in new data. We address this problem by proposing a novel method for Cross-Dataset Transfer Learning in Chest X-Ray images based on Unsupervised Image Translation architectures. Our Transfer Learning approach achieved Jaccard values of 88.20% on lung field segmentation in the Montgomery Set by using a pre-trained model on the JSRT dataset and no labeled data from the target dataset. Several experiments in unsupervised and semi-supervised transfer were performed and our method consistently outperformed simple fine-tuning when a limited amount of labels is used. Qualitative analysis on the tasks of clavicle and heart segmentation are also performed on Montgomery samples and pre-trained models from JSRT dataset. Our secondary contributions encompass several experiments in anatomical structure segmentation on JSRT, achieving state-of-the-art results in lung field (96.02%), heart (89.64%) and clavicle segmentation (87.30%).
{"title":"Deep Transfer Learning for Segmentation of Anatomical Structures in Chest Radiographs","authors":"H. Oliveira, J. A. D. Santos","doi":"10.1109/SIBGRAPI.2018.00033","DOIUrl":"https://doi.org/10.1109/SIBGRAPI.2018.00033","url":null,"abstract":"Segmentation of anatomical structures in Chest Posterior-Anterior Radiographs is a classical task on biomedical image analysis. Deep Learning has been widely used for detection and diagnosis of illnesses in several medical image modalities over the last years, but the portability of deep methods is still limited, hampering the reusability of pre-trained models in new data. We address this problem by proposing a novel method for Cross-Dataset Transfer Learning in Chest X-Ray images based on Unsupervised Image Translation architectures. Our Transfer Learning approach achieved Jaccard values of 88.20% on lung field segmentation in the Montgomery Set by using a pre-trained model on the JSRT dataset and no labeled data from the target dataset. Several experiments in unsupervised and semi-supervised transfer were performed and our method consistently outperformed simple fine-tuning when a limited amount of labels is used. Qualitative analysis on the tasks of clavicle and heart segmentation are also performed on Montgomery samples and pre-trained models from JSRT dataset. Our secondary contributions encompass several experiments in anatomical structure segmentation on JSRT, achieving state-of-the-art results in lung field (96.02%), heart (89.64%) and clavicle segmentation (87.30%).","PeriodicalId":208985,"journal":{"name":"2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124920366","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 : 2018-10-01DOI: 10.1109/SIBGRAPI.2018.00057
B. C. Benato, A. Telea, A. Falcão
While the number of unsupervised samples for data annotation is usually high, the absence of large supervised training sets for effective feature learning and design of high-quality classifiers is a known problem whenever specialists are required for data supervision. By exploring the feature space of supervised and unsupervised samples, semi-supervised learning approaches can usually improve the classification system. However, these approaches do not usually exploit the pattern-finding power of the user's visual system during machine learning. In this paper, we incorporate the user in the semi-supervised learning process by letting the feature space projection of unsupervised and supervised samples guide the label propagation actions of the user to the unsupervised samples. We show that this procedure can significantly reduce user effort while improving the quality of the classifier on unseen test sets. Due to the limited number of supervised samples, we also propose the use of auto-encoder neural networks for feature learning. For validation, we compare the classifiers that result from the proposed approach with the ones trained from the supervised samples only and semi-supervised trained using automatic label propagation.
{"title":"Semi-Supervised Learning with Interactive Label Propagation Guided by Feature Space Projections","authors":"B. C. Benato, A. Telea, A. Falcão","doi":"10.1109/SIBGRAPI.2018.00057","DOIUrl":"https://doi.org/10.1109/SIBGRAPI.2018.00057","url":null,"abstract":"While the number of unsupervised samples for data annotation is usually high, the absence of large supervised training sets for effective feature learning and design of high-quality classifiers is a known problem whenever specialists are required for data supervision. By exploring the feature space of supervised and unsupervised samples, semi-supervised learning approaches can usually improve the classification system. However, these approaches do not usually exploit the pattern-finding power of the user's visual system during machine learning. In this paper, we incorporate the user in the semi-supervised learning process by letting the feature space projection of unsupervised and supervised samples guide the label propagation actions of the user to the unsupervised samples. We show that this procedure can significantly reduce user effort while improving the quality of the classifier on unseen test sets. Due to the limited number of supervised samples, we also propose the use of auto-encoder neural networks for feature learning. For validation, we compare the classifiers that result from the proposed approach with the ones trained from the supervised samples only and semi-supervised trained using automatic label propagation.","PeriodicalId":208985,"journal":{"name":"2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)","volume":"77 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124684246","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 : 2018-10-01DOI: 10.1109/SIBGRAPI.2018.00062
B. Nassu, L. Lippmann, Bruno Marchesi, Amanda Canestraro, Rafael Wagner, Vanderlei Zarnicinski
Knowing the state of the disconnect switches in a power distribution substation is important to avoid accidents, damaged equipment, and service interruptions. This information is usually provided by human operators, who can commit errors because of the cluttered environment, bad weather or lighting conditions, or lack of attention. In this paper, we introduce an approach for determining the state of each switch in a substation, based on images captured by regular pan-tilt-zoom surveillance cameras. The proposed approach includes noise reduction, image registration using phase correlation, and classification using a convolutional neural network and a support vector machine fed with gradient-based descriptors. By combining information given in an initial labeling stage with image processing techniques to reduce variations in viewpoint, our approach achieved 100% accuracy on experiments performed at a real substation over multiple days. We also show how modifications to the standard phase correlation image registration algorithm can make it more robust to lighting variations, and how SIFT (Scale-Invariant Feature Transform) descriptors can be made more robust in scenarios where the relevant objects may be brighter or darker than the background.
{"title":"Image-Based State Recognition for Disconnect Switches in Electric Power Distribution Substations","authors":"B. Nassu, L. Lippmann, Bruno Marchesi, Amanda Canestraro, Rafael Wagner, Vanderlei Zarnicinski","doi":"10.1109/SIBGRAPI.2018.00062","DOIUrl":"https://doi.org/10.1109/SIBGRAPI.2018.00062","url":null,"abstract":"Knowing the state of the disconnect switches in a power distribution substation is important to avoid accidents, damaged equipment, and service interruptions. This information is usually provided by human operators, who can commit errors because of the cluttered environment, bad weather or lighting conditions, or lack of attention. In this paper, we introduce an approach for determining the state of each switch in a substation, based on images captured by regular pan-tilt-zoom surveillance cameras. The proposed approach includes noise reduction, image registration using phase correlation, and classification using a convolutional neural network and a support vector machine fed with gradient-based descriptors. By combining information given in an initial labeling stage with image processing techniques to reduce variations in viewpoint, our approach achieved 100% accuracy on experiments performed at a real substation over multiple days. We also show how modifications to the standard phase correlation image registration algorithm can make it more robust to lighting variations, and how SIFT (Scale-Invariant Feature Transform) descriptors can be made more robust in scenarios where the relevant objects may be brighter or darker than the background.","PeriodicalId":208985,"journal":{"name":"2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)","volume":"116 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117016106","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 : 2018-10-01DOI: 10.1109/SIBGRAPI.2018.00032
Daniel Souza, M. M. O. Neto
Bone age is a reliable metric for determining the level of biological maturity of children and adolescents. Its assessment is a crucial part of the diagnosis of a variety of pediatric syndromes that affect growth, such as endocrine disorders. The most commonly used method for bone age assessment (BAA) is still based on the comparison of the patient's hand and wrist radiograph to a bone age atlas. Such a method, however, takes considerable time, requires an expert rater, and suffers from high inter-rater variability. We present a deep-learning-based approach to estimate bone age from radiographs. It provides a fast, deterministic solution for bone-age assessment. We demonstrate the effectiveness of our method by using it to rate a set of 200 radiographs as part of a contest organized by the Radiological Society of North America. The results of this experiment have shown that our method's performance is similar to the one of a trained physician. Our system is available on-line, providing a free global service for doctors working in remote areas or in institutions with no BAA experts.
{"title":"End-to-End Bone Age Assessment with Residual Learning","authors":"Daniel Souza, M. M. O. Neto","doi":"10.1109/SIBGRAPI.2018.00032","DOIUrl":"https://doi.org/10.1109/SIBGRAPI.2018.00032","url":null,"abstract":"Bone age is a reliable metric for determining the level of biological maturity of children and adolescents. Its assessment is a crucial part of the diagnosis of a variety of pediatric syndromes that affect growth, such as endocrine disorders. The most commonly used method for bone age assessment (BAA) is still based on the comparison of the patient's hand and wrist radiograph to a bone age atlas. Such a method, however, takes considerable time, requires an expert rater, and suffers from high inter-rater variability. We present a deep-learning-based approach to estimate bone age from radiographs. It provides a fast, deterministic solution for bone-age assessment. We demonstrate the effectiveness of our method by using it to rate a set of 200 radiographs as part of a contest organized by the Radiological Society of North America. The results of this experiment have shown that our method's performance is similar to the one of a trained physician. Our system is available on-line, providing a free global service for doctors working in remote areas or in institutions with no BAA experts.","PeriodicalId":208985,"journal":{"name":"2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)","volume":"82 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122677438","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 : 2018-10-01DOI: 10.1109/SIBGRAPI.2018.00038
Ignacio Avas, Eduardo Fernández
Lighting intentions are the goals and constraints that designers like to achieve in a lighting design process. In this context, rendering problems are the kind of problems based on the rendering equation that are proposed to satisfy a set of lighting intentions. These problems are usually expressed as optimization problems. In this article is presented a novel method based on photon tracing, the VNS optimization metaheuristic, and the determination of the number of photons needed, which allows to handle a wider variety of lighting intentions without incurring in high computational costs. Moreover, the method developed shows to be efficient when the geometry is also a variable in the rendering problem. The techniques explained here could be included in a package used by architects or designers to aid in the lighting design process of architectural environments.
{"title":"A Photon Tracing Approach to Solve Inverse Rendering Problems","authors":"Ignacio Avas, Eduardo Fernández","doi":"10.1109/SIBGRAPI.2018.00038","DOIUrl":"https://doi.org/10.1109/SIBGRAPI.2018.00038","url":null,"abstract":"Lighting intentions are the goals and constraints that designers like to achieve in a lighting design process. In this context, rendering problems are the kind of problems based on the rendering equation that are proposed to satisfy a set of lighting intentions. These problems are usually expressed as optimization problems. In this article is presented a novel method based on photon tracing, the VNS optimization metaheuristic, and the determination of the number of photons needed, which allows to handle a wider variety of lighting intentions without incurring in high computational costs. Moreover, the method developed shows to be efficient when the geometry is also a variable in the rendering problem. The techniques explained here could be included in a package used by architects or designers to aid in the lighting design process of architectural environments.","PeriodicalId":208985,"journal":{"name":"2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)","volume":"137 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114144638","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 : 2018-10-01DOI: 10.1109/SIBGRAPI.2018.00027
E. F. Vernier, J. Comba, A. Telea
Computing treemap layouts for time-dependent (dynamic) trees is an open problem in information visualization. In particular, the constraints of spatial quality (cell aspect ratio) and stability (small treemap changes mandated by given tree-data changes) are hard to satisfy simultaneously. Most existing treemap methods focus on spatial quality, but are not inherently designed to address stability. We propose here a new treemapping method that aims to jointly optimize both these constraints. Our method is simple to implement, generic (handles any types of dynamic hierarchies), and fast. We compare our method with 14 state of the art treemaping algorithms using four quality metrics, over 28 dynamic hierarchies extracted from evolving software codebases. The comparison shows that our proposal jointly optimizes spatial quality and stability better than existing methods.
{"title":"A Stable Greedy Insertion Treemap Algorithm for Software Evolution Visualization","authors":"E. F. Vernier, J. Comba, A. Telea","doi":"10.1109/SIBGRAPI.2018.00027","DOIUrl":"https://doi.org/10.1109/SIBGRAPI.2018.00027","url":null,"abstract":"Computing treemap layouts for time-dependent (dynamic) trees is an open problem in information visualization. In particular, the constraints of spatial quality (cell aspect ratio) and stability (small treemap changes mandated by given tree-data changes) are hard to satisfy simultaneously. Most existing treemap methods focus on spatial quality, but are not inherently designed to address stability. We propose here a new treemapping method that aims to jointly optimize both these constraints. Our method is simple to implement, generic (handles any types of dynamic hierarchies), and fast. We compare our method with 14 state of the art treemaping algorithms using four quality metrics, over 28 dynamic hierarchies extracted from evolving software codebases. The comparison shows that our proposal jointly optimizes spatial quality and stability better than existing methods.","PeriodicalId":208985,"journal":{"name":"2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116933895","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 : 2018-10-01DOI: 10.1109/SIBGRAPI.2018.00059
F. L. Galvão, A. Falcão, A. Chowdhury
Methods for superpixel segmentation have become very popular in computer vision. Recently, a graph-based framework named ISF (Iterative Spanning Forest) was proposed to obtain connected superpixels (supervoxels in 3D) based on multiple executions of the Image Foresting Transform (IFT) algorithm from a given choice of four components: a seed sampling strategy, an adjacency relation, a connectivity function, and a seed recomputation procedure. In this paper, we extend ISF to introduce a unique characteristic among superpixel segmentation methods. Using the new framework, termed as Recursive Iterative Spanning Forest (RISF), one can recursively generate multiple segmentation scales on region adjacency graphs (i.e., a hierarchy of superpixels) without sacrificing the efficiency and effectiveness of ISF. In addition to a hierarchical segmentation, RISF allows a more effective geodesic seed sampling strategy, with no negative impact in the efficiency of the method. For a fixed number of scales using 2D and 3D image datasets, we show that RISF can consistently outperform the most competitive ISF-based methods.
{"title":"RISF: Recursive Iterative Spanning Forest for Superpixel Segmentation","authors":"F. L. Galvão, A. Falcão, A. Chowdhury","doi":"10.1109/SIBGRAPI.2018.00059","DOIUrl":"https://doi.org/10.1109/SIBGRAPI.2018.00059","url":null,"abstract":"Methods for superpixel segmentation have become very popular in computer vision. Recently, a graph-based framework named ISF (Iterative Spanning Forest) was proposed to obtain connected superpixels (supervoxels in 3D) based on multiple executions of the Image Foresting Transform (IFT) algorithm from a given choice of four components: a seed sampling strategy, an adjacency relation, a connectivity function, and a seed recomputation procedure. In this paper, we extend ISF to introduce a unique characteristic among superpixel segmentation methods. Using the new framework, termed as Recursive Iterative Spanning Forest (RISF), one can recursively generate multiple segmentation scales on region adjacency graphs (i.e., a hierarchy of superpixels) without sacrificing the efficiency and effectiveness of ISF. In addition to a hierarchical segmentation, RISF allows a more effective geodesic seed sampling strategy, with no negative impact in the efficiency of the method. For a fixed number of scales using 2D and 3D image datasets, we show that RISF can consistently outperform the most competitive ISF-based methods.","PeriodicalId":208985,"journal":{"name":"2018 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127509576","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：