Learning to Combine Bottom-Up and Top-Down Segmentation

Levin, Anat; Weiss, Yair
January 2009
International Journal of Computer Vision;Jan2009, Vol. 81 Issue 1, p105
Academic Journal
Bottom-up segmentation based only on low-level cues is a notoriously difficult problem. This difficulty has lead to recent top-down segmentation algorithms that are based on class-specific image information. Despite the success of top-down algorithms, they often give coarse segmentations that can be significantly refined using low-level cues. This raises the question of how to combine both top-down and bottom-up cues in a principled manner. In this paper we approach this problem using supervised learning. Given a training set of ground truth segmentations we train a fragment-based segmentation algorithm which takes into account both bottom-up and top-down cues simultaneously, in contrast to most existing algorithms which train top-down and bottom-up modules separately. We formulate the problem in the framework of Conditional Random Fields (CRF) and derive a feature induction algorithm for CRF, which allows us to efficiently search over thousands of candidate fragments. Whereas pure top-down algorithms often require hundreds of fragments, our simultaneous learning procedure yields algorithms with a handful of fragments that are combined with low-level cues to efficiently compute high quality segmentations.


Related Articles

  • A Novel Approach for Foreign Substances Detection in Injection Using Clustering and Frame Difference. Guiliang Lu; Yu Zhou; Yao Yu; Sidan Du // Sensors (14248220);2011, Vol. 11 Issue 10, p9121 

    This paper focuses on developing a novel technique based on machine vision for detection of foreign substances in injections. Mechanical control yields spin/stop movement of injections which helps to cause relative movement between foreign substances in liquid and an ampoule bottle. Foreign...

  • A Review on Multiple Haze Removing Techniques for Single Image. Kaur, Satinderpal; Bansal, Meenakshi // International Journal of Advanced Research in Computer Science;2016, Vol. 7 Issue 6, p105 

    Single image haze removal has been a difficult drawback owing to its ill-posed nature. We have a tendency to propose an easy however powerful color attenuation previous for haze removal from one input hazy image. By making a linear model for modeling the scene depth of the hazy image below this...

  • Development of an Intelligent Virtual Grader for Estimation of Fruit Quality. Singh Chauhan, Ajay Pal; Partap Singh, Amar // International Journal of Computer Applications;2013, Vol. 62, p25 

    Food industry is amongst the industries that largely use machine vision for inspection of produce. Machine vision systems benefit from specially designed digital image processing software to accomplish this task. Color provides valuable information in examining the freshness and estimating the...

  • Semi-Supervised Learning Based Social Image Semantic Mining Algorithm. AO Guangwu; SHEN Minggang // Journal of Multimedia;Feb2014, Vol. 9 Issue 2, p245 

    As social image semantic mining is of great importance in social image retrieval, and it can also solve the problem of semantic gap. In this paper, a novel social image semantic mining algorithm based on semi-supervised learning is proposed. Firstly, labels which tagged the images in the test...

  • Unsupervised Learning of Object Detectors for Everyday Scenes. Khan, Najeed Ahmed; Hogg, David C. // International Journal of U- & E-Service, Science & Technology;2014, Vol. 7 Issue 1, p159 

    This paper proposes an unsupervised learning framework in which models of objects' appearance classes are learned using their spatio and temporal information, from video. These models are used to detect objects of different classes in the everyday scene. The proposed technique combines...

  • 3-D Depth Reconstruction from a Single Still Image. Saxena, Ashutosh; Chung, Sung; Ng, Andrew // International Journal of Computer Vision;Jan2008, Vol. 76 Issue 1, p53 

    We consider the task of 3-d depth estimation from a single still image. We take a supervised learning approach to this problem, in which we begin by collecting a training set of monocular images (of unstructured indoor and outdoor environments which include forests, sidewalks, trees, buildings,...

  • Decision Forests: A Unified Framework for Classification, Regression, Density Estimation, Manifold Learning and Semi-Supervised Learning. Criminisi, Antonio; Shotton, Jamie; Konukoglu, Ender // Foundations & Trends in Computer Graphics & Vision;2011, Vol. 7 Issue 2/3, preceding p83 

    This review presents a unified, efficient model of random decision forests which can be applied to a number of machine learning, computer vision, and medical image analysis tasks. Our model extends existing forest-based techniques as it unifies classification, regression, density estimation,...

  • UNSUPERVISED LEARNING OF APPEARANCE CLASSES FROM VIDEO. Khan, Najeed Ahmed; Hogg, David // NED University Journal of Research;2014, Vol. 11 Issue 2, p15 

    This paper proposes an unsupervised learning framework in which models of multiple objects' appearance classes are learned from video. These models are used to detect objects of different classes in the scene. The proposed technique combines appearance and motion features in a weighted...

  • Unsupervised Texture Segmentation Using Active Contour Model and Oscillating Information. Guodong Wang; Zhenkuan Pan; Qian Dong; Ximei Zhao; Zhimei Zhang; Jinming Duan // Journal of Applied Mathematics;2014, p1 

    Textures often occur in real-world images and may cause considerable difficulties in image segmentation. In order to segment texture images, we propose a new segmentation model that combines image decomposition model and active contour model. The former model is capable of decomposing structural...


Read the Article


Sorry, but this item is not currently available from your library.

Try another library?
Sign out of this library

Other Topics