Optimizing kd%2Dtrees for scalable visual descriptor indexing.

Learning to combine multi%2Dresolution spatially%2Dweighted co%2Doccurrence matrices for image representation.

Dynamic Video Collage.

Image search by concept map.

Image search by concept map.

Interactive image search by 2D semantic map.

Tag refinement by regularized LDA.

Summarizing tagged image collections by cross%2Dmedia representativeness voting.

Linear Neighborhood Propagation and Its Applications.

Graph%2Dbased semi%2Dsupervised learning with multiple labels.

Bayesian video search reranking.

Finding image exemplars using fast sparse affinity propagation.

Joint multi%2Dlabel multi%2Dinstance learning for image classification.

Normalized tree partitioning for image segmentation.

Maximum Margin Clustering with Pairwise Constraints.

Augmented tree partitioning for interactive image segmentation.

Augmented tree partitioning for interactive image segmentation.

Graph%2Dbased semi%2Dsupervised learning with multi%2Dlabel.

Transductive video annotation via local learnable kernel classifier.

Optimized video scene segmentation.

Transductive multi%2Dlabel learning for video concept detection.

Semi%2DSupervised Classification with Universum.
The Universum data, defined as a collection of ”nonexamples” that do not belong to any class of interest, have been shown to encode some prior knowledge by representing meaningful concepts in the same domain as the problem at hand. In this paper, we address a novel semi-supervised classification problem, called semi-supervised Universum, that can simultaneously utilize the labeled data, unlabeled data and the Universum data to improve the classification performance. We propose a graph based method to make use of the Universum data to help depict the prior information for possible classifiers. Like conventional graph based semi-supervised methods, the graph regularization is also utilized to favor the consistency between the labels. Furthermore, since the proposed method is a graph based one, it can be easily extended to the multiclass case. The empirical experiments on the USPS and MNIST datasets are presented to show that the proposed method can obtain superior performances over conventional supervised and semi-supervised methods. 1
Semi%2DSupervised Classification with Universum.
The Universum data, defined as a collection of ”nonexamples” that do not belong to any class of interest, have been shown to encode some prior knowledge by representing meaningful concepts in the same domain as the problem at hand. In this paper, we address a novel semi-supervised classification problem, called semi-supervised Universum, that can simultaneously utilize the labeled data, unlabeled data and the Universum data to improve the classification performance. We propose a graph based method to make use of the Universum data to help depict the prior information for possible classifiers. Like conventional graph based semi-supervised methods, the graph regularization is also utilized to favor the consistency between the labels. Furthermore, since the proposed method is a graph based one, it can be easily extended to the multiclass case. The empirical experiments on the USPS and MNIST datasets are presented to show that the proposed method can obtain superior performances over conventional supervised and semi-supervised methods. 1
MSRA atT TRECVID 2008 High%2DLevel Feature Extraction and Automatic Search.

Joint Affinity Propagation for Multiple View Segmentation.
A joint segmentation is a simultaneous segmentation of registered 2D images and 3D points reconstructed from the multiple view images. It is fundamental in structuring the data for subsequent modeling applications. In this paper, we treat this joint segmentation as a weighted graph labeling problem. First, we construct a 3D graph for the joint 3D and 2D points using a joint similarity measure. Then, we propose a hierarchical sparse affinity propagation algorithm to automatically and jointly segment 2D images and group 3D points. Third, a semi-supervised affinity propagation algorithm is proposed to refine the automatic results with the user assistance. Finally, intensive experiments demonstrate the effectiveness of the proposed approaches. 1.
Image%2Dbased tree modeling.
Figure 1: A tree is modeled using a set of input photographs. We show some examples of input and resulting 3D tree models. If image information is not available, e.g. the foliage is missing, the user is able to sketch it (right). The models approximate the input images while forming botanically plausible branching structures. We present a method for producing 3D tree models from input photographs with only limited user intervention. An approximate voxel-based tree volume is estimated using image information. The density values of the voxels are used to produce initial positions for a set of particles. Performing a 3D flow simulation, the particles are traced downwards to the tree basis and are combined to form twigs and branches. If possible, the trunk and the first-order branches are determined in the input photographs and are used as attractors for particle simulation. The geometry of the tree skeleton is produced using botanical rules for branch thicknesses and branching angles. Finally, leaves are added. Different initial seeds for particle simulation lead to a variety, yet similar-looking branching structures for a single set of photographs.
Image%2DBased Modeling by Joint Segmentation.

Face recognition using spectral features.

Face recognition using spectral features.

Semi%2DSupervised Classification Using Linear Neighborhood Propagation.
In this paper, we address the general problem of learning from both labeled and unlabeled data. Based on the reasonable assumption that the label of each data can be linearly reconstructed from its neighbors ’ labels, we develop a novel approach, called Linear Neighborhood Propagation (LNP), to learn the linear construction weights and predict the labels of the unlabeled data from the labeled data. The main procedure is composed of two steps: (1) it computes the local neighborhood geometry according to the image information by solving a constrained least squares problem, (2) the obtained geometry will be used to predict the labels of unlabeled data by solving a linearly constrained quadratic optimization problem. LNP has at least three characteristics. Firstly, it is much practical in that only one free parameter, the size of neighborhood, requires user-tuning. Secondly it can be easily extended to out-of-sample examples. Finally the labels achieved from LNP can be sufficiently smooth with respect to the intrinsic structure collectively revealed by the labeled and unlabeled points. Many promising experimental results, such as object recognition, data ranking and interactive image segmentation, demonstrate the effectiveness and efficiency of the proposed approach. 1
Picture Collage.
In this paper, we address a novel problem of automatically creating a picture collage from a group of images. Picture collage is a kind of visual image summary — to arrange all input images on a given canvas, allowing overlay, to maximize visible visual information. We formulate the picture collage creation problem in a Bayesian framework. The salient regions of each image are firstly extracted and represented as a set of weighted rectangles. Then, the image arrangement is formulated as a Maximum a Posterior (MAP) problem such that the output picture collage shows as many visible salient regions (without being overlaid by others) from all images as possible. Moreover, a very efficient Markov chain Monte Carlo (MCMC) method is designed for the optimization. Applications to desktop image browsing and image search result summarization demonstrate the effectiveness of our approach. 1
Image%2Dbased plant modeling.
Creating a 3D model from multiple view images with complex objects is a challenging problem. The segmentation of registered 2D images and 3D points, reconstructed from the multiple view images, to meaningful objects is viewed as the fundamental stage. In this paper, we first present a joint segmentation framework for image-based plant and indoor modeling, which simultaneously segments 3D points and 2D images. Then we describe an image fragmentation method to extract small leaves for image-based tree modeling. The experimental results demonstrate that the segmentation is critical for image-based modeling. 1.
Visual object recognition using probabilistic kernel subspace similarity.
 
Multi%2Dview EM algorithm and its application to color image segmentation.
In this paper we propose a new algorithm, Multi-View Expectation and Maximization algorithm(Multi-View EM), to deal with real-world learning problems where there are some natural splits of features. Multi-View EM does feature split as Co-training and Co-EM, two successful semi-supervised learning algorithms in text learning tasks, but it considers multi-view learning problem in the framework of EM algorithm. The Multi-View EM algorithm has these impressing advantages comparing with Co-training and Co-EM: its convergence is theoretically guaranteed; it can deal with multiple views instead of only two views. We utilize it for color image segmentation and discuss the phenomenon that different weights for color view and coordinate view will lead to different segmentation results. 1.
Multi%2Dview EM algorithm and its application to color image segmentation.
In this paper we propose a new algorithm, Multi-View Expectation and Maximization algorithm(Multi-View EM), to deal with real-world learning problems where there are some natural splits of features. Multi-View EM does feature split as Co-training and Co-EM, two successful semi-supervised learning algorithms in text learning tasks, but it considers multi-view learning problem in the framework of EM algorithm. The Multi-View EM algorithm has these impressing advantages comparing with Co-training and Co-EM: its convergence is theoretically guaranteed; it can deal with multiple views instead of only two views. We utilize it for color image segmentation and discuss the phenomenon that different weights for color view and coordinate view will lead to different segmentation results. 1.
Probabilistic tangent subspace a unified view.
Tangent Distance (TD) is one classical method for invariant pattern classification. However, conventional TD need pre-obtain tangent vectors, which is difficult except for image objects. This paper extends TD to more general pattern classification tasks. The basic assumption is that tangent vectors can be approximately represented by the pattern variations. We propose three probabilistic subspace models to encode the variations: the linear subspace, nonlinear subspace, and manifold subspace models. These three models are addressed in a unified view, namely Probabilistic Tangent Subspace (PTS). Experiments show that PTS can achieve promising classification performance in non-image data sets. 1.
Kernel Trick Embedded Gaussian Mixture Model.
Abstract. In this paper, we present a kernel trick embedded Gaussian Mixture Model (GMM), called kernel GMM. The basic idea is to embed kernel trick into EM algorithm and deduce a parameter estimation algorithm for GMM in feature space. Kernel GMM could be viewed as a Bayesian Kernel Method. Compared with most classical kernel methods, the proposed method can solve problems in probabilistic framework. Moreover, it can tackle nonlinear problems better than the traditional GMM. To avoid great computational cost problem existing in most kernel methods upon large scale data set, we also employ a Monte Carlo sampling technique to speed up kernel GMM so that it is more practical and efficient. Experimental results on synthetic and real-world data set demonstrate that the proposed approach has satisfing performance. 1
Color Image Segmentation Kernel Do the Feature Space.

