A voting-based computational framework for visual motion analysis and interpretation

Nicolescu, Mircea; Medioni, Gérard

doi:10.1109/tpami.2005.91

Cited by 31 publications

(12 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, formulating object segmentation as motion segmentation using optical flow rests on the assumption of brightness constancy, which is violated at moving boundaries, resulting in poor estimates of object contours [33]. Object segmentation also tries to detect and segment the observed motions into semantic meaningful instances of particular activities from videos [17].…”

Section: Related Workmentioning

confidence: 99%

Unsupervised Object Modeling and Segmentation with Symmetry Detection for Human Activity Recognition

Cheng

Huang

2015

Symmetry

View full text Add to dashboard Cite

Abstract:In this paper we present a novel unsupervised approach to detecting and segmenting objects as well as their constituent symmetric parts in an image. Traditional unsupervised image segmentation is limited by two obvious deficiencies: the object detection accuracy degrades with the misaligned boundaries between the segmented regions and the target, and pre-learned models are required to group regions into meaningful objects. To tackle these difficulties, the proposed approach aims at incorporating the pair-wise detection of symmetric patches to achieve the goal of segmenting images into symmetric parts. The skeletons of these symmetric parts then provide estimates of the bounding boxes to locate the target objects. Finally, for each detected object, the graphcut-based segmentation algorithm is applied to find its contour. The proposed approach has significant advantages: no a priori object models are used, and multiple objects are detected. To verify the effectiveness of the approach based on the cues that a face part contains an oval shape and skin colors, human objects are extracted from among the detected objects. The detected human objects and their parts are finally tracked across video frames to capture the object part movements for learning the human activity models from video clips. Experimental results show that the proposed method gives good performance on publicly available datasets.

show abstract

Section: Related Workmentioning

confidence: 99%

Unsupervised Object Modeling and Segmentation with Symmetry Detection for Human Activity Recognition

Cheng

Huang

2015

Symmetry

View full text Add to dashboard Cite

show abstract

“…or video is an important aspect of modern computer vision, in fields such as video surveillance [36,30, and references therein], vehicle control [16], crowd behavior analysis [35], and other applications [40].…”

Section: Directions Regularization Analysis Of Principal Directions mentioning

confidence: 99%

Augmented-Lagrangian regularization of matrix-valued maps

Rosman

Tai

Kimmel

et al. 2014

Methods and Applications of Analysis

View full text Add to dashboard Cite

Abstract. We propose a novel framework for fast regularization of matrix-valued images. The resulting algorithms allow a unified treatment for a broad set of matrix groups and manifolds. Using an augmented-Lagrangian technique, we formulate a fast and highly parallel algorithm for matrixvalued image regularization.We demonstrate the applicability of the framework for various problems, such as motion analysis and diffusion tensor image reconstruction, show the formulation of the algorithm in terms of splitBregman iterations and discuss the convergence properties of the proposed algorithms.Key words. Regularization, Lie-groups, total-variation, split-Bregman, matrix-manifolds, diffusion-imaging, rotations, articulated motion.AMS subject classifications. 65K10, 58JXX.1. Introduction. Matrix-manifolds and Matrix-valued images have become an integral part of computer vision and image processing. Matrix-manifolds and groups have been used for tracking [43,56], robotics [38,58,10], motion analysis [45,20], image processing and computer vision [9,40,42,45,60], as well as medical imaging [4,39]. Efficient regularization of matrix-valued images is therefore highly important in the fields of image analysis and computer vision. This includes applications such as direction diffusion [29,53,59] and scene motion analysis [33] in computer vision, as well as diffusion tensor MRI (DT-MRI) regularization [5,15,25,50,54] in medical imaging.We present an augmented Lagrangian method for efficient regularization of matrix-valued images, or maps. We assume the matrix-manifold to have an efficient projection operator onto it from some embedding into a Euclidean space, and that the distortion associated with this mapping is not too large in term of the metric accompanying these spaces.Examples of matrix-manifolds that are of interest include the special-orthogonal and special-Euclidean Lie-groups and the symmetric positive-definite matrices, as well as Stiefel manifolds. We show that the augmented Lagrangian technique allows us to separate the optimization process into a regularization update step of a map onto an embedding-space, and a per-pixel projection step. An efficient regularization step is shown for the total-variation (TV, [48]) regularization, and a second-order regularization penalizing the Hessian norm. Both the regularization step and the projection steps are simple to compute, fast and easily parallelizable using consumer graphic processing units (GPUs), achieving real-time processing rates. The resulting framework unifies algorithms using in several application domains into one framework, since they differ only in the choice of projection operator. While such an optimization problem could have been approached by general saddle-point solvers such as [12], the

show abstract

“…This technique has been proven versatile, since it has successfully been adapted to problems well beyond the ones to which it was originally applied with excellent results. For example, this method has already been applied to a variety of problems in image and video processing, such as perceptual organization [26,43], image restoration [15], image segmentation [22,29], video segmentation [36,27], mesh analysis [17], 3D reconstruction [49] and dimensionality estimation [28]. Since the input data for most of these applications are not clouds of points, a common approach is to apply tensor voting as described in [26] to clouds of points derived from the original data.…”

Section: Introductionmentioning

confidence: 99%

Adaptation of Tensor Voting to Image Structure Estimation

Moreno

Pizarro

Burgeth

et al. 2012

Mathematics and Visualization

View full text Add to dashboard Cite

Tensor voting is a well-known robust technique for extracting perceptual information from clouds of points. This chapter proposes a general methodology to adapt tensor voting to different types of images in the specific context of image structure estimation. This methodology is based on the structural relationships between tensor voting and the so-called structure tensor, which is the most popu- lar technique for image structure estimation. The problematic Gaussian convolution used by the structure tensor is replaced by tensor voting. Afterwards, the results are appropriately rescaled. This methodology is adapted to gray-valued, color, vectorand tensor-valued images. Results show that tensor voting can estimate image structure more appropriately than the structure tensor and also more robustly.

show abstract

A voting-based computational framework for visual motion analysis and interpretation

Cited by 31 publications

References 43 publications

Unsupervised Object Modeling and Segmentation with Symmetry Detection for Human Activity Recognition

Unsupervised Object Modeling and Segmentation with Symmetry Detection for Human Activity Recognition

Augmented-Lagrangian regularization of matrix-valued maps

Adaptation of Tensor Voting to Image Structure Estimation

Contact Info

Product

Resources

About