Results 1  10
of
12
Efficient Subwindow Search: A Branch and Bound Framework for Object Localization
 IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
"... Most successful object recognition systems rely on binary classification, deciding only if an object is present or not, but not providing information on the actual object location. To estimate the object’s location one can take a sliding window approach, but this strongly increases the computational ..."
Abstract

Cited by 122 (10 self)
 Add to MetaCart
Most successful object recognition systems rely on binary classification, deciding only if an object is present or not, but not providing information on the actual object location. To estimate the object’s location one can take a sliding window approach, but this strongly increases the computational cost, because the classifier or similarity function has to be evaluated over a large set of candidate subwindows. In this paper, we propose a simple yet powerful branch and bound scheme that allows efficient maximization of a large class of quality functions over all possible subimages. It converges to a globally optimal solution typically in linear or even sublinear time, in constrast to the quadratic scaling of exhaustive or sliding window search. We show how our method is applicable to different object detection and image retrieval scenarios. The achieved speedup allows the use of classifiers for localization that formerly were considered too slow for this task, such as SVMs with a spatial pyramid kernel or nearest neighbor classifiers based on the χ²distance. We demonstrate stateoftheart localization performance of the resulting systems on the
Robust Least Square Baseline Finding using a Branch and Bound Algorithm
 in Document Recognition and Retrieval VIII, SPIE
, 2002
"... Many document analysis and OCR systems depend on precise identification of page rotation, as well as the reliable identification of text lines. This paper presents a new algorithm to address both problems. It uses a branchandbound approach to globally optimal line finding and simultaneously models ..."
Abstract

Cited by 23 (18 self)
 Add to MetaCart
Many document analysis and OCR systems depend on precise identification of page rotation, as well as the reliable identification of text lines. This paper presents a new algorithm to address both problems. It uses a branchandbound approach to globally optimal line finding and simultaneously models the baseline and the descender line under a Gaussian error/robust least square model. Results of applying the algorithm to documents in the University of Washington Database 2 are presented. Keywords: document analysis, layout analysis, skew detection, page rotation, text line finding, a#ne transformations 1.
Implementation Techniques for Geometric BranchandBound Matching Methods
 CVIU
, 2003
"... This paper describes matchlistbased branchandbound techniques and presents a number of new applications of branchandbound methods, among them, a method for globally optimal partial line segment matching under bounded or Gaussian error, point matching under a Gaussian error model with subpixel a ..."
Abstract

Cited by 16 (5 self)
 Add to MetaCart
(Show Context)
This paper describes matchlistbased branchandbound techniques and presents a number of new applications of branchandbound methods, among them, a method for globally optimal partial line segment matching under bounded or Gaussian error, point matching under a Gaussian error model with subpixel accuracy and precise orientation models, and a simple and robust technique for finding multiple distinct object instances. It also contains extensive reference information for the implementation of such matching methods under a wide variety of error bounds and transformations. In addition, the paper contains a number of benchmarks and evaluations that provide new information about the runtime behavior of branchandbound matching algorithms in general, and that help choose among different implementation strategies, such as the use of point location data structures and space/time tradeoffs involving depthfirst search
A Practical, Globally Optimal Algorithm for Geometric Matching under Uncertainty
 Proc. International Workshop on Combinatorial Image Analysis (IWCIA 2001
, 2001
"... Geometric matching under uncertainty is a longstanding problem in computer vision. This paper presents a simple and e#cient algorithm for finding globally optimal solutions to geometric matching problems under a wide variety of allowable transformations (translations, isometries, equiform transform ..."
Abstract

Cited by 12 (4 self)
 Add to MetaCart
(Show Context)
Geometric matching under uncertainty is a longstanding problem in computer vision. This paper presents a simple and e#cient algorithm for finding globally optimal solutions to geometric matching problems under a wide variety of allowable transformations (translations, isometries, equiform transformations, others) and a wide variety of allowable feature types (point features, oriented point features, line features, line segment features, etc.). The algorithm only requires an implementation of the forward transformation (modeltoimage) and an error model to be supplied. Benchmarks and comparisons of the algorithm in comparison with alignment and Hough transform methods are presented.
A Comparison of Search Strategies for Geometric Branch and Bound Algorithms
 In: European Conference on Computer Vision
, 2002
"... Over the last decade, a number of methods for geometric matching based on a branchandbound approach have been proposed. Such algorithms work by recursively subdividing transformation space and bounding the quality of match over each subdivision. No direct comparison of the major implementation str ..."
Abstract

Cited by 10 (2 self)
 Add to MetaCart
(Show Context)
Over the last decade, a number of methods for geometric matching based on a branchandbound approach have been proposed. Such algorithms work by recursively subdividing transformation space and bounding the quality of match over each subdivision. No direct comparison of the major implementation strategies has been made so far, so it has been unclear what the relative performance of the different approaches is. This paper examines experimentally the relative performance of different implementation choices in the implementation of branchandbound algorithms for geometric matching: alternatives for the computation of upper bounds across a collection of features, and alternatives the order in which search nodes are expanded. Two major approaches to computing the bounds have been proposed: the matchlist based approach, and approaches based on point location data structures. A second issue that is addressed in the paper is the question of search strategy; branchandbound algorithms traditionally use a "bestfirst" search strategy, but a "depthfirst" strategy is a plausible alternative. These alternative implementations are compared on an easily reproducible and commonly used class of test problems, a statistical model of feature distributions and matching within the COIL20 image database. The experimental results show that matchlist based approaches outperform point location based approaches on common tasks. The paper also shows that a depthfirst approach to matching results in a 50200 fold reduction in memory usage with only a small increase in running time. Since matchlistbased approaches are significantly easier to implement and can easily cope with a much wider variety of feature types and error bounds that point location based approaches, ...
Efficient Discovery of Spatial Associations and Structure with Application to Asteroid Tracking
, 2005
"... The problem of finding sets of points that conform to a given underlying spatial model is a conceptually simple, but potentially expensive, task that arises in a variety of domains. The goal is simply to find occurrences of known types of spatial structure in the data. However, as we begin to examin ..."
Abstract

Cited by 3 (1 self)
 Add to MetaCart
The problem of finding sets of points that conform to a given underlying spatial model is a conceptually simple, but potentially expensive, task that arises in a variety of domains. The goal is simply to find occurrences of known types of spatial structure in the data. However, as we begin to examine large, dense, and noisy data sets the cost of finding such occurrences can increase rapidly. In this thesis I consider the computational issues inherent in extracting modelbased spatial associations and structure from large amounts of noisy data. In particular, I discuss the development of new techniques and algorithms that mitigate or eliminate these computational issues. I show that there are several different types of structure in both the data and the problem itself that can often be exploited to this end. Primarily, I describe a new type of treebased search algorithm that uses a variable number of tree nodes to adapt to both structure in the data and search state itself. While the problem of finding known types of spatial structure arises in a wide range of domains, the primary motivating problem throughout this thesis is the task of asteroid
Application of the FisherRao Metric to Structure Detection
, 2005
"... Abstract Certain structure detection problems can be solved by sampling a parameter space for the different structures at a finite number of points and checking each point to see if the corresponding structure has a sufficient number of inlying measurements. The measurement space is a Riemannian m ..."
Abstract

Cited by 2 (2 self)
 Add to MetaCart
(Show Context)
Abstract Certain structure detection problems can be solved by sampling a parameter space for the different structures at a finite number of points and checking each point to see if the corresponding structure has a sufficient number of inlying measurements. The measurement space is a Riemannian manifold and the measurements relevant to a given structure are near to or on a submanifold which constitutes the structure. The probability density function for the errors in the measurements is described using a generalisation of the Gaussian density to Riemannian manifolds. The conditional probability density function for the measurements yields the Fisher information which defines a metric, known as the FisherRao metric, on the parameter space. The main result is a derivation of an asymptotic approximation to the FisherRao metric, under the assumption that the measurement noise is small. Using this approximation to the FisherRao metric, the parameter space is sampled, such that each point of the parameter space is near to at least one sample point, to within the level of accuracy allowed by the measurement errors. The probability of a false detection of a structure is estimated. The feasibility of this ap1 proach tostructure detection is tested experimentally using the example ofline detection in digital images. Index Termsasymptotic approximation, false detection, FisherRao metric, heat equation, Hough transform, line detection, parameter space, Riemannian manifold. 1
Optimal Line and Arc Detection on RunLength Representations
"... The robust detection of lines and arcs in scanned documents or technical drawings is an important problem in document image understanding. We present a new solution to this problem that works directly on runlength encoded data. The method finds globally optimal solutions to parameterized thick line ..."
Abstract

Cited by 1 (1 self)
 Add to MetaCart
(Show Context)
The robust detection of lines and arcs in scanned documents or technical drawings is an important problem in document image understanding. We present a new solution to this problem that works directly on runlength encoded data. The method finds globally optimal solutions to parameterized thick line and arc models. Line thickness is part of the model and used during the matching process. Unlike previous approaches, it does not require any thinning or other preprocessing steps, no computation of the line adjacency graphs, and no heuristics. Furthermore, the only searchrelated parameter that needs to be specified is the desired numerical accuracy of the solution. The method is based on a branchandbound approach for the globally optimal detection of these geometric primitives using runs of black pixels in a bilevel image. We present qualitative results of the algorithm on images used in the 2003 GREC arc segmentation contest.
On the Use of Interval Arithmetic in Geometric Branch and Bound Algorithms
, 2003
"... Branch and bound methods have become established methods for geometric matching over the last decade. This paper presents techniques that improve on previous branch and bound methods in two important ways: they guarantee reliable solutions even in the presence of numerical roundoff error, and they e ..."
Abstract

Cited by 1 (0 self)
 Add to MetaCart
Branch and bound methods have become established methods for geometric matching over the last decade. This paper presents techniques that improve on previous branch and bound methods in two important ways: they guarantee reliable solutions even in the presence of numerical roundoff error, and they eliminate the need to derive bounding functions manually. These new techniques are compared experimentally with recognitionbyalignment and previous branch and bound techniques on geometric matching problems. Novel methods for nonlinear baseline finding and globally optimal robust linear regression using these techniques are described.
Line detection with adaptive random samples
, 2011
"... This paper examines the detection of parameterized shapes in multidimensional noisy grayscale images. A novel shape detection algorithm utilizing random sample theory is presented. Although the method can be generalized, line detection is detailed. Each line in the image corresponds to a point in th ..."
Abstract

Cited by 1 (0 self)
 Add to MetaCart
This paper examines the detection of parameterized shapes in multidimensional noisy grayscale images. A novel shape detection algorithm utilizing random sample theory is presented. Although the method can be generalized, line detection is detailed. Each line in the image corresponds to a point in the line parameter space. The method creates hypothesis lines by randomly selecting parameter space points and tests the surrounding regions for acceptable linear features. The information obtained from each randomly selected line is used to update the parameter distribution, which reduces the required number of random trials. The selected lines are reestimated within a smaller search space with a more accurate algorithm like the Hough transform (HT). Faster results are obtained compared to HT, without losing performance as in other faster HT variants. The method is robust and suitable for binary or grayscale images. Results are given from both simulated and experimental subsurface seismic and ground penetrating radar (GPR) images when searching for features like pipes or tunnels.