A heuristic method has been developed for registering two sets of 3-D curves obtained by using an edge-based stereo system, or two dense 3-D maps obtained by using a correlation-based stereo system. Geometric matching in general is a difficult unsolved problem in computer vision. Fortunately, in many practical applications, some a priori knowledge exists which considerably simplifies the problem. In visual navigation, for example, the motion between successive positions is usually approximately known. From this initial estimate, our algorithm computes observer motion with very good precision, which is required for environment modeling (e.g., building a Digital Elevation Map). Objects are represented by a set of 3-D points, which are considered as the samples of a surface. No constraint is imposed on the form of the objects. The proposed algorithm is based on iteratively matching points in one set to the closest points in the other. A statistical method based on the distance distribution is used to deal with outliers, occlusion, appearance and disappearance, which allows us to do subset-subset matching. A least-squares technique is used to estimate 3-D motion from the point correspondences, which reduces the average distance between points in the two sets. Both synthetic and real data have been used to test the algorithm, and the results show that it is efficient and robust, and yields an accurate motion estimate.

This paper provides a review of shape analysis methods. Shape analysis methods play an important role in systems for object recognition, matching, registration, and analysis. Researchin shape analysis has been motivated, in part, by studies of human visual form perception systems.

A great deal of work has been done on the evaluation of information retrieval systems for alphanumeric data. The same thing can not be said about the newly emerging multimedia and image database systems. One of the central concerns in these systems is the automatic characterization of image content and retrieval of images based on similarity of image content. In this paper, we discuss effectiveness of several shape measures for content based similarity retrieval of images. The different shape measures we have implemented include outline based features (chain code based string features, Fourier descriptors, UNL Fourier features), region based features (invariant moments, Zemike moments, pseudo-Zemike moments), and combined features (invariant moments & Fourier descriptors, invariant moments & UNL Fourier features). Given an image, all these shape feature measures (vectors) are computed automatically, and the feature vector can either be used for the retrieval purpose or can be stored in the database for future queries. We have tested all of the above shape features for image retrieval on a database of 500 trademark images. The average retrieval efficiency values computed over a set of fifteen representative queries for all the methods is presented. The output of a sample shape similarity query using all the features is also shown.

We investigate the problem of retrieving similar shapes from a large database; in particular, we focus on medical tumor shapes (“Find tumors that are similar to a given pattern.”). We use a natural similarity function for shape-matching, based on concepts from mathematical morphology, and we show how it can be lower-bounded by a set of shape features for safely pruning candidates, thus giving fast and correct output. These features can be organized in a spatial access method, leading to fast indexing for range queries and nearest-neighbor queries. In addition to the lower-bounding, our second contribution is the design of a fast algorithm for nearest-neighbor search, achieving significant speedup while provably guaranteeing correctness. Our experiments demonstrate that roughly 90 percent of the candidates can be pruned using these techniques, resulting in up to 27 times better performance compared to sequential scan.

We first describe a multiresolutional discretization scheme for the boundary of a planar connected shape, by means of the Haar wavelet transform. This scheme yields a hierarchy of polygons of increasing complexity that adaptively captures the shape's boundary at increasing resolutions. Next, we present a new morphological transform called the Chordal Axis Transform for planar shapes that leads to a constructive definition of the skeleton of a discretized shape. We describe an efficient, adaptive algorithm based on the Constrained Delaunay Triangulation of polygons for skeletonizing the shape. The resulting skeleton is connected and faithfully captures the structure of the shape. Finally, we obtain a simple quantitative descriptor of the morphology of a discretized shape based on its Delaunay triangulation, and introduce the concept of morphological congruence. This leads to a graph theoretic characterization of the nature of structural differences between shapes that are morphologicall...

Abstract not found

Abstract—With inspiration from psychophysical researches of the human visual system, we propose a novel aspect and a method for performance evaluation of contour-based shape recognition algorithms regarding their robustness to incompleteness of contours. We use complete contour representations of objects as a reference (training) set. Incomplete contour representations of the same objects are used as a test set. The performance of an algorithm is reported using the recognition rate as a function of the percentage of contour retained. We call this evaluation procedure the ICR test. We consider three types of contour incompleteness, viz. segment-wise contour deletion, occlusion, and random pixel depletion. As an illustration, the robustness of two shape recognition algorithms to contour incompleteness is evaluated. These algorithms use a shape context and a distance multiset as local shape descriptors. Qualitatively, both algorithms mimic human visual perception in the sense that recognition performance monotonously increases with the degree of completeness and that they perform best in the case of random depletion and worst in the case of occluded contours. The distance multiset method performs better than the shape context method in this test framework.

A novel shape description method based on the Morphological Signature Transform (MST) is presented in this dissertation. The MST uses multiresolution morphological image processing by non-convex multiple structuring elements. A binary image which contains the object shape to be described is represented by means of a multiresolution pyramid. The method is based on the successive morphological erosions of the input image at di erent resolutions by primary and rotated structuring elements. The areas of successively eroded images are computed for each structuring element at each pyramid level. The obtained set of numbers is arranged into vectors, ordered, and used as a shape descriptor. Experimental results demonstrate that the method is robust against noise and invariant to translation, rotation, and scale change. A new method for the selection of the optimal structuring element is presented in the second part of dissertation. For a given class of shapes the optimal structuring element for MST method is selected by means of a genetic algorithm. The optimization criteria is formulated to enable a robust shape matching. Experiments have been performed on a class of model shapes. The proposed optimal shape description method is applied to the problem of shape matching which

Simple nonlinear filters are often used to enforce "hard" syntactic constraints while remaining close to the observation data; e.g., in the binary case it is common practice to employ iterations of a suitable median, or a one-pass recursive median, openclose, or closopen filter to impose a minimum symbol run-length constraint while remaining "faithful" to the observation. Unfortunately, these filters are - in general - suboptimal. Motivated by this observation, we pose the following optimization: Given a finite-alphabet sequence of finite extent, y = fy(n)g N \Gamma1 n=0 , find a sequence, b x = fbx(n)g N \Gamma1 n=0 , which minimizes d(x; y) = P N \Gamma1 n=0 dn (y(n); x(n)) subject to: x is piecewise constant of plateau run-length M . We show how a suitable reformulation of the problem naturally leads to a simple and efficient Viterbi-type optimal algorithmic solution. We call the resulting nonlinear input-output operator the Viterbi Optimal Runlength-Constrained Approximation...

Cutaneous melanocytic lesions, commonly known as moles, are mostly benign; however, some of them are malignant melanomas, the most fatal form of skin cancer. Because the survival rate of melanoma is inversely proportional to the thickness of the tumor, early detection is vital to the treatment process. Many dermatologists have advocated the development of computer-aided diagnosis systems for early detection of melanoma. One of the important clinical features differentiating benign nevi from malignant melanomas is the lesion border irregularity. There are two types of border irregularity: texture and structure irregularities. Texture irregularities are the small variations along the border, while structure irregularities are the global indentations and protrusions that may suggest either the unstable growth in a lesion or regression of a melanoma. An accurate measurement of structure irregularities is essential to detect the malignancy of melanoma. This thesis extends the classic curvature scale-space filtering technique to locate all structure irregular segments along a melanocytic lesion border. An area-based index, called irregularity index, is then computed for each segment. From the individual irregularity index, two important new measures, the most significant irregularity index and the overall irregularity index, are derived. These two indices describe the degree of irregularity along the lesion border . A double-blind user study is performed to compare the new measures with twenty experienced dermatologists' evaluations. Forty melanocytic lesion images were selected and their borders were extracted automatically after dark thick hairs were removed by a preprocessor called DullRazor. The overall irregularity index and the most significant irregularity index we...