Showing papers in "Pattern Recognition Letters in 1996"

TL;DR: A new method to calculate the spatio-temporal correlation efficiently in a parametric eigenspace representation for moving object recognition that reduces the computational cost of correlation-based comparison between image sequences.

TL;DR: Genetic Algorithms have been used in an attempt to optimize a specified objective function related to a clustering problem and it is shown that the proposed method may improve the final output of K-Means where an improvement is possible.

TL;DR: A Fuzzy C-Means-based clustering method guided by an auxiliary (conditional) variable is introduced that reveals a structure within a family of patterns by considering their vicinity in a feature space along with the similarity of the values assumed by a certain conditional variable.

TL;DR: Validation of fuzzy partitions induced through c-shells clustering is considered, and a new set of indices are shown to be capable of validating the structure characterized by the shell clustering algorithms.

TL;DR: Examples are provided to show Dempster-Shafer evidence theory's ability to take into account a large variety of situations, which actually often occur and are not always well managed by classical approaches.

TL;DR: A Boolean characterization of three-dimensional simple points in cubic grids is proposed, which requires the checking of 5 basic configurations in the neighborhood of the point considered.

TL;DR: A new technique for pattern recognition by fuzzy integral is presented, and it is shown how to estimate importance of features, and their interaction.

TL;DR: A novel method based on data-dependent densities in a Choquet fuzzy integral is shown to outperform neural networks, Borda and weighted Borda counts, and Sugeno fuzzy integral.

TL;DR: Two new geodesic distance transforms for gray-scale images are presented that give a distance map in which the distance value of a single point corresponds to the length of the shortest discrete 8-path to the nearest background point.

TL;DR: Three applications representing different approaches to car detection and tracking problems are presented, and artificial neural networks are utilized to solve the essential subproblems.

TL;DR: A clustering algorithm that combines the advantages of fuzzy clustering and robust statistical estimators is presented and an extension of the algorithm to deal with an unknown number of clusters is proposed.

TL;DR: A new algorithm is proposed which has a sublinear time complexity and keeps the other features unchanged and is based on the recently introduced algorithm LAESA.

TL;DR: A general algorithm that performs basic mathematical morphology operations, like erosions and openings, with any arbitrary shaped structuring element in an efficient way is presented.

TL;DR: An experimental analysis of the use of different neural models for the supervised classification of multisensor remote-sensing data is presented, and the k -nearest neighbour statistical classifier is considered in order to evaluate the validity of the aforementioned neural networks, as compared with that of classical statistical methods.

TL;DR: This paper studies the four algorithms for clustering m objects in c clusters and compares their computational performance, and test these algorithms on several clustering problems from the literature as well as several random problems.

TL;DR: By accommodating for mixed pixels in the classification, more accurate, appropriate and useful outputs may be derived.

TL;DR: A new definition for the neighbors of an arbitrary point P is proposed, free from any user-specified parameter and can be used for pattern classification, clustering and low-level description of dot patterns.

TL;DR: It is argued that a straightforward fair comparison demands automatic classifiers with no user interaction, as this conflicts with one of the main characteristics of neural networks, their flexibility, the question whether they are better or worse than traditional techniques might be undecidable.

TL;DR: In this article, two initialisation methods are developed that are particularly suited to problems involving very large data sets and good results are obtained.

TL;DR: In this letter, skew estimation of Roman script is considered, the lowermost and uppermost pixels of some selected characters of the text which may be subject to Hough transform for skew angle detection are considered.

TL;DR: A new fast and exact algorithm for determining the k-NN classification of multichannel image data, and a new distance transform algorithm are described that demonstrates the effectiveness of the new classification algorithm.

TL;DR: Given two fuzzy subsets on support set S in a metric (e.g., Euclidean) space, a Hausdorff-like distance is defined that is a metric.

TL;DR: This letter presents a new methodology for determining the compatibility coefficients required for performing graph matching by probabilistic relaxation, and illustrates the application of this new theory in two graph matching applications.

TL;DR: This article demonstrates how the inclusion of context-related information in the entropy measure can be a simple matter, as is demonstrated in this article.

TL;DR: A parameter space to image space Hough transform locating ellipse centers is developed and the average time complexity of the algorithm is O(N2).

TL;DR: How the Fuzzy C-Means (FCM) classification algorithm can be used in this application and enhanced with an adaptive capability, enabling the supervisory control system to detect process departure from normal operating conditions is described.

TL;DR: Two methods for reducing the overfitting — a common problem in high-dimensional classification schemes — are presented, and the superiority of their combination is demonstrated.

TL;DR: In this paper, the notion of periodic lines is introduced and it is shown that periodic lines can be used to generate a wide range of granulometric functions via structural decomposition.

TL;DR: Representations that uniquely characterize images modulo the planar geometric similarity transformations are presented, and they are invariant with respect to translation, rotation, scaling, and their combination.

TL;DR: Experimental results show that the proposed approach is effective in reduction of color space dimensionality and edge detection in color images.