Showing papers in "Pattern Recognition Letters in 1995"

TL;DR: A Bayesian formulation and a weighted majority vote (with weights obtained through a genetic algorithm) are implemented, and the combined performances of 7 classifiers on a large set of handwritten numerals are analyzed.

TL;DR: A hybrid numericsymbolic method that integrates an extended version of the K-means algorithm for cluster determination and a complementary conceptual characterization algorithm for clusters description is proposed.

TL;DR: The first labeled sample reduces the risk from 1 2 to 2R ∗ (1−R∗ ) and subsequent labeled samples in the training set reduce the probability of error exponentially fast to the Bayes risk.

TL;DR: In this paper, two color matching methods, the Reference Color Table Method (CTM) and the Distance Method (DM), were proposed for image retrieval. And the results show that both the new methods perform better than the existing method, and that the reference color table method gives the best results.

TL;DR: In this work, two models of the fish-eye transform are presented and the validity of the transformations is demonstrated by fitting the alternative models to a real fish-eyes lens.

TL;DR: A simple and efficient histogram-based approach is presented for multi-level thresholding that uses Gaussian kernel smoothing to detect peaks and valleys in a multimodal histogram, and uses a local maximum curvature method to detect points of discontinuity in a unimodalhistogram.

TL;DR: A genetic algorithm is applied for selecting a reference set for the k-Nearest Neighbors rule and the results are commented together with those obtained with the standard k-NN, random selection, Wilson's technique, and the MULTIEDIT algorithm.

TL;DR: A new image detection method based on background differencing and edge detection techniques, which separates the objects from their backgrounds and works well under various lighting and weather conditions is developed.

TL;DR: This work proposes a new method of texture analysis and classification based on a local center-symmetric covariance analysis, using Kullback (log-likelihood) discrimination of sample and prototype distributions, which finds that the new covariance measures perform better than Laws' approach despite their measure of texture pattern and grey-scale.

TL;DR: An interpretation phase is proposed, to complement usage of the Kohonen self-organizing feature map (SOFM) method, which segments the SOFM output, using an agglomerative contiguity-constrained clustering method.

TL;DR: The proposed method is an improvement over the differential box-counting method of Sarkar and Chaudhuri (1992, 1994) and is more accurate and efficient.

TL;DR: A method is proposed for finding decision boundaries, approximated by piecewise linear segments, for the classification of patterns in R 2, using an elitist model of a genetic algorithm that yields minimum misclassification.

TL;DR: A novel method of supervised clustering and classification using genetic algorithms, motivated by human perception and Kohonen's method, is presented, which drives the genetic search towards the smallest ellipsoid or set ofEllipsoids, which maximizes the number of correctly classified examples, and minimizes thenumber of misclassified examples.

TL;DR: A new 3D parallel thinning algorithm for medial surfaces that works in cubic grids with the 6-connectivity is proposed, based on a precise definition of end points which are points belonging to surfaces or curves.

TL;DR: A new sequential thinning algorithm, which uses both flag map and bitmap simultaneously to decide if a boundary pixel can be deleted, as well as the incorporation of smoothing templates to smooth the final skeleton, is proposed in this paper.

TL;DR: The generated sets of moment invariants, including higher-order moments, have been used for the recognition of English alphabets, such that the recognition rates have improved to over 98.8%.

TL;DR: A hierarchical, agglomerative clustering methodology is presented in which composite symbolic objects are formed using a cartesian join operator whenever symbolic objects were selected for agglomersation based on both similarity and dissimilarity.

TL;DR: A modification to the traditional concept of fuzzy connectedness is described, which extends the basic ideas to grey-level objects and a non-iterative method for detection of suboptimal paths is proposed, which simplifies the computation of such a measure.

TL;DR: This work clarifies the possible scope of steerability by Fourier decompositions, and approximate steerability with a limited number of basis functions, and the singularity that occurs when steering the scale.

TL;DR: An efficient method for tracking several objects moving through a sequence of monocular images against a non-uniform background is proposed which locks on each object as long as it moves in the scene.

TL;DR: The integrated function Algorithm for binarization of gray level document images is improved by smoothing, a new print pixel identification strategy, and a postprocessing step removing false print objects.

TL;DR: A new method is proposed to detect circular arcs that includes the determination of its center, radius, length, and midpoint using the Hough transform.

TL;DR: The initial segmentation is done by introducing the concept of rank of a point and the procedure generates polygons that are insensitive to rotation and scales and remains reliable in presence of noise.

TL;DR: A framework for solving the 3-D registration problem in medical imaging based on a canonical genetic algorithm (cGA) and an adaptive search space scaling technique is presented which operates by successive activations of the cGA procedure.

TL;DR: A simple and fast symmetry detection algorithm that works directly on the grey-scale image and does not require any prior segmentation of the input image is developed.

TL;DR: A simple and fast algorithm for polygonal approximation of digital curves is proposed, based on a greedy iterative elimination of a point with the currently minimal error value.

TL;DR: A method is proposed that employs genetic algorithms to determine a disparity map optimizing both the compatibility between corresponding points and the map continuity, and it is confirmed that use of GAs is effective when applied to stereo matching problems.

TL;DR: This paper proposes a 3D fully parallel thinning algorithm and demonstrates a possible way for designing a new generation of 3D Parallel thinning algorithms.

TL;DR: Moments and moment invariants can be normalized such that they become completely insensitive to the absolute intensity of the studied image, and this normalization can be used to further improve the recognition algorithm.

TL;DR: The Fuzzy k-Ellipses (FKE) algorithm was employed for the detection of the outer and inner contours of the heart's left ventricle from MRI images of transverse sections of the thorax.