Showing papers on "Bhattacharyya distance published in 1980"

Decorrelation Methods of Texture Feature Extraction

Abstract: This paper presents the development and evaluation of a visual texture feature extraction method based on a stochastic field model of texture. Results of recent visual texture discrimination experiments are reviewed in order to establish necessary and sufficient conditions for texture features that are in agreement with human discrimination. A texture feature extraction technique involving autocorrelation function measurement of a texture field, combined with histogram representation of a statistically decorrelated version of the texture field, is introduced. The texture feature extraction method is evaluated in terms of a Bhattacharyya distance measure.

Spectral distance measures between Gaussian processes

Abstract: Utilizing asymptotic results from prediction theory of multivariate stationary random processes and from regression theory for multivariate stationary processes, we develop asymptotic (large sample) expressions for the Chernoff coefficient, Bhattacharyya distance, I -divergence and J -divergence between two s -dimensional, covariance stationary Gaussian processes on the basis of n discrete-time samples. The expressions are given in terms of the two spectral density matrices F_{1}(\lambda), F_{2}(\lambda) derived from the two autocovariance matrix sequences, and of the spectral density matrix M(\lambda) related to the sequence of differences of mean vectors. The resulting spectral expressions are useful in a variety of applications, as discussed in the paper.

A Decision Theory Approach to the Approximation of Discrete Probability Densities

Abstract: The problem of approximating a probability density function by a simpler one is considered from a decision theory viewpoint. Among the family of candidate approximating densities, we seek the one that is most difficult to discriminate from the original. This formulation leads naturaliy to the density at the smallest Bhattacharyya distance. The resulting optimization problem is analyzed in detail.

The performance evaluations of the correlations of digital images - using figures of merit

Abstract: The performance of two dimensional correlations applied to a set of images of a scene, each exhibiting a different characteristic of it, is presented. The probability density function for the correlation output is assumed to be Gaussian. Both edge and area correla-tions are used. The Bayes probability of error, Chernoff and Bhattacharyya error bounds and Fisher's criteria are used as figures of merit to characterize the performance of the matching algorithm. Empirical results are given for a set of four images of a scene, i.e., down-looking, target-looking, real and synthetic images. These results show that for real images the target-looking view has better performance than the down-looking view and for synthetic images the down-looking view outperforms the target-looking view.