Motivation: Combinations of classifiers have been found useful empirically, yet no formal proof exists about their generalization ability. Our goal is to develop a combination of kernel machines for which it is possible to prove generalization bounds. We believe that this is possible by further elaborating the arguments presented in [6], which may provide insights on boosting methods and view-b...

Even with the advent of more sophisticated, data-hungry methods, boosted decision trees remain extraordinarily successful for fast rigid object detection, achieving top accuracy on numerous datasets. While effective, most boosted detectors use decision trees with orthogonal (single feature) splits, and the topology of the resulting decision boundary may not be well matched to the natural topolo...

In this paper we propose a new method for texture segmentation based on the use of texture feature detectors derived from a decorrelation procedure of a modi®ed version of a Pseudo-Wigner distribution (PWD). The decorrelation procedure is accomplished by a cascade recursive least squared (CRLS) principal component (PC) neural network. The goal is to obtain a more ecient analysis of images by c...

Freehand 3D ultrasound can be acquired without a position sensor by using image-based positioning methods. In-plane motion can be tracked using image registration between nearby images. Elevational probe motion can be determined from the decorrelation between images. However, a freehand scan involves rotational as well as translational motion, and this affects the decorrelation. If this effect ...

My objective in this effort is to process and interpret acoustic monitoring data taken on both mine shapes and localized biological treatments during the 1995 ORCAS experiment. The ORCAS data allows examination of the acoustic scattering from mines and biological treatments for up to two months. Four types of results can be produced using the data: backscattering, scan-to-scan decorrelation, cu...

We present a straightforward way to use temporal decorrelation as preprocessing in linear and post-nonlinear independent component analysis (ICA) with higher order statistics (HOS). Contrary to the separation methods using second order statistics (SOS), the proposed method can be applied when the sources have similar temporal structure. The main idea is that componentwise decorrelation increase...

The effects of refractive turbulence along the path on the aerosol speckle field propagation and on the decorrelation time are studied for coherent pulsed lidar systems.

SAR interferometry requires a thorough phase control throughout the imaging process. The phaseaccuracy is affected bydecorrelation effects between the individual surveys. The paper presents the probability density function of the interferometric phase and analyzes quantitatively the influence of decorrelation. In panicular, phase aberrations typical for SAR processors are studied. The dependenc...