Electrocardiogradia De Alta Frecuencia: Estudio Del Espectro De Potencia En La Cuantificacion De Componentes De Alta Contour Motion Estimation Using 2D Distribution Of Displacement Vectors Field

839

CONTOUR MOTION ESTIMATION USING 2D DISTRIBUTION OF DISPLACEMENT VECTORS FIELD

L. Urbano, F. D’Alvano, C. Lollett ,G. Gonzalez, T. Adrian de Perez, C. Bruscianelli.
Grupo de Procesamiento de Seiiales - Departamento de ElectrBnica y Circuitos Universidad SimBn Bolivar Apartado 89000, Caracas 1080A - Venezuela Phone: (+582) 9063630 9063666 Fax:(+582) 906363 1 e-mail: lurbano@usb.veABSTRACT
The motion information obtained from a sequence of time-varying two-dimensional images gives fundamental clues about the three-dimensional structure of the observed scene. For many vision systems, the main task is to extract that information. In this work, we present a quantitative and qualitative analysis of the 2D motion field obtained applying optical flow algorithms (with either globalor local conditions), and pel-recursive algorithms to image sequences (using real and synthetic data). Special emphasis is placed on assessing the performance of both techniques in scene segmentation. The experiments were carried out in two stages: the first one consisted of a qualitative analysis of the results of the motion field estimation process for a known motion field; the second one wasthe quantitative evaluation of the applicability of the results to scene segmentation, pondering sensitivity and specificity as figures of merit. The results are presented and discussed. classified as parametric and nonparametric [l]. The first ones are based on il modeling of the 2-D optical flow field in terms of rigid motion and structural parameters. These parametric models can be straightlyused for either 3-D image analysis, such as in objectbased image compression and passive navigation, or improved 2-D motion estimation. The second family of methods, which are refered to as optical flow estimation, or the correspondence problem solution, aims to estimate motion projected onto the image plane in terms of instantaneous pixel velocities or frame-to-frame pixel correspondences. Inthis category we find, among others, block-based, pel-recursive and Bayesian methods, and those based on the optical flow equation. These schemes are usually the foundation (lower levels) of many image analysis hierarchies. Much of the work in dynamic scene analysis depends on the availability of segmented frames. If accurate information on the location of visible boundary elements is available,sophisticated techniques may be used to track the moving objects and resolve ambiguities due to occlusion [2]. When assessing the performance of motion estimation methods, a fundamental problem is the lack of a simple, but subjectively meaningful, quantitative measure. In this work, we present a quantitative and qualitative analysis of the application of the 2D distribution of displacement vectorfields, obtained using nonparametric methods.

1. INTRODUCTION

-

The motion estimation of a shifting object from a sequence of images is of prime interest in computer vision. Many different analysis hierarchies can be used to extract the information present in a sequence of time-varying two-dimensional images. However, a widespread feature of them is the displaying of the motion informationas a 2-D distribution of displacement vectors, called the optical flow field. In certain applications, the optical flow (also called image flow) field contains all the necessary motion information for solving the particular computer vision problem, such as recovering the 3-D velocity field, segmenting the image into parts corresponding to different moving objects, or reconstructing the 3-Dstructure of surfaces in the viewed scene. The 2-D motion estimation methods have been

2. NONPARAMETRIC 2-D MOTION METHODS 2.1 Gradient Algorithm or Optical Flow Algorithm.
Optical Flow Algorithms are based on the optical flow equation [3]. This methods attempt to provide an estimate of the optical fllow field in terms of spatiotemporal image intensity gradients, and are used in conjunction with...