An Isotropic 3 3 Image Gradient Operator

Presentation at Stanford A.I. Project 1968 02/2014;
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we introduce the use of dispar- ity maps to alleviate the problem of appearance- based robot localization due to changes in illumi- nation. We describe how to use disparity maps for appearance-based localization and we compare the results obtained using disparity maps with those obtained with other techniques commonly used to reduce the efiect of illumination on im- ages: histogram equalization and gradient-based fllters. The results we present show that disparity maps are sensitive enough to rotations and trans- lations of the robot and that they are less sensitive to changes in illumination than previously used techniques. Consequently, we show that disparity maps are a valid alternative to achieve a robust appearance-based robot localization.
    International Conference on Advanced Robotics, Coimbra, Portugal; 01/2003
  • Source
  • [Show abstract] [Hide abstract]
    ABSTRACT: Abstract Edge detection and feature extraction are widely used in image processing and computer vision applications. In this paper we propose a new model for detecting edges in digital images. The basic idea is to model the image as a grid of the electric charges of a plan surface in electrostatic balance and, to affect for each block of size 3x3 pixels, a negative sign to the pixels with gray levels greater than a calculating threshold. The method operates in two phases: the first one consists in establishing the neighbourhood forces exerting on the central pixel for each defined block of the image, and the second one determines the horizontal and vertical masks corresponding to the resulting forces followed by the convolution operation. In order to validate and to illustrate the interest of the proposed approach, some experiments were made on a collection of synthetic and real images. As an objective quantitative measure, we have used Pratt's Figure of Merit to compare edge preservation performances to other methods.


Available from
Jun 2, 2014