Browse Prior Art Database

Cluster Estimation in Hough Space for Image Analysis

IP.com Disclosure Number: IPCOM000059787D
Original Publication Date: 1986-Jan-01
Included in the Prior Art Database: 2005-Mar-08
Document File: 3 page(s) / 21K

Publishing Venue

IBM

Related People

Biland, HP: AUTHOR [+2]

Abstract

The problem of finding shapes in images is important, e.g., for automated manufacturing and inspection. One powerful approach for image analysis is the so-called Hough transform technique which maps the image space into the Hough space. One essential step in the evaluation is the de- termination of clusters in the Hough space, and present article suggests a method for finding such clusters. The Hough space is usually represented by an n-dimensional array of numbers. Each element of the image defines a hypersurface in the Hough space. The hypersurfaces are determined by the shape to be identified in the image. As the mapping from the image space to the Hough space proceeds, these hypersurfaces are accumulated in the array.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 3

Cluster Estimation in Hough Space for Image Analysis

The problem of finding shapes in images is important, e.g., for automated manufacturing and inspection. One powerful approach for image analysis is the so-called Hough transform technique which maps the image space into the Hough space. One essential step in the evaluation is the de- termination of clusters in the Hough space, and present article suggests a method for finding such clusters. The Hough space is usually represented by an n-dimensional array of numbers. Each element of the image defines a hypersurface in the Hough space. The hypersurfaces are determined by the shape to be identified in the image. As the mapping from the image space to the Hough space proceeds, these hypersurfaces are accumulated in the array. Certain locations are incremented more frequently and local maxima (clusters) are produced in the array. As an example, a straight line of an image is represented as a bundle of lines in the Hough space, each Hough space line representing one point of the image line; the common intersection point of the Hough space lines, the cluster, represents the straight line of the image. The coordinates of the clusters are the fundamental parameters for image analysis and image recognition. Traditional techniques extract these parameter values by methods such as smoothing and subsequent local maximum finding, cluster analysis or pattern matching which are all computationally expensive. The objective of the present disclosure is to describe a new efficient algorithm to identify clusters in the Hough space. In the new technique, assume a 2-dimensional Hough space of size M v N. Each point in the Hough space is represented by a value H(m,n), with m=0...M and n=0...N. In general, the clusters can be found by generating projection vectors (accumulating the H-values in rows or columns) and then finding the maxima in these projection vectors. However, merely adding up the array values row- or column-wise leads to a poor representation of clusters as maxima in the projection vectors. The technique proposed h...