Browse Prior Art Database

Image Analysis Algorithm for Detecting Defects in Repetitive Shapes in Single Images

IP.com Disclosure Number: IPCOM000109471D
Original Publication Date: 1992-Aug-01
Included in the Prior Art Database: 2005-Mar-24
Document File: 1 page(s) / 74K

Publishing Venue

IBM

Related People

Kirtley, KB: AUTHOR

Abstract

Disclosed is an image analysis algorithm and software implementation for inspecting single gray scale images which contain repetitive two-dimensional shapes. Figure 1 shows an image with four equivalent two-dimensional shapes. Figure 2 shows a similar image with a defective shape. The algorithm is able to inspect Figure 2 and identify the defective shape without any knowledge of Figure 1. The two-dimensional shapes must be the darkest or brightest objects in the image but may exist on a nonuniform background. In addition, the good two-dimensional shapes must be similar though the shapes that are defective can vary in both size and shape. It is presumed that within an image only a small number of the two-dimensional shapes are defective.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 83% of the total text.

Image Analysis Algorithm for Detecting Defects in Repetitive Shapes in Single Images

      Disclosed is an image analysis algorithm and software
implementation for inspecting single gray scale images which contain
repetitive two-dimensional shapes.  Figure 1 shows an image with four
equivalent two-dimensional shapes. Figure 2 shows a similar image
with a defective shape.  The algorithm is able to inspect Figure 2
and identify the defective shape without any knowledge of Figure 1.
The two-dimensional shapes must be the darkest or brightest objects
in the image but may exist on a nonuniform background.  In addition,
the good two-dimensional shapes must be similar though the shapes
that are defective can vary in both size and shape.  It is presumed
that within an image only a small number of the two-dimensional
shapes are defective.

      The algorithm is based on a template matching technique in
which each of the two-dimensional shapes is compared to an ideal
template.  While the size and shape of the two-dimensional shapes may
vary from image to image, within a given image, the twodimensional
shapes should be equivalent.  To account for variations between
images, the ideal template is created separately for each image.

      In order to create an ideal template within an image, a window
is placed around each two-dimensional shape.  The gray scale values
in these windows are added and then averaged on a pixel by pixel
basis to create an average image of t...