Browse Prior Art Database

Querying Image Databases Using Computed Texture Features

IP.com Disclosure Number: IPCOM000105067D
Original Publication Date: 1993-Jun-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 2 page(s) / 50K

Publishing Venue

IBM

Related People

Dom, B: AUTHOR [+5]

Abstract

Described is a method for improving a way in which computed image texture can be used to enhance image database querying. See Fig. 1. for a block diagram.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 65% of the total text.

Querying Image Databases Using Computed Texture Features

      Described is a method for improving a way in which computed
image texture can be used to enhance image database querying.  See
Fig. 1.  for a block diagram.

1.  Step 1 - For each image in the database, identify interesting
    regions of the image, for which one may possibly want to query
    later.  These regions may be the entire image, or some subset of
    the image.  This identification of regions could be done
    manually, it could use automatic object segmentation, or it might
    involve a regular partitioning of the image (such as into N X N
    blocks).
2.  Step 2 - For every identified region, calculate specific computed
    "texture measures", describing visual properties of the region
    [*].
3.  Step 3 - When making a query, describe the images one wishes to
    retrieve in terms of "texture measures".  This can be done in a
    number of ways.  For example, if one has a particular image (or
    region of an image) and one wants to find all images in the
    database with (or containing) similar texture, then one can
    simply calculate the texture features of the sample image.  This
    way, a small set of images can serve as example textures for the
    purpose of defining a query.  Alternatively, one could
    artificially generate texture features to be matched against.
    Artificial texture feature generation can be done by...