IBM Books

Image, Audio, and Video Extenders Administration and Programming

QBIC catalogs

A QBIC catalog is a set of files that hold data about the visual features of images. The Image Extender uses this data to search for images by content.

You create a QBIC catalog for each column of images in a user table that you want to make available for searching by content. When you create a QBIC catalog you identify the features for which you want the Image Extender to analyze, store, and later query data. You can also add or drop features from a QBIC catalog after the catalog is created.

A QBIC catalog can hold data for the following image features:

Average color
The sum of the color values for all pixels in an image divided by the number of pixels in the image. (A pixel

is the smallest element of an image that can be assigned color and intensity.) For example, if 50% of an image consists of blue pixels and the other 50% red pixels, the image has an average color value of purple. Average color is used to search for images that have a predominant color. If an image has a predominant color, the average color will be similar to the predominant color.

Histogram color
Measures the distribution of colors in an image against a spectrum of 64 colors. For each of the 64 colors, histogram color identifies the percentage of pixels in an image that have that color. For example, the histogram color of an image might be 40% white pixels, 50% blue, and 10% red; none of the pixels in the image have any of the remaining colors in the histogram spectrum. Histogram color is used to search for images that have a variety of colors.

Positional color
The average color value for the pixels in a specified area in an image. For example, the upper right-hand corner of an image might show a bright yellow sun; the positional color of this area of the image is bright yellow. Positional color is used to search for images that have a predominant color in a particular area.

Texture
Measures the coarseness, contrast, and directionality of an image. Coarseness indicates the size of repeating items in an image (for example, pebbles versus boulders). Contrast identifies the brightness variations in an image (light versus dark). Directionality indicates whether a direction predominates in an image (as in the vertical direction of a picket fence) or does not predominate (as in an image of sand). Texture is used to search for images that have a particular pattern.

To make an image available for searching by content, you catalog the image. When you catalog an image, the Image Extender analyzes the image, by computing the feature values for the image, and stores the values in a QBIC catalog.

When you search for an image by content, your query identifies one or more features for the search (such as average color), a source for each feature (such as an example image), and a target set of cataloged images. The Image Extender computes the feature value of the source and compares it to the cataloged feature values for the target images. It then computes a score that indicates how similar the feature values of the target images are to the source.

You can have the Image Extender return the images whose features are most similar to the source. The Image Extender will return the handle of each image and the image score. You can also have the Image Extender return only the score of a single image.


[ Top of Page | Previous Page | Next Page | Table of Contents | Index ]