IT-Based Tools to Support New Product Design: A Case Study of a Design Consultancy Firm - Information Technology and Product Development - page 73

major free Web-based image database, delivering relevant pictures as thumbnails
and links to the websites related to these images.
The algorithms used for the image search engine integration are commercially
available. They might also look for clues from the image s context (for example,
words or phrases that are close to the image) or the meta-information found at the
top of the HTML coding. Analysis of an image s text and context is used to exclude
images. This is the same principle used by Web image filters.

4.4.5 Image Collection Manager

The image collection manager is an advanced tool featuring a number of innova-
tive content-based search techniques. The core of the tool is based on a series of
prototype C++ and Python-based searching algorithms. These make use of a multi-
resolution wavelet decomposition algorithm to query image databases. Image search
queries may be expressed in three different ways, either by using a rough sketch
scanned or drawn directly into the computer using a sketching tool, or by using a
similar image, or by using meta-data key words.
By clicking on the “Draw” tab the user can draw a brief sketch of the image
or images that they are looking for. The color and shape are also used to narrow
the search parameters. For example, by drawing a simple orange triangle using the
sketch tool, the image collection tool returns ten best matches along with a score
indicating how similar they are to the actual sketch, in this case, a picture of the
pyramids (66.3% matched to original) from a collection of 404 images in total
(see Fig. 4.6).

Fig. 4.6 Interface of the image collection manager