Web image search engine evaluation

Abstract EN

Image search engine is a specialized search engine for retrieving images.

It is rapidly growing with the growth in digital images on the web.

The essential role of image search engine is to retrieve image results that are relevant to the users and provide them more reliable and comfortable services.

The main challenge for image search engine is to retrieve images with content matching much more than text matching.

This paper investigates the performance of three image search engines.

The paper is realized in two phases.

In the first phase, three image search engines, namely, Google, Yahoo and Ask, are selected.

Then, twenty queries are determined from two researches.

Each query is run on each image search engine separately and first twenty images retrieved are classified as being relevant or non-relevant.

Afterwards, precision ratios are calculated at various cut-off points.

In the second phase, image features, namely, color and shape are determined.

Some of the images retrieved from first phase are analyzed according to their features.

The results of first phase indicated that Google has the best overall retrieval effectiveness with 95 % precision ratio, followed by Yahoo with 91 %, and Ask at the last with 83.7 %.

Furthermore, the results from second phase showed that two images with similar color histograms can possess different content.

As well as for shape feature, the number of edges is not efficient to identify relevant images.