Beyond Google Images Part 2

This series of posts is adapted from the Library Information Seminar “Beyond Google Images”, which looks at issues and problems faced when sourcing images from the web and gives examples of a range of useful image resources available to College staff and students.

For Part 1: Introduction, click here.

Part 2: Google Images and other search engines describes the features of image search engines, and considers how they address the criteria for image searching discussed in Part 1: findability, quality and copyright.


Part 2: Google Images and other search engines

google-images2

What are image search engines?

Search engines are the major tools for finding information on the web. There are many examples out there, but the big three are Google, Yahoo and Microsoft’s Bing (formerly Live Search) – all of which have image search features. There are also a number of specialist search engines that search solely for images. While the specifics vary, they all work on similar principles. You enter a text query describing the image you want and are presented with a display of results in the form of thumbnail images. Clicking on a thumbnail retrieves the image itself or its web page. You are often given a choice of related searches and the option of filtering your results according to size, colour, layout etc. For reasons of space, this article will focus on Google, but if you are interested there is a good review of the major examples available on this site.

Search engines index the web using programs that “crawl” or scan millions of web pages, using complex algorithms to extract relevant information such as titles, common keywords etc. This information is stored and used to retrieve results that (hopefully) match the queries entered by searchers. Even with images, the search engine relies mostly on the textual content of the page containing an image – which can cause problems if the text is not relevant to the image’s subject matter or visual content. To address this, search providers are exploring visual searching, in which an image is used as the query and results retrieved by matching colours and shapes. Such efforts remain experimental and imperfect, and for the time being text searching remains the norm. If you are interested in visual searching, try a search on Google Images and look for thumbnails with a “Find similar images” link below them. You’ll probably find the results are variable, to say the least!

How useful are image search engines for our needs?

So how well do search engines (specifically Google Images) address the three main problems facing the image searcher – findability, quality and copyright?

'Signpost, Hokuchin-dake' by shirokazan on Flickr

‘Signpost, Hokuchin-dake’ by shirokazan on Flickr.

Findability
As mentioned in Part 1, the World Wide Web is a very fragmented image source. Relevant imagery may be scattered across many different personal, commercial and academic websites: on blogs, photo sharing sites or in the collections of private or public galleries, to name just a few examples. The greatest strength of search engines, and particularly the big ones, is their reach. They allow us to find imagery from across the internet, and across national and cultural boundaries, very quickly because they index millions of web pages. This reach can work against us, however – the massive quantity of material indexed can make finding the right image like looking for the proverbial needle in a haystack unless we want something very specific. There are also large numbers of high quality images on the web which are invisible to search engines because they are held in databases and cannot be indexed by web-crawling programs.

Search engines have trouble with the central problem of image findability – the need to translate between the visual and the verbal. Google doesn’t know what an image is of – it relies on the text surrounding it, and as suggested above this isn’t always relevant. This can often lead to strange results if our search query describes the content or subject matter we require rather than an artist or a title. Things are complicated further by the fact that the same word can have different meanings in different contexts, but Google does not differentiate between these, e.g. “painting” as in “oil painting” or as in “painting and decorating”.

The way search engines present their results also limits our options for exploratory browsing, as opposed to quick and specific searching. We are usually presented with a series of thumbnails that match our original query, but we are given no context for the images, and they are not organised thematically or conceptually in a way that would allow us to explore related imagery or subjects easily. To change the direction of our search we have to think of new keywords and start again from scratch.

The upshot of all this is that while Google can be very efficient at finding a specific named work, person, place or thing, it is less useful for finding something very conceptual or with specific visual content.

'Blurry Bicycle Commute' by roland on Flickr

‘Blurry Bicycle Commute’ by roland on Flickr

Quality
Image quality is one of the biggest weaknesses of search engines. Many results are from personal web pages such as blogs, which require that images are in a compressed format such as JPEG which saves disc space but reduces quality. Such images are usually formatted for display in web browsers as opposed to printed reproduction, meaning that they are only a few hundred pixels long and wide. This can look poor in an essay or other assignment. If the person who posted an image scanned it from a book or an original work, there is no guarantee that the colour balance and quality of reproduction is accurate – not everyone is a scanning expert! Finally, because search engines index all sources of imagery indiscriminately, from gallery sites to collections of holiday snaps, images may be aesthetically mediocre or lack interest.

'copyleft' by eflon on Flickr.

‘copyleft’ by eflon on Flickr.

Copyright
The fact that an image is retrieved by a seach engine does not mean you are allowed to use it! Some people, businesses and institutions post images online but explicitly deny any right to reproduce them in any form. In other cases, images have been posted in flagrant breach of copyright, whether scanned from a book or simply copied from another website. As such, determining whether an image can be used legally is sometimes difficult. For this reason, most search engines feature prominent disclaimers that images “may be subject to copyright”. Clicking through to the site hosting the image may provide further information, but not always. If you are planning on reproducing or manipulating an image in any way beyond using a single copy for reference, you should exercise extreme caution when using search engines.

The following examples will hopefully illustrate the problems outlined above.

Example searches on Google Images

Click on a query to see the results in Google Images. These links will open in a new window.

Search 1: peter doig reflection what does your soul look like

This search demonstrates that Google Images can be useful if you have an exact named image in mind. But look at the results closely. Are these good quality images? Would they look good reproduced in an essay? Which one has the most accurate colour reproduction? How many of the websites have permission to reproduce the image? Do any of them give you permission to reproduce it?

Search 2: art nouveau

This search demonstrates some limitations when looking at a broader subject. The results include diverse examples of Art Nouveau design including graphics, typefaces, jewellery and architecture. Where can we explore from here? Are the images organised in a way that is useful to us? “Related searches” are available, but are these exhaustive? What are we looking at? How can we find out? Are these all “genuine” Art Nouveau designs? Are they significant or important works? Do the web pages themselves provide useful and authoritative information?

Search 3: painting interior people reading

This search tries to find images of paintings with specific visual content. How relevant are the results? Is there anything strange or irrelevant? Why has this happened?

Conclusions

While these are serious limitations, there is no question that search engines remain very useful sources of imagery. While in many cases a search engine may be your first port of call in an image search, it shouldn’t always be your last. The later episodes of this series will show you other resources that address the issues mentioned above, and explain why you should sometimes look beyond Google Images.

12. January 2010 by jfurnes1
Categories: e-resources, library information seminars | Tags: , , | Comments Off on Beyond Google Images Part 2

University of Edinburgh Privacy Policy and cookie information