Category Archives: images

Google experiments with Image Swirl

Having made Google Image Options (including colour) and Similar Images available as part of their standard image search, Google are now playing around with Image Swirl in Google Labs. According to Google it “builds on new computer vision research to cluster similar images into representative groups in a fun, exploratory interface”. In practice it is a combination of similar images and the Wonderwheel.

One of my image test searches is Edvard Munch and Swirl came back with 12 thumbnails of stacked images (12 is the standard number of stacks) :

Click on a group of stacked  images and another set of images “swirl” into view in the form of the wonderwheel:

And you can keep on clicking on groups/stacks of images but still keep the “history” of your selections.

I was pleasantly surprised by the clustering or stacking of the images. I thought that by the time I had reached ‘level 3’ of my browsing each stack would be just different versions of the same image or images with similar colour composition. My Edvard Munch level 3 selection, however, came up with a selection of landscapes with different colours. They did, though, seem to have similar ‘patterns’, for examples paths or what could be interpreted as paths as a major component of the image.

Phil Bradley has also reviewed Google Swirl and comments “Bing is going to have their work cut out to try and catch up.” Far too polite, Phil. I’d say “Bing, eat your heart out!”

Google Swirl looks very promising and I shall be monitoring its progress with interest.

Searching for images by colour

This is not a frequently asked question on my workshops but when it is raised by one of the participants it generates a great deal of interest amongst the rest. So far I have come across three that I would recommend trying.

The first is Exalead’s Chromatik, which is part of the Exalead Labs experimental area. This enables you to search a selection of Flickr images by colour and optionally by keyword. You first select one or more colours or hues from a palette which are added to a bar below the palette. You can adjust the proportions of  the colours in the photos by moving the separators between the colours in the bar. Luminosity can be toggled between bright and dark, and saturation between colourful and grey levels. The last option in the list is to search for specific images using keywords (I assume this searches the titles, tags and descriptions associated with the Flickr images). The implication is that once you have selected your colours you can then limit your search to particular objects. In practice, if you search for colour followed by keyword, Chromatik ignores your colour choices and searches only on your keywords. If, for example, you want to search for apples of a particular colour you must first search on apples and then pick your colours.

It pays to keep the number of colour choices to two or three, even if you require very specific colours, as this will give you a wider range of images to choose from. When the thumbnails are displayed you can hover over the best match and select “show images with same colors”. Click on an image and it is displayed full size, but in order to see further information about it you have to right click and select properties. This will give you a URL for the original image on Flickr but only for the image itself. It does not take you to the “full” Flickr page for the photo, which means that  you cannot check ownership and copyright.

The second tool is Multicolr Search Lab from Idée Inc. This uses “10 million of the most “interesting” Creative Commons images on Flickr”. As with Chromatik you select colours from a palette. You can select up to ten colours and click on the same colour several times if you wish to increase its prominence in the photo. Unfortunately there is no keyword search. On the plus side, if you find an image you like simply click on the image to go straight to its page on Flickr where you can double check the copyright situation.

And of course there is Google’s image search. Carry out a search on your keywords in Google images and above the results there is an option to select a colour. There are only twelve colours from which to choose and you can only select one but it works well enough. If you want to search only Creative Commons images then carry out the first stage of your search in the Advanced Image Search screen and select the appropriate option from the Usage Rights menu.

Getty Images wins £2,000 over unauthorised web use of photo

If  nothing else, this is a good example of what can happen if you fail to check the rights associated with photographs and images found on the web, and then use them for your own commercial purposes. Some people refuse to accept that just because an image is on the web does not mean that you can do what you want with it. If you do not want to pay for an image, there are plenty of sources of public domain and Creative Commons images but even then there may be conditions and some restrictions on their use (see my posting Free-to-use images might not be).

In this particular case, a removals firm used a Getty photograph on their web site without paying for it. Getty found out about it because it uses tracking technology to detect the unauthorised use of pictures.

You have been warned!

Free-to-use images might not be

You may have already read that Google now includes a creative commons license filter option in its Advanced Image search screen. Creative Commons is a series of licenses that can be applied to a variety of works such as images, video and PowerPoint presentations and they specify what you can and cannot do with those works. Information on the licenses can be found on the Creative Commons web site at

Google does not use the CC terminology but has instead generated a pull down menu with the options: labelled for reuse, labelled for commercial reuse, labelled for reuse for modification,  and labelled for commercial reuse for modification.


There is another option at the top of the list that is the default: not filtered by license. I had to think twice about this one because my first thoughts were that this was for public domain images. It is not. The “not filtered” option is all images. I ran the license options past a few people over the past week and they all immediately assumed that the default option is for images that you can use as you want.  A couple, though, then asked how “labelled for reuse” differed from this and then they became totally confused by the whole thing. To make it worse,  the licenses as listed by Google do not cover all the possible CC license conditions, for example attribution and share alike. So once you have done your search you still have to check the full license for the image that you wish to use. Furthermore, very few people are aware that you have to cite the license and any attribution as requested by the author.

Google says in its help files:

“By returning these search results, Google isn’t making any representation that the linked content is actually or lawfully offered under a Creative Commons license. It’s up to you to verify the terms under which the content is made available and to make your own assessment as to whether these terms are lawfully applied to the content.”

The accuracy and validity of the Google implied license was raised recently in The Register: The tragedy of the Creative Commons . It comments:

“Since there’s no guarantee that the licence really allows you to use the photo as claimed, then the publisher (amateur or professional) must still perform the due diligence they had to anyway. So it’s safer (and quicker) not to use it at all.”

I disagree with that: I recommend using it as a first level filter but then check with the original web site regarding the details of the license. At least you won’t be spending hours wading through “all rights reserved” images.

If you do use the license filter you will notice that many of the photos come from Flickr, which is owned by Yahoo!. Yahoo! has had a Creative Commons filter on its Image Advanced Search screen for a long time but only on the US site, not the UK. A far better way of searching CC Yahoo images is to go straight into Flickr at  This gives you a description of the different licenses and you can search images assigned that license. This assumes, of course, that the person who has uploaded the image is the owner of that image and there are stories that this is not always so. But how paranoid do you have to be? With respect to Flickr my approach is to take the photographer’s word for it unless there are serious inconsistencies in their photostream, for example the  meta data associated with the photos suggests that they were in Armenia, New Zealand and Peru on the same day!

So where do you go for images that really are free to use.  There is a trick you can use in Google  to pull up just public domain images. Carry out your search on the standard Image search screen and when the results come up add


to the end of the string in your br0wser address bar, and press enter. (Thanks to Barry Schwartz at Search Engine Land for this tip) . The test searches I have tried so far come up with photos from NASA, US government sites and Wikimedia Commons.  NASA is a safe bet for public domain images as are US government web pages, although there are a few exceptions but these are clearly labelled with any copyright restrictions.. A recent spat between Wikimedia Commons and the UK’s National Portrait Gallery  – National Portrait Gallery bitchslaps Wikipedia: Hands off our photos! – has thrown suspicion on the validity of CC and public domain licenses attached to its photos. This appears to have been an isolated incident, though, and the high resolution images have now been removed if you are accessing the site from the UK.

Another source of public domain images is MorgueFile, which is a small database of high resolution photos but you may have to play around with your search terms before you find exactly what you want.

If you are looking for photos of buildings or locations in the UK then head straight for Geograph.  This aims to collect geographically representative photographs and information for every square kilometre of Great Britain and Ireland. Anyone can upload photos provided that they adhere to the guidelines and attach a Creative Commons Attribution-Share Alike 2.0 Generic license.

Geograph has saved me so much time. A few months ago I was trying to find a photo of the Great Expectations pub in Reading, Berkshire. Google, Yahoo and Live (now Bing) insisted on giving me photos of people reading a copy of  Charles Dickins’s Great Expectations  while sitting in a pub in Berkshire. The image I wanted was probably somewhere in the list but I was not prepared to trawl through hundreds of results to find it. I typed in Great Expectations Reading into Geograph and I was there in a couple of seconds. Brilliant!


If you are interested in finding out more about finding and using images head for JISC Digital Media – Still Images.

Geograph British Isles – photograph every grid square

Geograph British Isles, sponsored by the Ordnance Survey, aims to collect geographically representative photographs and information for every square kilometre of Great Britain and Ireland.

According to a recent article in the Daily Telegraph the three founders of the project – Paul Dixon, Gary Rogers and Barry Hunter – think of it  as a “modern Domesday Book”.  It was  started in February 2005 and  has apparently  built up a large following in Canada, New Zealand and Australia among people searching for pictures of their ancestors’ home towns.

You can find photos by browsing the map or by searching on keywords. The Advanced Search has options for grid reference, post code, place name and centre of county. You can specify the distance in kilometres (up to 10) from any of the above. The only option that did not work for me was post code.   Other advanced search criteria include contributor, a drop down list for category e.g. weir, date submitted and date taken. All submitters are required to assign a Creative Commons Attribution-ShareAlike licence to  their photographs and to allow the right to use the work commercially, so this site is ideal if you are looking for photos that you can use in a presentation.

Geograph search results for Caversham weir:

Geograph search results for Caversham weir

If you wish to submit a photo you first have to register.  Once you have logged in, you have to  give the grid reference for your photograph.  Somehow I missed the easy route the first time I tried this and spent ages trying to work out the exact OS reference. The easier and better way to do this is to use the Map Placename Application. This uses Google Maps and you simply home in on your location. The  grid reference is automatically generated and you move on to step 2, which is where you specify the image file you wish to upload. On the same screen you need to supply  the grid reference of the “primary photo subject”, but if you have used the map to find the location this should be filled in automatically. For the photographer position you just drag and drop the relevant circle onto the map.

The next step is to add a title, description, primary geographical category e.g. floods, AA phone box, bus stop.  The date taken is automatically extracted from the EXIF (Exchangeable image file format) information but this can be changed manually – useful  if you never got around to setting up the correct date and time on your camera! Finally, you have to confirm that you agree to the  Creative Commons rights assigned to your photo.

There is a lot on this site and it may not always be obvious to users how to  search, and for those who wish to submit photos it does require time and effort to upload images.  From the searchers’ point of view it is  worth it: the highly structured  records ensure that precision and relevance is high. Family snaps are rejected!  The FAQ clearly states that while people can be in the photo, they must not be the photo.  If you are  looking for photographs of locations in Great Britain and Ireland this is an excellent place to start, but be warned  – it is addictive.

TASI Tutorial: Internet for Image Searching   

TASI (Technical Advisory Service for Images) has launched a tutorial on Internet for Image Searching. This is a  free-to-use online tutorial that, according to the announcement, is “to assist staff and students within the education sphere in locating images for use in both teaching and learning”. I would recommend, though, that anyone who searches for images on the Internet either for personal or business use should work their way through this tutorial.

The tutorial starts off by demonstrating that just because you have found a photo or graphic via Google Images does not mean that it is free for you to use as you want, and it is not always easy to find copyright and ownership information.  Similarly it points out that although a Flickr photo may be given a Creative Commons licence It is not uncommon for people to upload images that are not their own and make them available under Creative Commons licences. And there is more than one form of Creative Commons: if you don’t know what they are follow this tutorial or go to

As well as taking you through the legal issues, copyright and licensing models the tutorial list sites that offer free photos, Creative Commons images, royalty free and commercial stock photos. Note that some of the free photo sites are for personal non-commercial use only and that fees apply if you wish to use them for business purposes.

The emphasis of the tutorial is on “finding copyright cleared images which are available free; facilitating quick, hassle-free access to a vast range of online photographs and other visual resources”.

An added bonus is that as you progress through the tutorial you can add sites to a free ‘link basket’, which can be saved, printed out or sent to you by email.

Live Search Images adds face search

First Exalead adds an option to limit your image search to faces, then Google, and now Live Search has joined the gang. In terms of ease of use, it is not as slick as Exalead’s but not quite as clunky as Google’s. You first of all carry out a search in Live Images and then add filter:face to your search search strategy or filter:portrait. If you want to look for black and white images you add filter:bw. At present you have to remember the commands but they say they are looking at how to make these features intuitively accessible through a drop-down menu or some other means.

On my image test searches on I cannot honestly say it was better or worse than Exalead or Google. None of them are perfect. They do remove most of the non-people images but all three also lose relevant faces and ‘portraits’.

Google image search looks for faces

Hot on the heels of Exalead’s new face recognition search option, Google has launched a similar feature. Unlike Exalead, which has a ‘Face’ option under ‘Narrow your search’, Google requires you to add &imgtype=face to the end of the URL of your results page. As Phil Bradley says in his blog “A simple button would suffice guys!”. Phil also reports that, although clunky, Google’s face search seems to return more and better results than Exalead’s. My own experience is variable: sometimes Exalead is better, sometimes Google. Which just goes to prove that you really do need to know your way around more than one search engine.

Exalead UK and new search features

Exalead now has a UK version of its search engine that includes an option to limit your search to UK pages only. Both and UK have added a Wikipedia search and options to limit your results to blogs or forums. Alternatively, you can choose to exclude those types of sites. The Wikipedia search includes a “Narrow your search” panel on the results page that lists ‘tags’ for categories, related terms, people, location and organizations.

Exalead Wikipedia search

Exalead has also launched a new version of its image search with over one billion images indexed. The new ‘Face’ filter enables you to narrow your search results to images containing faces. It is not a hundred per cent accurate and sometimes excludes images that are of faces and includes some in which there are no faces at all, but it is close enough. Other options include size of image, wallpapers, image colour, layout and file type.