Darren pointed to a new feature in Google images search letting you to filter results by color ! The results looks very accurate that you will be interested to know how google are indexing colors ? Are they using image recognition techniques to index every image in their database ? That will be crazy if you think about it, but the answer could be easily found in something launched about three years ago, an ajax game, letting everyone to tag images and get scored on that : The image labeler.

Because the results are very accurate and fast, it can only be something based on a real human indexing ! You can see below for example if you search for a car model, you can get the result you are looking for, but imagine you want to see only gray cars in results, the filter can really give you what you are looking for. While if you used "gray" as keyword in search you will get something else completely different.

I find the color filter implementation very successful. Image search is getting more and more powerful search features and I expect to see more amazing filtering options, who knows !