Pictures on the Internet have to look often. It is regrettable to note, but it is not always possible to go and take a picture. Yes, this is the easiest way, but often it is unacceptable. That is why a variety of Yandex come to our aid. Pictures, Guglokartinki and Flickers.
Yandex has an excellent feature - if you find a picture, next to it, it will show a list of sites where there are the same ones, as well as their sizes. Accordingly, you can choose the maximum possible size.
I myself use Flickr.com for two reasons. The first is a huge number of high resolution images. Second, I know English quite well, so I can make sensible requests.
Image search is not perfect on all services. And now I will explain why. For example, I need to find an image on which three people are leaning against a wall. And to be precise, the photo should be taken a little on top. And how do you order me to enter a request?
"Three sit in the corner"? "People sit wall"? Or maybe “A photo of three young men sitting leaning against a wall, view at an angle of 65 degrees, sepia”? As you understand, none of these requests gave me the necessary information on any of the three resources.
The result - you need to go and shoot, then process. And I
just need to make a sketch for the layout. Sketch! Another fact that the layout will be. To hold a photo session, not knowing whether it is useful, it is expensive.
So, why all this. In my opinion, the perfect image search is when there is an easel and a pencil in front of you (not on the table, in the browser). You take and sketch draw what you want to find. For example, like this:

you set tags, for example, "people", "wall", "sepia". Mark key elements of the picture, such as: Man, Man, Wall. The engine analyzes the entered data, including the figure, and offers the most similar images by composition. I spent exactly 3 minutes on this sketchy picture. I think it is less than the time spent on the selection of queries and browsing dozens of pages.
')
The perfect image search is by composition.I think that it is possible to realize this, we implemented the same “Faces” function in iPhoto'09!