How to find a picture on the Internet?
How many times did it happen that, having found a photo on your disk, you could not remember where it came from? Or they could not understand at all what is depicted on it? And how many pictures have to shovel in search of “something like this, but a little different?” Not to mention the perennial need to find the original of the photo in its miniature. Now this problem will help solve the new search service from Google. In addition to image search, Google now allows you to search for images and photos. In this case, the existing image is used as a sample for the search. If you are working in the Google Chrome browser, you can simply drag the image into the search box, and Google will display information about all sites that have such a picture. In addition, Google will offer the most likely name for this image. Owners of browsers Chrome and Firefox can also get plug-ins that allow you to search by right-clicking on the original image. It must be said that this idea is not revolutionary. Ever since 2008, there is TinEye – a search engine specializing in searching images on the Internet. TinEye became famous for the fact that with her help the identity of an unknown soldier was established from a photograph. The system managed to find a file with a complete list of people depicted in the photo. But, despite the rather respectable age of the site TinEye, the young service from Google already now has more powerful search algorithms and a larger image database. In addition, unlike TinEye, Google offers not only completely matching, but also similar images, which undoubtedly will appeal to web-designers. Technically, the image search process is almost the same as a regular text search. Google’s robots bypass the World Wide Web and index images and photos encountered in their path. Of course, building an index of images is much more complicated than indexing text documents — for each picture, characteristic elements are selected, which are then described in a certain way, sorted and stored in a database. Thus, each image gets its own unique key. During the search, the sample downloaded by the user is also represented as a key, according to which a comparison with the keys already in the database is made. Naturally, complete coincidence is possible only if the photos are absolutely identical. However, the description method used makes it possible to recognize similar images regardless of the file format, size, color and other characteristics that do not affect the meaning of the image. Recognition is even possible in terms of the original image. Of course, all this is possible only if there is a similar picture on the Internet and a description of it. So for personal photos while you can not be afraid. But if you like to load pages on social networks with your photos, you can not worry – soon Google will know about you. For example, the system identified Andrei Arshavin and Madonna with absolute confidence. The only thing that could confuse Google was the mirrored image. But the first step is the hardest, search algorithms will be improved.