Washington, May 28 : Researchers at the Massachusetts Institute of Technology (MIT) have discovered that just a few pixels of information is required to be able to identify the subject of an image.
Antonio Torralba, an assistant professor in the Computer Science and Artificial Intelligence Laboratory, said that his team's discovery might help in the advancement of online images' automated identification, and thereby enable computers to see like humans do.
He and his colleagues are currently trying to determine the smallest numerical representation that can be derived from an image, so that it will provide a useful indication of its content.
He said that such a short representation would be an important step toward making it possible to catalogue the billions of images on the Internet automatically, which in turn would provide a way to index pictures people download from digital cameras onto their computers, without having to go through and caption each one by hand.
Torralba even envisions that the technique may eventually give rise to robots with the ability to sense of the data coming from their cameras, and figure out where they are. "We're trying to find very short codes for images, so that if two images have a similar sequence (of numbers), they are probably similar--composed of roughly the same object, in roughly the same configuration," Science Daily quoted Torralba as saying.
The researcher says that where one image is identified with a caption or title, other images matching its numerical code would likely show the same object-like a car, tree, or person-and thus the name associated with one picture could be transferred to other images.
"With very large amounts of images, even relatively simple algorithms are able to perform fairly well" in identifying images this way, says Torralba.
During a study, he and his colleagues reduced images in various low levels in order to determine how many images at each level people could identify.
"We are able to recognize what is in images, even if the resolution is very low, because we know so much about images. The amount of information you need to identify most images is about 32 by 32," said Torralba, adding that by contrast, even the small "thumbnail" images shown in a Google search were typically 100 by 100.
The researcher, however, admits that his findings are still preliminary, and that problems associated with identifying the more-unusual subjects would always remain.
"There are many words you hear very often, but no matter how long you have been living, there will always be one that you haven't heard before. You always need to be able to understand (something new) from one example," he said.
He will present his findings this June in Alaska at a conference on Computer Vision and Pattern Recognition.