We know Google as a well advanced search engine and retrieves in maximum all the information needed. Here we use certain key words to retrieve information about them. Such as… if we want to know about Apple iPhone, then we simply type Apple iPhone in search option and we get the websites which contain information about that so called Apple iPhones. And we want any images related to the keyword, and then we use Google Images and several other search engines like photo bucket which intake input keywords in the form of alphabets or words or sentences.
But… have you ever wondered why only alphabets or words are only used for searching??!! If you have thought that images may also help in such searching instead of “keywords”, then I must say guys… this article is really interesting to you…
.JPG FILES IN PLACE OF KEYWORDS:
By seeing the side heading you may be imagining the things like when you have got a beautiful picture without any relevant information of it, but you wanted to know about that. Let’s say you have great scenery in your desktop and wanted to learn anything about it but don’t have any idea where that place is from. Then it’s highly impossible for any one to search entire internet to learn about that place. Instead, if you have a search engine which accepts the scenery of your choice as input and retrieves the information about it, will it not be wonderful??!!
If your answer is yes it is… then we really are lucky enough to have such search engines with nearly 50% efficiency in retrieving the comparative images or information from their own data base or from the whole internet.
In such search engines we use images as input and get the comparative images as outputs.
HOW DOES THESE SEARCH ENGINES WORK? :
These search engines work by 3major rules. They are comparing the input image with the images of their database using
- Texture and
Using these 3 major factors of any image, these search engines compare the input image and the images from their database.
Implies that, first the search engine reads the color of the input image and retrieves the images of the same or approximate color components from its database. By this, number of images retrieved will be lesser then the total number of images in the database.
Later, these images are again compared with the texture of input image. Every image contains its own texture as we know. Here texture implies the particular information about the spatial arrangement of color or intensities in an image or selected region of an image. By using that information images from the database are further filtered by comparing the texture qualities between input and the rest images.
In the third case, the remaining images are further filtered by comparing the intensities of input image and database images. The interpretation of an intensity image depends strongly on the characteristics of the camera called the camera parameters. Here intensity may also be referred to as the crowd present in the image or may be the density of the image also.
By writing the efficient algorithms on these 3major factors, search engines are continuing their trails. But frankly to say, none of them have succeeded in their work. At maximum the efficiency percentage is 40% and rest of all cases end in failure.
But never leave the hope for the bright future. Revolution in search engines still awaits and we students are the persons who are going to create such miracles.
PREDICTIONS ON THESE KINDS OF IMAGE SEARCH ENGINES:
The important and key draw back of these search engines is efficiency. If you give salman khan as input, the output may be Mahesh babu depending upon the color, texture and intensity of the input image.
In some cases (according to my notice), these search engines use the name of the input image as a keyword and tries to find the images with relative names from its database. That means, if you give wanted poster from bollywood, it may show gopi chand’s wanted poster from tollywood.
Such limitations can be overcome by efficient algorithms which read the human’s face as well. In future, search engines may come which read the humans present in the image and compare those human resemblances in the images from their database. This kind of search engine definitely brings up some positive result to the above limitations and may become legend of its kind.
Gopichand ( MGIT ECE 3rd year)