Basis of the image search engine.

This blog as the name suggests is about image search engine & what we can think of to improve or innovate it.

Lately Google has improved its image search engine by going from the usual "text around image" to "actual image" search, basically it scans the zillions of images stored in WWW to correlate them with the text that you have inputted.

Well what I thought a few months back was have an image search engine which scans our given image and correlates it with the most similar ones. Basically its like,

Text search -> "the input is text"
Image search -> "the input should be an image!"

Yes it sounds totally weird and unnecessary but take a situation where you have a photo of an person/object but have no clue about its name/location/etc. So we can try to "google" it by saying "house" or "a long lost friend" or "river side hotel" etc. which will give you 0.00000000001% of finding the copy of it in some website.

Taking a practical view of this idea. Its like a face recognition(in case of a persons photo). Now that is already implemented but the application here is totally different and more useful. We can also incorporate the text around the image to verify the result. The minor tunings are many in this case.

Well this post is open for your suggestions.