How does Image Search Work?

As with many technical questions I have a problem in the distinction between "How does something work?" and "What does it do?", this is an example.

If you make an Internet search you will get a lot of the latter explainations, especially from sources such a Quora. While it is important to know how to make an Image Search such explainations don't give much of an idea how it is done. In fact, Google are unlikely to give away their secrets.

Why I am interested in this

Having suspected that many Impressions on this site are from those making searches for names of local residents I decided to replace an HTML re-write of a page that I had found with an image of it. The theory was that the page, and hence the image, could not be found by making a textual search.

I would be very surprised if the users who make Impressions on my site are capable of making a Reverse Image Search.

Top

How Google explain how a Reverse Image Search works

The quotes are from the smartframe.io blog post by Matt Golowczynski - 13/11/2020

" As Google explains in this video, an image that’s used for a reverse search is analyzed for distinctive points, colors, lines and textures. The query that’s generated from this analysis is then compared with billions of possible images that Google is able to access. Should an exact match be found, Google will display this, indicating the different sizes of images it has been able to find. Related images, along with relevant web pages, will also typically be shown. "

The Youtube video - 22 sec - I don't really think this really helps:

The Google Reverse Image Search Explained

" First, it may bring up many similar images that may have some kind of useful or interesting information attached to them. An example of this would be when you’ve captured an image of a plant or insect whose name you’re unsure of. If you use this for a reverse image search, and a number of similar images showing this same subject are returned, there’s a good chance that a few will bear the same name in their descriptions. "

Facial Recognition

The explanations above do not mention facial recognition (well on my first read).

The Process of performing a Reverse Image Search

As I understand it there are a couple of ways that you can do this. These are described on the blog post that I refer to.

The process involves the submission of an image to Google by means of a Context Menu option (or possibly by the copy and pasting of an image directly into the Google search for Images). The use of the Context menu does appear to upload a copy of the image the search for the image name does a different search.

The Image Search seems to work quite well but it use to determine whether the image has been "boot-legged" remains to be seen.

How does Image Search Work?

Links