Wednesday, March 13, 2019

How Google Reverse Image Search Works? Explained - SingleWindowTech

Google Reverse Image Search

Reverse Image Search  - Definition

Most of us turn to Google to search for information and images, but sometimes you may want to look up an image you already have. Maybe you want to find out where it originated, expose it as a fake photo, or verify the claims of whoever posted it. Whatever your motivation, this is known as a reverse image search.

What is Search Google by Image?

The technology behind this is Reverse Image Search which has got a good base in current times in the market. Reverse image search is a Content-Based Image Retrieval (CBIR) query technique that involves providing the system with a sample image that it will then base its search upon; in terms of information retrieval, the sample image is what formulates a search query. In particular, the reverse image search is characterized by a lack of search terms. This effectively removes the need for a user to guess at keywords or terms that may or may not return a correct result. Reverse image search also allows users to discover content that is related to a specific sample image, popularity of an image, and discover manipulated versions and derivative works.

Google acquired Neven Vision in 2006, a key player in face and image recognition biometrics. This allows Google to recognize faces, places, objects, and automatically apply tags to all the images on the Internet.

How Google Reverse Image Search Works?

Google uses some combination of techniques and "ranks the results using a proprietary algorithm". The basic algorithm used is Auto-Encoders.
  • Whenever Google's bots (spider bots) finds an image while crawling the web, first that image could be re-sized to the desired dimension. (for example, a 1024*768 pixel image could be re-sized to a predetermined 32*32 pixel image like in the above case).
  • Then it could try and find this N-dimensional compression form for the re-sized image (N is 8 in the above example) using the Auto-Encoder network.
  • Since most of the images are colored, each image could be decomposed into R, G, and B color spaces and then finding the compressed form in each color space would be a much easier way to go.
  • So, Google has all of its images in the form of easily searchable N-Dimensional array format, i.e. now every image is just an N-Dimensional array where N is much smaller than the original dimensions of the image.
  • Now all that Google's search engine needs to do when someone asks for a visually similar image of her/his photo is, just like above, it finds the N-dimensional compression form of this new image and then it could find all the points that are nearest to it in the N-dimensional space and rank them according to the distance.
Next approach which can be used behind this is Visual Similarity Search, a set of algorithms are used to analyze image attributes such as color, shape, texture, luminosity, complexity, objects and regions. These attributes are stored for indexing and matching Google seems to use this in combination with keywords to refine searches on extremely large collections.

Feature Detection: There are three commonly used feature detection algorithms for matching image deformation such as blur, rotation, scale, and illumination change. 
They are SIFT, PCA-SIFT, and SURF.
  1. SIFT - Scale-invariant feature transform is slow and not good at illumination changes, while it is invariant to rotation, scale changes, and affine transformations.
  2. SURF - Speeded Up Robust Features is fast and has good performance as the same as SIFT, but it is not stable to the rotation and illumination changes.
  3. PCA-SIFT - Principal Components Analysis-SIFT is the best but it has problems with image blur.

How to do google reverse image search in Android?

To upload an image to perform the reverse image search on, you need to get to the desktop version of Google’s image search. Here’s how:
  1. Go to images.google.com in your browser.
  2. You want the desktop version, so you’ll need to request it. In Chrome, tap the three dots at the top right to open the More menu.
  3. Tick the Desktop site option.
  4. Tap on the wee camera icon to get the option to upload an image.

How to do google reverse image search on the iPhone? 

Well, it has an equivalent track to follow in iOS, but the route is slightly different. Here’s how:
  1. Go to images.google.com in your browser.
  2. You want the desktop version, so you’ll need to request it. In Safari, tap the Share icon, with the arrow pointing up, to open the share menu.
  3. Scroll along the bottom options until you find Request Desktop Site and tap it.
  4. Tap on the tiny camera icon to get the option to upload an image.

You might also like:

No comments:
Write comments

Featured Post

Xiaomi MI Soundbar - Specifications, Price and Review

Image Courtesy - Xiaomi Before talking about the sound, let's first talk about the picture. The modern day televisions are impro...