PixMatch - Idée Inc. - The Visual Search Company

PixMatch automated image matching

PixMatch is an API-based, automated image matching solution. Given an image to look for, it locates identical or modified images within or between large scale image collections.

What can PixMatch do?

PixMatch is a general image matching engine that allows you to perform large scale image comparisons for a variety of tasks:

Consolidate duplicate and modified images within an image collection
Compare incoming images against your image blacklist
Identify incoming duplicates before they are added to a collection
Identify modified versions of original images
Help comply with copyright law
Moderate user-generated content
Filter and classify images

How does PixMatch work?

PixMatch does not use file metadata or keywords to detect image matches. It uses specialized digital image fingerprinting techniques to identify image matches despite resizing, cropping, rotation, flips or occlusion. Unlike hash algorithm-based image matching, PixMatch can deal with a broad range of image modifications and rarely produces false-positive matches.

PixMatch functions in a live online environment, so it can index images as they are submitted, and identify matches on-the-fly in less than a second.

About the API

PixMatch is delivered as a hosted Web Services API. It runs over HTTP using a REST protocol and JSON formatted responses. The API is very straight forward to use and is completely independent of whatever hardware or software your systems are built with. REST/JSON bindings are available in all major programming languages.

The major functions supported by the API are:

SEARCH: Input is an image to search for. The response is a list of matching images in your collection.
ADD: Input is an image. The image will be fingerprinted and added to your collection index.
DELETE: Input is an image ID. The image is removed from your collection index.

Requirements and specifications

PixMatch is a completely hosted solution–you need only to integrate the API with your existing software and send us an image collection to be fingerprinted. Once we receive the images, they are stored on our secure servers and PixMatch is ready to go. Only you will have access to your image collection.

Updates to your image collection are sent to us via HTTP. An image can be indexed and added to your collection in a fraction of a second.

PixMatch typically supports image collection sizes from half a million to tens of millions of images, but can work with very small collections or be scaled for billions of images. PixMatch can cope with millions of queries per day, and uses the same robust engine as TinEye web search.

Getting PixMatch results

Query images are submitted via the 'SEARCH' function of the API. Typical search time is less than one second.

Image match results are returned by the API, and each image is given a confidence score between 0 and 100. A score between 90 and 100 is considered excellent. Lower scores can still be matches, but images will have been modified in some way. You may set a threshold to filter out images with low scores.