Digital video fingerprint


The current media industry becomes more complex and different. Each year media volumes increase, and it becomes harder and harder to manage it. Media data goes through from the source to a client via a big quantity of equipment and potentially are subject to be distorted or modified. At the same time, the number of channels and delivery ways is growing, and this is makes a problem for monitoring. This leads to the system high cost buy and run.

The above-mentioned problems may be solved by a technology named as video fingerprinting.

What is the digital video fingerprints?

The digital video fingerprints is a technology which allows to identify and code the content in the same way as a human perception system does. At the same time the video content analysis is performed and certain unique characteristics outlined . These unique characteristics are called as Digital Fingerprints. As a human finger prints which are characterized as lines, arches of the fingertips of a certain human. The received fingerprints are stored in a database and at any time may be used to identify video (files or streaming) at different stages of its lifetime.

Difference from Watermarking technology

One of the characteristics of the video fingerprint compared to Watermarking technology is that it does not require any content modification and, therefore, any hardware on the side of the content source. In the Table below some key differences are presented.


The digital fingerprints should satisfy the following requirements for the successful usage with real tasks:

  • must be stable to different types of distortions and noise, format change and compression, different types of visual distortions, to temporary and spatial transformations etc.
  • must be effective in a sense of that it should not be complicated for calculations and require significant computational power.
  • must be effective in terms quick search over a big database.

Technical specification of the developed system

Fingerprints formation speed
To build the digital video fingerprints we use our own product named as TAPe-technology. For each frame of video sequence a set of characteristic is formed (digital print) identifying the frame. At this point two actions performed - integration (initial description formation) and spatial differentiation (analysis). This is allows us getting a high speed of construction - about 0.8ms per frame (for one core Intel Core 3GHz with Mpeg decoding). Moreover, as a result of the 1st operation is the full noise killing. The latter becomes as an element of image and in some cases helps to outline informative image elements.
Fingerprint size
A set of characteristics is very small and takes about 17 bytes per frame and building of digital prints for video raw allows lessen its volume even more. Therefore, our technology makes it possible to do 24 hours of video to be fit about 25 Mb (recorded with the speed of 50 frames per second)!
Using TAPe-technology allows getting invariant nature to all types of broadcasted noise and distortions (frequent to both analogue and digital signals) such as: change of color, brightness, shifts, macroblocks appearance etc.
Comparison and video search by its prints
For comparison and search of targeted video according to database we use innovated algorithm that identifies "visual similarity" of frames over digital prints as a measure of distance between them. The lesser the distance - the higher "visual similarity". The algorithm differs with its high operational speed and accuracy. So to find a video in one hour of video it is needed just 0.3 ms. Alfa and beta mistakes comprise 0.0165 and 0.0040 correspondingly. I.e. error level of the algorithm is very low. To be noted that for reaching such results the video length should be 5 sec and more but the minimum video length is 0.6 s!
The developed technology allows getting compact and at the same time stable to any types of distortions video fingerprints regardless the format and type of initial video stream. It is also does not require any specific hardware and allows building, comparing and searching of video fingerprints in the real-time mode with high accuracy using standard equipment. All this gives possibilities to solve effectively a large scope of tasks for media asset management, its protection and monitoring.

Made on