INTERNATIONAL ORGANISATION FOR STANDARDISATION

ORGANISATION INTERNATIONALE DE NORMALISATION

ISO/IEC JTC1/SC29/WG11

CODING OF MOVING PICTURES AND AUDIO

ISO/IEC JTC1/SC29/WG11/N11823

January 2011, Daegu, KR

Title:               Description of MPEG-7 Tools for Image Signature Description

Source            :           Video Subgroup

Authors:         Miroslaw Bober and Stavros Paschalakis

Status:             Approved

-----------------

In recent years people have been generating and distributing an ever increasing amount of image data. A recent survey of prominent web sites shows that Flickr has over 2 billion images, Photobucket has over 4 billion and Facebook has 1.7 billion. There are hundreds of billions of images on the Internet, and even users’ personal databases can have tens or hundreds of thousands of images. At the same time, there are few tools which one can use to efficiently identify or search for a specific image, possibly in an edited or modified form, either on the Internet or in one’s own personal collection. The MPEG-7 Image Signature Tools address this problem by providing an interoperable solution for image identification.

In contrast to previous MPEG-7 visual descriptors which were designed to provide access to similar content, the Image Signature is a content-based descriptor designed specifically for image identification, i.e. designed for the fast and robust identification of the same or modified image in web-scale or personal databases. Such a type of descriptor is also commonly known as a fingerprint and has a strong advantage over watermarking techniques in that it does not require any modification of the content and can be used readily with all existing content.

The applications for the MPEG-7 Image Signature are numerous. These include media usage monitoring, e.g. tracking and recording statistics such as distribution and frequency of content usage, web-page linking, e.g. using images to imply links between web-pages, as is currently done for text, rights management and monetization, e.g. detection of possible copyright infringement or content monetization online (for content owners) or identification of the copyright owner (for content consumers), and personal image collection management and de-duplication.

The Image Signature is the result of extensive collaborative effort within MPEG-7, aiming at the delivery of an optimised interoperable image identification solution. In order to achieve fast searching and high robustness, the Image Signature combines two complementary approaches in image representation: a global signature, where the signature is extracted from the entire image, and a local approach, where a set of local signatures are extracted at salient points in the image. In terms of content identification performance, the MPEG-7 evaluation process tested the robustness of the Image Signature to a wide range of common modifications, such as text/logo overlay, rotation, cropping, colour changes, etc., and achieved an overall success rate of ~99.29% at a false alarm rate of less than 0.05 parts-per-million for the global signature, and ~98.04% at a false alarm rate of less than 10 parts-per-million for the complete signature. These success rates are achieved with search speeds in the order of 80 million and 100,000 matches per second for the global and complete signatures respectively. The Image Signature is also extremely compact, at only 1024 bits per image for the global signature and up to 7424 bits for the complete signature.

The MPEG-7 Image Signature Tools comprise four amendments to the MPEG-7 standard, specifying the extraction, decoding and syntax of the Image Signature [1], providing a reference software implementation [2], specifying the conformance conditions and dataset [3], and describing the Image Signature matching procedure that was used in the MPEG-7 evaluation process. These resources will ease the development of systems that comply with standard, and it is anticipated that the Image Signature Tools will find wide adoption in image identification applications.

References

[1] ISO/IEC 15938-3:2002/AMD 3:2009, Information Technology – Multimedia content description interface – Part 3: Visual, Amendment 3: Image signature tools

[2] ISO/IEC 15938-6:2003/AMD 3:2010, Information Technology – Multimedia content description interface – Part 6: Reference software, Amendment 3: Reference software for image signature tools

[3] ISO/IEC 15938-7:2003/AMD 5:2010, Information Technology – Multimedia content description interface – Part 7: Conformance testing, Amendment 5: Conformance testing for image signature tools

[4] ISO/IEC 15938-8:2002/AMD 5:2010, Information Technology – Multimedia content description interface – Part 8: Extraction and matching of image signature tools