Live Demo | Info Request | Careers
 

Image-Filter, porn filtering

LTU Image-Filter is a ground-breaking image-based software that can automatically analyze the content of images and video streams.

This enables companies and organizations to perform :

  • Porn filtering / reliable filtering of objectionable visual content (e.g. X-rated)
  • Protection of sensitive visual assets
  • Customized automation of content monitoring processes

Image-Filterhas been designed for :

  • OEM/VAR integration with content filtering solutions & e-mail management software solutions for businesses and ISPs.
  • Direct online content monitoring for government and education organizations, and custom email management systems.

 

 

Features


 

  • Real-time image analysis and indexing
    System users can filter content in real-time with LTU's Image-Filter. First, the system analyzes an image or video dataflow, based on its visual features (shape, color, texture, etc). This dataflow is then translated into a digital signature (DNA), which is used to identify a variety of visual categories that must be monitored for protection purposes (i.e. porn filtering / adult content or sensitive corporate documents).
  • Severity Cursor
    System administrators can define customized acceptance thresholds for the filtering system.
  • Score computation
    Image-Filter returns a customized score and/or alert for each image, allowing an administrator to take action accordingly (block, quarantine, deliver, etc).
  • Blacklist / White list customization System administrators can modify image blacklists and white lists in order to block or grant access to particular images for certain users.

 

Architecture


  • Scalability; compliant with high availability clustered architecture.
  • Computation cache (for large volumes of images).
  • Parallelization on multiprocessors architecture.
  • Flexible architecture allows for light software updates.

 

System requirements


 

Server: Image-Filter is a server-side software optimized for high volumes and scalability.

  • Intel based processors (1 GHz Pentium or better).
  • RAM: 256 MB per processor (minimum).
  • Current OS: Linux (RedHat 7.1 or higher, SuSe 7

Client-side Platform: Client side platforms are unrestricted and currently include Windows (XP, 2000 & NT 4.0), AIX (4.2.1 or higher), and SUN Solaris 8 or higher.
Compatible mail platforms are Microsoft Exchange, Lotus Domino and SendMail.

 

Integration


  • Standard API & SDK (Java, Perl, C/C++, PHP).
  • Easy integration (HTTP/XML).
  • Complementary with existing filtering systems (text, URL, mail).
  • Open API allows remote administration via HTT

 

Supported content and visual categories


  • Over 20 supported formats, including jpeg, gif, bmp, psd, etc.
  • Color photography, cartoons, clip art, buttons, banners.
  • Several recognition categories are currently available.( Examples: Pornography, Cartoons, Website graphics, Technical drawings)
  • Expansion of existing recognition categories and development of new categories are customizable on demand.

 

References


 

"The tests that we have conducted show that LTU filtering solution can really bring an added-value to an integrated solution as the one we intend to develop in the context of the NetProtect project."
NetProtect Report - a European Prototype of Internet Access Filtering.
Download the Report (PDF)

"The Software Image-Filter of LTU was tested with a variety of different test data comprising adult image content as well as "harmless" content such as images of company logos, people, coin collections etc. The software produced good results."
JOANNEUM RESEARCH - the largest independent R&D institution in Austria.

 

More about porn filtering


Filtering images using text is rarely effective. For instance, an image entitled “Image001.gif” or “teddybear.jpg” provides no information about the content of the image. Thus, to filter an image effectively, one must view the actual content.  

As such, the founders of LTU, have spent over seven years at MIT, Oxford and INRIA designing a computer software that automatically (and quickly) sees and analyzes the content of images to provide information regarding its “appropriateness”. Providing the world access to this revolutionary technology is precisely the mission of our company.

Image-Filter is the image recognition and filtering system that has been designed, developed and commercialized by LTU.

Key applications include filtering of harmful content on the internet (e.g. personal pages, user-generated content, auction sites, search engines) or on telcos devices (MMS) and the monitoring of email attachments (non business-related content, sensitive proprietary content).

Generally speaking, LTU software has been designed to fit smoothly into existing monitoring tools. Image-Filter is capable of analyzing visual content in any format (gif, jpg, bmp, mpg, avi…), and originating from any source (a web page, email attachment, hard disk, video stream, scanner, etc).

Image-Filter visual thesaurus can be extended to include new categories according to the client’s needs. Other key functionalities include Black and White lists: the administrator can add images to black and white lists to block or grant access to particular images (e.g. proprietary content, sensitive content).

 

Acceptance levels


 

Within 0.6 second (average CPU time) of inputting an image into the system, Image-Filter outputs a score for the submitted image. In the particular case of pornography (one of Image-FilterTM’s core filtering features), a score of “0” corresponds to a “totally harmless image” and a score of “100” corresponds to “highly pornographic content”. This score is generally reported to a human administrator who uses the analysis to determine how his/her organization will respond to the content (e.g. the administrator may only want to see the images which score higher than “50”).

footer bar
LTU Technologies: Making sense of visual content