Pornographic Image Recognition Software Employed by Chinese Government

These days online discussion grows regarding Chinese government’s deployment of Green Dam Censorware System, a piece of software aiming at blocking adult content or political content-sensitive websites. It is mandated that every computer that will be sold in China after July 1st, 2009 needs to install this software.

The part I am interested in is its pornographic image recognition engine, developed by Jinhui Technologies in Zhengzhou. Pornographic image identification is an interest computer vision application. How does this work? 

Intuitively, the Pornographic image filter detects large area of skin area based on image color information, and then removes images with large face regions using OpenCV’s face detector. More details are disclosed in “A Technical Analysis of the ‘Green Dam-Youth Escort’ Software”,

The process of image detection begins when visual data is obtaining as the image is in queue to be screened, first normalizing the image’s size, then separating areas of skin tone from those without skin tone; analysis of the relationship between areas of skin tone is followed by removal of noises, then extraction of the area’s characteristics, which are then input into a trained SVM (note: Support Vector Machine) classifier. Once the image has been deemed pornographic it is sent to a human face detector; if a human face is not the primary component, the image is then classified as pornography.

The method of detecting pornographic image based on color and face detector is efficient and effective. There are related patent and technical paper along this line. Chinese government nowadays has a more strict regulations on intellectual properties of software and patents. I believe that the Jinhui company has disclosed the potential IP issues to Chinese government, or worked out a workaround.

One voice rises because Green Dam’s use of OpenCV prior to version 3.174 did not include the required license. According to a study by University of Michigan, this problem was addressed in the 3.174 filter update by adding the required OpenCV license.

This “green” activity will have great social benefits to Chinese youth.


  • By Mosalam, July 15, 2009 @ 12:05 pm

    The technical paper you’ve referred to, though interesting but it doesn’t include enough empirical tests. For example what happens with a medical website that has some photos of skin diseases? Is the Chinese government going to block that kind of websites and make trouble for researchers?

  • By Harry Wang, July 20, 2009 @ 10:02 pm


    Thanks for your insight. You bring up a good point. Yes, based on skin color and face detection heuristics, the problem of Pornographic Image Recognition is by no means solved. However, considering the problem is to block adult website for young kids, the technology in the paper (H. Rowley, Y. Jing, S. Baluja , Large Scale Image-Based Adult Content Filtering. in Int. Conf. on Computer Vision Theory & Applications 2006) should work to some extent because there is small probability that a young kid needs to do research on skin disease.

    Nevertheless, it is always good to extend the work to handle this problem in future computer vision research.

  • By Jai Pillai, July 5, 2010 @ 9:52 am

    The method seems to be relying too much on the skin detection part. I heard that it doesnt work for all races as the skin color changes. They should also start detecting key words and using them along with the vision techniques.

  • By mark, October 12, 2010 @ 4:12 am

    Doing my research I find one amazing free to download book about Computer Vision.
    This book presents research trends on computer vision, especially on application of robotics, and on advanced approaches for computer vision (such as omnidirectional vision).
    The contents of this book allow the reader to know more technical aspects and applications of computer vision.

    The intended audience is anyone who wishes to become familiar with the latest research work on computer vision, especially its applications on robots.
    This book features representative work on the computer vision, and it puts more focus on robotics vision and omnidirectional vision.
    This is the link where you can find it:

  • By Full-Learning Blogspot, February 5, 2011 @ 6:01 am

    Good news, Atleast now we can save our children

  • By Adrian, January 22, 2014 @ 8:07 pm

    These methods inherently rely on skin detection techniques, which are inherently unstable in sub-optimal lighting conditions. And of course, depending on the race of the person, the skin detection model would have to be changed. These types of works need further empirical evaluation and more focus on actually “modeling” the type of data they want to flag.

Other Links to this Post

RSS feed for comments on this post. TrackBack URI

Leave a comment