A pornographic image and video filtering application using optimized nudity recognition and detection algorithm

College

College of Computer Studies

Department/Unit

Software Technology

Document Type

Conference Proceeding

Source Title

10th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment, and Management

Publication Date

2018

Abstract

The combination of multimedia technology and Internet provides an amiss channel for pornographic contents accessible by certain sensitive groups of people. Furthermore, the same channel provides the easiest medium to distribute illicit images and videos without an autonomous content supervision process. In this study, an application was developed grounded from a pixel-based approach and a skin tone detection filter to identify images and videos with a large skin color count and considered as pornographic in nature. With nudity detection algorithm as the foundation of the system, all multimedia files were preprocessed, segmented, and filtered to analyze skin- colored pixels by processing in YCbCr space and then classifying it as skin or non-skin pixels. Afterwards, the percentage of skin pixels relative to the size of the frames is calculated to be part of the mean baseline for nudity and non-nudity materials. Lastly, the application classifies the files as nude or not, and then filter it. The application was evaluated by supplying a dataset of 1,239 multimedia files (Images = 986; Videos = 253) collected from the Web. On the final testing set, the application obtained a precision of 90.33% and accuracy of 80.23% using the supplied dataset.

html

Disciplines

Computer Sciences

Keywords

Image processing—Digital techniques; Pornography

Upload File

wf_no

This document is currently not available here.

Share

COinS