Content-Based Pornography Image Detection
碩士 === 義守大學 === 資訊工程學系 === 92 === The Internet is so popular that every one can easily get almost any kind of information from it. Of course, pornography image is included. How to properly prevent children from accessing these porn-images is deeply concerned by most of parents. This disse...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | zh-TW |
| Published: |
2004
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/18966804927023008443 |
| Summary: | 碩士 === 義守大學 === 資訊工程學系 === 92 === The Internet is so popular that every one can easily get almost any kind of information from it. Of course, pornography image is included. How to properly prevent children from accessing these porn-images is deeply concerned by most of parents.
This dissertation presents an efficient pornography image detection system that utilizes multiple low-level features. The skin regions of an input picture are first detected. Then color histogram, intensity moments of skin region and bit-slice edge moments within the skin region are calculated to form a feature vector. The adult image can be detected by comparing the feature vector with the feature database obtained offline. Experimental results indicate that the proposed system achieve detection rate of 90.8% with a false alarm rate of 9.45% on a test set of 1000 adult images and 10000 non-adult images. In addition, it is very computationally efficient and suitable for real time applications.
|
|---|