Skin detection and lightweight encryption for privacy protection in real-time surveillance applications
Imtiaz, Muhammad Babar
Ashgar, Mamoona Naveed
MetadataShow full item record
An individual’s privacy protection is the concerning issue in surveillance videos. Existing research work for individual’s identification on the bases of their skin detection is focused either on different human skin detection techniques, or on protection. This research paper considers both lines of research and proposes a hybrid scheme for human skin detection and protection by utilizing color information in dynamically varying illumination and environmental conditions. For the purpose, the dynamic and explicit skin detection approaches are implemented simultaneously considering the multi-color-space i.e. RGB, perceptual (HSV) and orthogonal (YCbCr) color-spaces and then the human skin is detected by the proposed Combined Threshold Rule (CTR) based segmentation by considering the advantages of three multi-color-spaces. The comparative qualitative and quantitative detection results with average 93.73% accuracy imply that the proposed scheme gain considerable accuracy without incurring the training cost. Secondly, once skin detection has been performed, the detected skin pixels (including false positives) are encrypted with state-of-the-art Advance Encryption Standard with Cipher Feedback Mode (AES-CFB) rather than applying selective encryption on complete video. The proposed scheme preserves the behavior of the subjects within the video, hence can be useful for further image processing and behavior analysis and if required can be decrypted by the authorized user. The experimental results show that average encryption time is 8.268 sec and Encryption Space Ratio (ESR) with an average 7.25 % for HD cricket video (1280 x 720 pixels/frame) strongly imply that to protect a person within a video, the method of encrypting skin detection is preferable. Thirdly, performance comparison between the proposed method in term of Correct Detection Rate (CDR) with an average 91.5%, RGB with 85.86%, HSV with 80.93% and YCbCr with an average 84.8% imply that proposed method has high potential to detect the skin accurately. Furthermore, the security analysis performed confirms that proposed scheme could be a suitable choice for real-time surveillance applications working on resource constrained devices.
The following license files are associated with this item: