Privacy-preserving Motion Detection for HEVC-compressed Surveillance Video

• DOI: 10.1145/3472669
• Publication Date: 2022-01-27T19:44:21Z
• Authors: Changming Liu; Xiaojing Ma; Sixing Cao; Jiayun Fu; Bin B. Zhu
• Keywords: Motion Detection; Quarter-pixel motion

In the cloud era, a large amount of data is uploaded to and processed by public clouds. The risk privacy leakage has become major concern for users. Cloud-based video surveillance requires motion detection, which may reveal people in video. Privacy-preserving allows detection while protecting privacy. existing scheme [ 25 ], designed detect on encrypted H.264-compressed videos, does not work well more advanced compression schemes such as HEVC. this article, we propose first method HEVC-compressed videos. It adopts novel approach that exploits inter-prediction reference relationships among coding blocks regions. partition pattern number bits each block used prior art are also help Spatial temporal consistency moving object Kalman filtering applied segment connected/merged regions, remove noise background motions, refine trajectories shapes detected objects. Experimental results indicate our achieves high recall, precision, F1-score videos both low resolutions with various scenes. similarly accuracy ] Our proposed incurs no bit-rate overhead very computational complexity encryption HEVC