SEMANTIC
COMPRESSION OF VIDEOS IN CCTV
Sergey. A. Kuzmin
SUAI,
Saint-Petersburg,
Russia
Received
April 9, 2012
Abstract. In the first
part of article the possibilities of video compression by downscaling of
parameters (bit depth, resolution) are studied and rejected. In the second part
of the article the approach to compression, which utilizes saliencies of signal,
is studied. Algorithms of object detection and semantic compression are
described. The key feature of compression algorithm is supplement of video
stream by information about parameters of objects. Final part consists of
experimental study of compression ratio of proposed algorithm.
Keywords: object detection,
parameters of object, video compression, JPEG, metadata.