Game Development Reference
In-Depth Information
must take a long time before tampering is noticed, particularly when one operator
monitors multiple cameras.
10.3 Intelligent Video Coding System
Although the existing video analysis system can provide the intelligent service for
users by integrating lots of computer vision techniques, they are usually with high
complexity and their performance is affected by the quality of compressed video.
In addition, for some intelligent operations, some basic operations must be carried
out, e.g., video decoding, object extraction and region of interesting extraction, which
consume lots of power in computation. Therefore, the intelligent video coding system
is proposed to integrate some basic operations into the compression stream in camera
coding process, which can provide an “intelligent” video stream. In addition, it should
provide a higher compression performance to reduce the bandwidth and storage
limitation for surveillance video.
10.3.1 Motivation of Intelligent Video Coding System
The past decades have seen an explosion in the use of video surveillance in many
places from public security to traffic control and in-home nanny monitoring. Nowa-
days, different numbers of cameras are connected to recording and display devices
via transmission networks, and constitute the so-called video surveillance systems
at different scales. According to a recent report by IDC (Ali and Aggarwal
2001
),
by 2020, as much as 5,800 exabytes of surveillance videos will be stored, transmit-
ted, and analyzed. Traditionally, the industry employs the coding standards such as
MPEG-4 and AVC/H.264 (Ayers and Shah
2001
) that are not originally designed
for generic videos to compress surveillance videos. If we still follow this technology
road map, in future several years, the growth rate of surveillance videos will be much
higher than the video compression rate that AVC/H.264 and even its next genera-
tion development, HEVC/H.265 (Comaniciu et al.
2003
), can achieve. As shown in
consequently presenting an unprecedented challenge for high-efficiency and low-
complexity surveillance video coding technology. It will bring out that there is not
enough space to store these video data, e.g., for common commercial application,
the surveillance video only can be stored for recent three months.
Besides, the resolution of surveillance cameras, the bandwidth of transmission net-
work, the capacity of storage systems, as well as the recognition capability of video
analysis software, all these may influence the performance of a video intelligent sys-
tem in various degrees. In general, the higher the resolution of cameras, the better the
quality of the captured videos or images, which in turn may enhance the visual expe-
rience in monitoring and improve the video analytics (e.g., face recognition). At the
