Cryptography Reference
In-Depth Information
704 pixel/line576 line/frame8 bit/byte3 byte/pixel25 frame/s
= 243.3
Mbit/s.
A two-hour movie of the above video signal would need 219 Gbytes of
storage, which is 51 times the capacity of Digital Versatile Disk (DVD). How-
ever, after compression using MPEG-2 video coding standard, the bit rate
of this video can be reduced to about 1.5 Mbits/s with no noticeable quality
degradations. This is equivalent to an compression ratio of approximately 160.
There exists a huge amount of redundancy in the uncompressed video signal.
2.1.1 Redundancies in Digital Video
It is known that there are very strong correlations between pixels, lines and
neighboring frames of a video signal. In a typical video signal, most parts
of the scenes change very slowly, or are still. This is especially true in the
background area. Only the foreground objects may experience large changes.
In order to effectively compress the video signal, it is important to understand
the various types of redundancies in the video signal.
Statistical Redundancy
Statistical redundancy is related to the similarity, correlation and predictabil-
ity of data. Digital video signal consists of still images that vary with the time,
and each of these images is called a frame. Thus depending on whether it lies
inside a frame or in between frames, statistical redundancy can be further
classified into spatial redundancy and temporal redundancy.
Spatial Redundancy: Spatial redundancy refers to the similarities within a
frame of the video signal. For example, the neighboring pixels in a frame
have a similar brightness, color and saturation. Typically we can use run-
length coding, predictive coding and/or transform coding to reduce the
spatial redundancy.
Temporal Redundancy: Often there are only small changes between adjacent
frames within a video sequence. The background may be fixed or only
change very slowly while only the objects in the foreground are moving.
Even when the camera pans over a scene, the change of the pixels in the
scene is uniform and can be predicted. Inter frame prediction and motion
compensation are an effective way of removing temporal redundancy.
Coding Redundancy
The processed video data, after predictive coding or transform coding, may
show a statistically skewed distribution that can be exploited for further data
compression. For example, after predictive coding, the error signal will show a
biased distribution where the majority of the prediction errors are either very
