Cryptography Reference
In-Depth Information
2.2 DCT Based Intra Frame Video Coding
A Digital Video Signal consists of a sequence of images. Video Coding has
much in common with still image compression. In general, video coding con-
sists of two parts, Still Image Compression Based Intra Coding is used to
remove the spatial redundancy within an frame, and motion prediction and
compensation based Inter coding is used to remove the Temporal Redundancy.
During the 1980s and 1990s, Discrete Cosine Transform (DCT) [3] which
is based coding algorithms and international standards were developed. They
alleviate storage and bandwidth limitations imposed by Digital Still Image
and Motion Video Applications. Many DCT-based standards are now widely
used and accepted worldwide. These include JPEG, H.26x, and MPEG [4].
In addition to being DCT-based, many processing functions and coding prin-
ciples are common to all these standards. The principles for image coding
in most of the standards can be summarized as follows: Divide the picture
into 88 blocks, Determine relevant picture information, Discard redundant
or insignificant information, and Encode relevant picture information using
the least number of bits. Common functions of the basic coding scheme are,
DCT, Quantization, Zigzag Scanning, Run Length Coding, and Entropy Cod-
ing. Fig. 2.2 shows a typical still image compression system. The rest of this
section includes a discussion of the various compression techniques used in
Fig. 2.2.
HVS
weighting
table
Huffman
or
arithmetic
Block
formatting
Zigzag
scanning
DCT
Quantizer
RLC
VLC
Input
Y/C
b
/C
r
Compressed
data stream
DC
DPCM
coding
Fig. 2.2.
DCT Based Image Compression Process.
2.2.1 Predictive Coding
Predictive coding [5] was known as early as 1955. In a typical image with
natural scene, there usually exist many large areas with uniform colors and
textures. Although the values of the neighboring pixels are not exactly the
same, their differences are small when compared to that of pixel. Big differ-
ences only happen along the edges and contribute to only a very small part to
