Page 110 - C:\Users\msi\OneDrive\Documents\Flip PDF Corporate Edition\E-Commerce\

The MPEG video compression standard provides two basic schemes: discrete-

transform-based compression for the reduction of' spatial redundancy and block-
based motion compensation for the reduction of temporal (motion) redundancy.

During the initial stages of DCT compression, both the full motion MPEG and
still image JPEG algorithms are essentially identical. First an image is converted
to the YUVcolor space (a luminance/chrominance color space similar to that used
for color television). The pixel data is then fed into a discrete cosine transform,

which creates a scalar quantization (a two dimensional array representing various
frequency ranges represented in the image) of the pixel data. Following
quantization, a number of compression algorithms are applied, including run-

length and Huffman encoding. For full motion video (MPEG I and 2), several
more levels of block based motioncompensated techniques are applied to reduce
temporal redundancy with both causal and noncausal coding to further reduce

spatial redundancy.

The MPEG algorithm for spatial reduction is lossy and is defined as a hybrid

which employs motion compensation, forward discrete cosine transforms (DCF),
a uniform quantizer, and Huffman coding. Block-based motion compensation is
utilized for reducing temporal redundancy (i.e. to reduce the amount of data
needed to represent each picture in a video sequence). Motion-compensated

reduction is a key feature of MPEG.

Moving Picture Types

Moving pictures consist of sequences of video pictures or t1'ame'S that are played
back a fixed number of frames per second. To achieve the requirement of random
access, a set of pictures can be defined to form a group of pictures (GOP)

consisting of one or more of the following three types of pictures. Intra pictures
(1)
Unidirectionally predicted pictures (U) Bidirectiomflly predicted

pictures (B) A Gap consists of consecutive pictures that begin with an
intrapeptide. The intrapeptide is coded without any reference to any other picture
in the group.
Predated pictures are coded with a reference to a past picture, either an

intrapeptide or a unidirectionally predicted picture. Bidirectionally predicted
picture is never used as references Motion Compensation for Coding MPEG.