Page 111 - C:\Users\msi\OneDrive\Documents\Flip PDF Corporate Edition\E-Commerce\
P. 111

Macro Blocks

               For  the  video coding algorithm  recommended by  CCITT, CIF and  QCIF are
               divided into a hierarchical block structure consisting of pictures, groups of blocks

               (GOBs), Macro Blocks (MBs), and blocks. Each picture frame is divided into 16
               x 16 blocks. Each Macroblock is composed of four 8 x 8 (Y) luminance blocks
               and two 8 x 8 (Cb and Cn) chrominance blocks. This set of six blocks, called a
               macroblock; is the basic hierarchical component used for achieved a high level

               of compression.

               Motion compensation

               Motion compensation is the basis for most compression algorithms for visual
               telephony and full-motion video. Motion compensation assumes that the current
               picture is some translation of a previous picture. This creates the opportunity for

               using prediction and interpolation. Prediction requires only the current frame and
               the reference frame.
               Based on motion vectors values generated, the prediction approach attempts to

               find the relative new position of the object and confirms it by comparing some
               block  exhaustively.  In  the  interpolation  approach,  the  motion  vectors  are
               generated in relation to two reference frames, one from the past and the next
               predicted frame

               The best-matching blocks in both reference frames are searched, and the average
               is taken as the position of the block in the current frame. The motion vectors for
               the two reference, frames are averaged.


               Picture Coding Method
               In this coding method, motion compensation is applied bidirectionally. In MPEG

               terminology, the motion-compensated units are called macro blocks (MBs).

               MBs  are  16  x  16  blocks  that  contain  a  number  of  8  x  8  luminance  and
               chrominance  blocks.  Each  16  x  16  macro  block  can  be  of  type  intrapeptide,
               forward-predicted, backward predicted, or average.


               MPEG Encoder
               Figure  below  shows  the  architecture  of  an  MPEG  encoder.  It  contains  DCT

               quantizer, Huffman coder and Motion compensation. These represent the key
               modules in the encoder Architecture of MPEG Encoder:
   106   107   108   109   110   111   112   113   114   115   116