1-2
Introduction
Final Rev F
Copyright 1996 by LSI Logic Corporation. All rights reserved.
Resolutions are about 352 pixels horizontally up to about 288 lines
vertically for MPEG-1 and 720 x 576 for MPEG-2 (main prole/main
level). The L64005 is capable of resolutions up to 720 x 576 for either
MPEG-1 or MPEG-2.
Display frame rates range from 24 to 30 frames per second.
1.1.1
Video Encoding
For a video signal to be compressed, it must be sampled, digitized, and
converted to luminance and color difference signals (Y, Cr, Cb). The
MPEG standard stipulates that the luminance component (Y) be sampled
with respect to the color difference signals (Cr and Cb) by a ratio of 4:1.
That is, for every four samples of Y, there is to be one sub-sample each
of Cr and Cb, because the human eye is much more sensitive to lumi-
nance (brightness) components than to color components. Video sam-
pling takes place in both the vertical and horizontal directions. Once
video is sampled, it is reformatted, if necessary, into a non-interlaced sig-
nal. An interlaced signal contains only part of the picture content (every
other horizontal line, for example) for each complete display scan.
The encoder must also choose which picture type to use. A picture cor-
responds to a single frame of motion video, or to a movie frame. There
are three picture types:
Intracoded pictures (I-pictures) are coded without reference to any
other pictures.
Predictive-coded pictures (P-pictures) are coded using motion-
compensated prediction from the past I or P reference pictures.
Bidirectionally predictive-coded pictures (B-pictures) are coded using
motion compensation from a previous and a future I or P-picture.
A typical coding scheme contains a mixture of I, P, and B-pictures. Typ-
ically, an I-picture may occur every half a second, to give reasonably fast
random access, with two B-pictures inserted between each pair of I- or
P-pictures.
Once the picture types have been dened, the encoder must estimate
motion vectors for each
macroblock in the picture. A macroblock consists
of a 16-pixel by 16-line section of luminance component and two spatially
corresponding 8-pixel by 8-line sections, one for each chrominance com-
ponent.