Browse Prior Art Database

Inverse Quantizer Matrix Compression

IP.com Disclosure Number: IPCOM000117368D
Original Publication Date: 1996-Feb-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 2 page(s) / 75K

Publishing Venue

IBM

Related People

Ngai, AY: AUTHOR [+2]

Abstract

The Inverse Quantizer Matrix Compression is a scheme used in the Quantizer logic for MPEG II Encoder Chip. The compression scheme reduces the number of bits needed to represent and store the inverse quantizer matrix value. The scheme's intention is to reduce cell count and, therefore, save chip real estate.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Inverse Quantizer Matrix Compression

      The Inverse Quantizer Matrix Compression is a scheme used in
the Quantizer logic for MPEG II Encoder Chip.  The compression scheme
reduces the number of bits needed to represent and store the inverse
quantizer matrix value.  The scheme's intention is to reduce cell
count and, therefore, save chip real estate.

      The quantizer matrix is an 8x8 array of individual quantization
factors that are used to quantize the corresponding AC coefficient
from each 8x8 block of the macroblocks.  The AC coefficient is
divided by the quantization factor which ranges from +1 to +256.  The
Quantizer receives the matrix table prior to processing a picture and
stores the table for the duration of the process.  Depending on the
processing parameters, there could be two tables, for Intr and
non-Intra, or for tables, for Intra-luminance, non-Intr-luminance,
Intra-chrominance, and non-Intra-chrominance.

      Inverse Quantizer Matrix - Implementing a divide in hardware
requires a lot of cycles while a multiply can be done in one cycle.
To simplify the Quantizer's hardware implementation of the MPEG
algorithm, multiply the AC coefficient by one over the quantization
factor instead of dividing by the straight value.  Using the inverse,
saves cycles, but the number of bits required for a binary
representation of the quantizer matrix is increased.

      Compression Scheme - The absolute value of 1 to 256, can be
represented in binary within nine bits.  Representing the inverse of
2 to 256, which is all fraction, takes at least 16 bits in binary
using rounding or truncation.  Since using the inverse quantizer
matrix instead of the straight value almost doubles the size of
registers and GRAMs for handling the quantization factor, ways to
decrease or compress the value are investigated.

      The significant bits of a binary fraction start from the left
most position and move to the right adding leading zeros as the size
of...