Browse Prior Art Database

Reconstructive Inverse Quantization of Motion Picture Expert Group Data

IP.com Disclosure Number: IPCOM000117474D
Original Publication Date: 1996-Mar-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 6 page(s) / 180K

Publishing Venue

IBM

Related People

Carr, JD: AUTHOR

Abstract

Disclosed is a macro circuit design for the inverse quantization of Motion Picture Expert Group (MPEG) data, within a reconstruction pipeline. After searching a reference picture for a best match of data to be encoded, the resultant encoded best match data must be reconstructed to determine how closely it resembles the source.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 33% of the total text.

Reconstructive Inverse Quantization of Motion Picture Expert Group
Data

      Disclosed is a macro circuit design for the inverse
quantization of Motion Picture Expert Group (MPEG) data, within a
reconstruction pipeline.  After searching a reference picture for a
best match of data to be encoded, the resultant encoded best match
data must be reconstructed to determine how closely it resembles the
source.

      This circuit accepts the quantization parameters applied in
encoding the data, optimizes the process of inverse quantization, and
pipelines the resultant block data to the Inverse Discrete Cosine
Transform (IDCT) macro.

      The Reconstructive Inverse Quantization (RIQ) process begins
with the setup of parameters, that remain constant, at least for the
duration of a macroblock of data.  These parameters are:  intra dc
precision, stepsize, chroma format, MPEG1 or MPEG2, and quantization
matrix values.

      Within a macroblock of data, which is a 16x16 section of the
picture, there are six or eight blocks of data per macroblock.  A
block is an 8x8 matrix of coefficients, or sixty four values.  If the
chroma format is 4:2:0, then there are four blocks of luminance and
two blocks of chrominance within the macroblock as shown in Fig. 1.
If the chroma format is 4:2:2, then there are four blocks of
luminance and four blocks of chrominance within the macroblock as
shown in Fig. 1 below.
            4:2:0                             4:2:2
   y1    y2      u1      v1          y1    y2      u1      v1
   (0)   (1)     (4)     (5)         (0)   (1)     (4)     (5)
   y3    y4                          y3    y4      u2      v2
   (2)   (3)                         (2)   (3)     (6)     (7)
       y         cr      cb              y         cr      cb
  luminance    chrominance          luminance    chrominance
  Fig. 1.  Chroma Format - Blocks Per Macroblock

      There are two types of blocks, intra and non-intra (or
inter).  An intra block, consists of one DC coefficient (which is the
first coefficient in the block), and 63 AC coefficients.  A non-intra
block, consists of all AC coefficients.  The Inverse Quantization
unit must distinguish between and handle both these types of blocks.

      DC Coefficients

      When the block being processed is intra, the first coefficient
received is identified as a DC coefficient and is dequantized based
on the intra dc precision parameter, according to the following
table.
          intra_dc_precision      multiplier
                00                    8
                01                    4
                10       ...