Browse Prior Art Database

Simultaneous Huffman Decoding and Descaling for Decompression of Transform Encoded Signals

IP.com Disclosure Number: IPCOM000108193D
Original Publication Date: 1992-Apr-01
Included in the Prior Art Database: 2005-Mar-22
Document File: 1 page(s) / 54K

Publishing Venue

IBM

Related People

Feig, E: AUTHOR [+3]

Abstract

Transform coding of data, such as JPEG image compression coding, involves quantizing the transform coefficient and then encoding the results with some entropy coder, such as Huffman. Decoding is usually done by first Huffman decoding, then descaling, to yield approximates of the original transform values, then inverse transforming. This invention performs the Huffman decoding and descaling simultaneously.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 98% of the total text.

Simultaneous Huffman Decoding and Descaling for Decompression of Transform Encoded Signals

      Transform coding of data, such as JPEG image compression
coding, involves quantizing the transform coefficient and then
encoding the results with some entropy coder, such as Huffman.
Decoding is usually done by first Huffman decoding, then descaling,
to yield approximates of the original transform values, then inverse
transforming.  This invention performs the Huffman decoding and
descaling simultaneously.

      Typically, the final stage of Huffman decoding is done by table
lookup.  For example, when the JPEG algorithm is used, a Huffman
codeword is sent to indicate the "magnitude category" of the
coefficient (the length, in bits, of the absolute value of the
coefficient), and then several extra bits are sent which uniquely
determine the coefficient value.   The coefficient value may be
determined by a lookup step, where the index into the lookup table is
based on the magnitude category and the extra bits.

      This invention proposes to incorporate the scaling into the
Huffman table lookup.  The values in the lookup tables are the scaled
DCT coefficients.  Since different coefficients are scaled by
different constants, one will need several such lookup tables.  In
applications where one needs not too many coefficients, this will not
tax the decoding architecture so that the data handling penalty
outweighs the gains of not having to descale separately.

  ...