Browse Prior Art Database

Progressive Coding With Discrete Cosine Transforms

IP.com Disclosure Number: IPCOM000121057D
Original Publication Date: 1991-Jul-01
Included in the Prior Art Database: 2005-Apr-03
Document File: 1 page(s) / 57K

Publishing Venue

IBM

Related People

Feig, E: AUTHOR

Abstract

Progressive transmission of images using or Discrete Cosine Transform (DCT) techniques reproduce an image in various stages, the clarity of the image improving with each stage. The early images are corrupted by block artifacts. These are inherent in these methods, where the sizes of the various progressively reconstructed images are identical.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 60% of the total text.

Progressive Coding With Discrete Cosine Transforms

      Progressive transmission of images using or Discrete
Cosine Transform (DCT) techniques reproduce an image in various
stages, the clarity of the image improving with each stage. The early
images are corrupted by block artifacts.  These are inherent in these
methods, where the sizes of the various progressively reconstructed
images are identical.

      For various browse applications, images of varying sizes may be
appropriate.  Thus one can incorporate ideas of subband coding and
DCT coding into our progressive scheme. A preferred embodiment
entails partitioning the image to 8 x 8 blocks, and taking the 8 x 8
quantized DCT of each block. The quantization is a function of the
desired quality and average desired compression rate.  The output of
each block is sectioned into four subblocks:  the top-left 4 x 4
subblock, labeled LL (low-low), the top-right 4 x 4 subblock, labeled
LH (low- high), the bottom-left 4 x 4 subblock, labeled HL
(high-low), and the bottom-right 4 x 4 subblock, labeled HH
(high-high).

      The LL subblock may be, in turn, subband coded again; the
remaining three subblocks are coded with any of the good lossless
coding techniques used in image processing, such as Huffman coding or
arithmetic coding (including Q-coding), utilizing a zig-zag pattern
as described in the JPEG standards draft proposal, or any other
scanning order.

      The LL subblocks may be used to create an ima...