Browse Prior Art Database

Using Implicit Phase to Allow Progressive Coding to Run at Higher Speeds

IP.com Disclosure Number: IPCOM000122630D
Original Publication Date: 1991-Dec-01
Included in the Prior Art Database: 2005-Apr-04
Document File: 4 page(s) / 171K

Publishing Venue

IBM

Related People

Equitz, W: AUTHOR [+2]

Abstract

Disclosed is an improvement for progressive arithmetic coding of images. Progressive arithmetic coding of images is a method for efficiently compressing an image in a way which allows lower resolution versions of the image to be quickly decoded before the final higher resolution version. The innovation described involves reformulating the way that a lower resolution image is handled, allowing one to include spatial phase information in model templates implicitly, rather than explicitly. The new method has several advantages, including reducing the time required to process an image, introducing illegal model context states which can be used for other purposes, and sometimes increasing the compression performance.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 32% of the total text.

Using Implicit Phase to Allow Progressive Coding to Run at Higher
Speeds

      Disclosed is an improvement for progressive arithmetic
coding of images.  Progressive arithmetic coding of images is a
method for efficiently compressing an image in a way which allows
lower resolution versions of the image to be quickly decoded before
the final higher resolution version. The innovation described
involves reformulating the way that a lower resolution image is
handled, allowing one to include spatial phase information in model
templates implicitly, rather than explicitly.  The new method has
several advantages, including reducing the time required to process
an image, introducing illegal model context states which can be used
for other purposes, and sometimes increasing the compression
performance.

      An example of the state of the art (1) for handling reference
images while doing progressive arithmetic coding is shown in Example
1.  Here, a sample 4 x 6 high-resolution image (with each "X"
representing a pixel) is shown with its corresponding 2 x 3
low-resolution image (with each letter in the range A-F representing
a pixel).  The low-resolution image has one-fourth as many pixels as
does the high- resolution image.  The high-resolution image can be
imagined to be broken up into 2 x 2 blocks of pixels, and the
"spatial phase" or "phase" of a high-resolution pixel identifies the
orientation of that pixel within such a block (top-left,
bottom-right, etc.).
                          XXXXXX
                          XXXXXX         ABC
                          XXXXXX         DEF
                          XXXXXX
                          Hi-res        Lo-res
Example 1.  A 4 x 6 hi-res image and corresponding 2 x 3 lo-res
image.

      When progressively encoding an image one would typically use a
model template, such as the one shown in Example 2.  Here the "?"
represents the pixel being encoded and "@" represents the
low-resolution pixel associated with the pixel being encoded.  The
pixel represented by a "@" is a part of the model template.
                            0
                           000          @0
                          00?           00
                         Hi-res        Lo-res
Example 2.  A differential layer model template.

      The values of the pixels corresponding to the differential
layer model template are explicitly combined with the two bits of
spatial phase information for the pixel encoded to form a "context",
such as shown in Example 3.
                        PPLLLLHHHHHH
Example 3.  Bits in a "context".

      The context bits marked "H" correspond to high-resolution model
template pixe...