Browse Prior Art Database

Variable Precision Walsh Hadamard Transform for Data Compression of Text

IP.com Disclosure Number: IPCOM000050880D
Original Publication Date: 1982-Dec-01
Included in the Prior Art Database: 2005-Feb-10
Document File: 3 page(s) / 27K

Publishing Venue

IBM

Related People

Lasher, D: AUTHOR [+2]

Abstract

This article describes a technique for processing a camera-captured text image to compress the amount of data that must be transmitted over a telecommunications link for reconstruction at a remote receiver without significant loss of quality. The image is divided into 8 x 8 pel (picture element) matrices, each of which is transformed using a Walsh-Hadamard transform. Background matrices are suppressed. Matrices with text can have different bit masks applied to them so that the amount of data sent is related to the complexity of the matrix.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 3

Variable Precision Walsh Hadamard Transform for Data Compression of Text

This article describes a technique for processing a camera-captured text image to compress the amount of data that must be transmitted over a telecommunications link for reconstruction at a remote receiver without significant loss of quality. The image is divided into 8 x 8 pel (picture element) matrices, each of which is transformed using a Walsh-Hadamard transform. Background matrices are suppressed. Matrices with text can have different bit masks applied to them so that the amount of data sent is related to the complexity of the matrix.

In carrying out the technique, a complete frame (e.g., 348 pels horizontally by 256 pels vertically) is captured and subdivided into eight pel by eight pel matrices. This results in 48 matrices horizontally repeated 32 times vertically. Each of the 48 matrices in a horizontal group is transformed using a Walsh- Hadamard transform. A mask is applied to each matrix to reduce the number of bits required to quantify the data in each element. A fixed mask may be used or, via program control, a unique mask can be dynamically chosen for each matrix to permit simple subpictures to use fewer bits and complex subpictures to use more bits.

The first element of each transformed matrix quantifies a background color of the original matrix. The other 63 elements of the transformed matrix are used to determine whether the original matrix contained text information. If the value of any of the 63 elements falls outside of certain limits, it is assumed that the original matrix contained text.

When a transformation has been completed and transmission is to begin, a 48-bit word (one bit per horizontal group) is first transmitted to identify the matrices containing background infor...