Browse Prior Art Database

METHOD FOR AUTOMATICALLY GENERATING GRAY TEMPLATES

IP.com Disclosure Number: IPCOM000027843D
Original Publication Date: 2000-Feb-29
Included in the Prior Art Database: 2004-Apr-09
Document File: 4 page(s) / 163K

Publishing Venue

Xerox Disclosure Journal

Abstract

Disclosed is an automatic system to design loose gray templates for line width tagging using a tree-structured classifier. It is necessary to tag appropriate pixels to receive more or less ink to control line width growth in a rendering system. Determination of sufficient templates by inspection is tedious and time-consuming. It is also difficult to design the templates simultaneously for multiple color separations and different rendering algorithms. Scanned lines with accompanying noise further complicate the design process. Consequently, the templates may not fully represent the variety of images encountered in a production system and yield tagging errors.

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 54% of the total text.

Page 1 of 4

XEROX DISCLOSURE JOURNAL

METHOD FOR AUTOMATICALLY GENERATING GRAY TEMPLATES John C. Handley

Proposed Classification
U. S. C1. 399/411 Int. C1. G03g 15/00

FIG. 7

24

30

FlG.2

Xerox Disclosure Journal - Vol. 25, No. 1 JanuaryiFebruary 2000 55

[This page contains 1 picture or other non-text object]

Page 2 of 4

METHOD FOR AUTOMATICALLY GENERATING GRAY TEMPLATES (Cont'd)

Disclosed is an automatic system to design loose gray templates for line width tagging using a tree-structured classifier.

It is necessary to tag appropriate pixels to receive more or less ink to control line width growth in a rendering system. Determination of sufficient templates by inspection is tedious and time-consuming. It is also difficult to design the templates simultaneously for multiple color separations and different rendering algorithms. Scanned lines with accompanying noise further complicate the design process. Consequently, the templates may not fully represent the variety of images encountered in a production system and yield tagging errors.

Line width control requires that gray lines of various widths be adjusted so that the printed line widths are correct. Different line widths require different adjustments, sometimes thinning and at other times, thickening. The actual operation requires knowledge of the line width. The affected pixels are those on the edges and white pixels adjacent to the edge. One can also consider comer pixels or any pixels where the image needs to be adjusted before printing.

Figure 1 illustrates an 11 x 11 window 10 having a portion of a gray image containing lines one pixel wide. A pixel 12 having a value of 15 is a gray edge pixel from a width 1 line. It may need to be adjusted before rendering. How it is adjusted depends on the width of the line to which it belongs and it must be tagged appropriately. One can use its context to decide whether the pixel is a gray edge pixel of line width 1.

The heart of the system is a classifier, which must be estimated from a training set. The classifier takes the pixel values surrounding given pixel 12 and estimates its class, say one of (G1 ,. . .,G5, Wl,.. .,WS}, where Gi means a gray edge pixel of a line of width i and Wi means a white border pixel of a line. The training set for the classifier can consist of analytic or scanned line data. Typically one would generate lines at given widths in various directions. A separate image for each line width would make establishing ground- truth easier. A training vector consists of pixel values surrounding a given pixel, plus the pixel's true class. In Fig.1, a training vector consists of the 121 values in the 11 x 11 window 10 plus the class label G1. A tree structured-classifier can be estimated using one of several popular methods, including Programs for Machine Learning C4.5 and Classijication and Regression Trees (CART), or can be built from scratch from the extensive literature. Regardless of which estimator is used, the result is a binary...