Browse Prior Art Database

Improved Pel Selection for Grey Scale Representation

IP.com Disclosure Number: IPCOM000041433D
Original Publication Date: 1984-Jan-01
Included in the Prior Art Database: 2005-Feb-02
Document File: 3 page(s) / 27K

Publishing Venue

IBM

Related People

Haigh, DC: AUTHOR

Abstract

The technique described in this article addresses the problem of how to choose which of the picture elements (pels) in a rectangle to make black, having previously determined how many should be black. A technique that is widely used to represent grey scales with pels that can only be either black or white is to divide the image into equal sized rectangles (typically 3 to 5 pels per side) and then to make an appropriate proportion of the number of pels inside each rectangle black, the appropriate proportion of black and white pels being determined by the level of grey that it is desired to represent. If the size of each rectangle is m X n pels, then it is possible to represent (m X n)+1 different grey levels, ranging from completely white to completely black.

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 52% of the total text.

Page 1 of 3

Improved Pel Selection for Grey Scale Representation

The technique described in this article addresses the problem of how to choose which of the picture elements (pels) in a rectangle to make black, having previously determined how many should be black. A technique that is widely used to represent grey scales with pels that can only be either black or white is to divide the image into equal sized rectangles (typically 3 to 5 pels per side) and then to make an appropriate proportion of the number of pels inside each rectangle black, the appropriate proportion of black and white pels being determined by the level of grey that it is desired to represent. If the size of each rectangle is m X n pels, then it is possible to represent (m X n)+1 different grey levels, ranging from completely white to completely black. The grey level to be represented can be obtained by suitably averaging the original grey levels of the pels in each rectangle and then choosing the most appropriate representable grey level. Three different techniques have been used to solve this problem. The simplest is to use a fixed pattern of black pels for each representable grey level, the patterns being chosen to minimize fringing effects which might occur in large areas of uniform greyness. A popular improvement of this technique is to use 'ordered dither' which causes the pels to be chosen pseudo-randomly. Thirdly, the original grey levels can be modulated by some function, such as a sinusoid, and the result compared against a fixed threshold. These techniques have the undesirable effect of blurring or destroying fine detail in the image; for example, a black line one pel wide which crosses a rectangle is almost certain to be fragmented. This is why fine text processed by such techniques becomes illegible. The technique proposed by this invention preserves as much detail as possible, subject to the constraint of representing the chosen grey level. The proposed technique is as follows. Having decided to make B pels black, where 0 < B < (m X n), refer to the original grey levels of each pel, choose the B blackest pels and make these black. To choose between equally grey pels, it is necessary to use one of the techniques of the prior art, but by applying the proposed technique first, a significant amount of information from the original image will have been preserved. That this simple technique is sufficient to preserve fine detail is shown by the following example. If a white 3 X 3 rectangle is crossed by a thin black line so that 3 of the 9 pels of the rectangle are black, the initial processing should indicate that the average grey level of this rectangle is best represented by 3 black pels. The 3 blackest pels are, of course, the 3 black pels, so these will be made black in the output image by...