1、Resolution as it Relates to Photographic and Electronic Imaging Approved As American National Standards Institute (ANSI) June 20, 2000 A Publication of 1100 Wayne Avenue Suite 1100 Silver Spring, Maryland 20910 USA Tel: 301-587-8202 www.aiim.org COPYRIGHT Association for Information annex B, Convert
2、ing from numbers of bits to number of grey levels; and annex C, Sample Quality Index calculations (with examples). Suggestions for improvement of this technical report are welcome and should be sent to the Chair, AIIM Standards Board, 1100 Wayne Avenue, Silver Spring, Maryland 20910-5699. At the tim
3、e it approved this technical report, the AIIM Standards Board had the following members: Name of Rep resentat ive Herbert J. White, II, Chair Robert Breslawski Betsy Fanning William Neale Louis H. Sharpe, II Herman Silbiger Christopher D. Thompson O rcian ization Represented Genealogical Society of
4、Utah Eastman Kodak Company Association for Information and Image Management International U.S. TAG to TC 171 Picture Elements Technologies APPLICOM, Inc. Recognition Research, Inc. The Document Quality and Control Committee, CIO, had the following members at the time it processed and approved this r
5、eport. Name of Rep resentat ive Michael J. Badal Jan Bastien Sandra Behel John Breeden Robert Breslawski, Chair Adele Carboni, CRM Myron Chace Richard R. Conger Walter Cybulski Suzanne Dodson Eric Erickson W. Camden Gass Organization Represented Badal Associates Agfa Gevaert NV Energen Corporation V
6、irginia Retirement Systems Eastman Kodak Company Gateway 2000 Library of Congress Conger Consultants National Library of Medicine University of British Columbia Genealogical Society of Utah Deere now line pairs per millimeter (ip/mm) is used. Originally, a line consisted of a black line (or black sp
7、ace) and an equally wide white space (.e., one real line, which when scanned requires at least two scan lines). Figure 1, Arrangement of test patterns in IS0 resolution test chart no. 2 (actual size), is a ruler for measuring resolution. The width of each dark line is equal to the width of the adjac
8、ent space. Such a line and space are called a line pair - the associated number gives the number of line pairs per millimeter (ip/mm). These test targets are imaged (though a lens) onto the imaging component (e.g., film) being evaluated and visual judgements are made on the image of the target with
9、a magnification aid. The group of most finely spaced visible lines is noted. This is the group where all five lines and spaces can just be distinguished or resolved in both directions by the observer. The spatial frequency of this line group is the resolution for the imaging component of conditions
10、under evaluation. Because the spatial frequency of this line group measures the limiting spatial frequency capabilities, it is sometimes also called limiting resolution. Figure 1 -Arrangement of test patterns in IS0 resolution test chart no. 2 (actual size) Note: Illustration not to be used for test
11、ing. To determine resolution, one image is a known target and visually determines the finest pattern on the output image that is just detectable. Resolution is a detection criterion. If our visual system detects it, it is resolved; if not, it is not resolved. See annex A, Converting between digital
12、resolution and dot pitch, for commonly used measures, for an explanation of conversion from Ip/mm to Ip/i and other conversions. 4 Association for Information and Image Management International COPYRIGHT Association for Information the most familiar of these in micrographics is a films gamma or cont
13、rast). Excessive flare from poorly designed optical paths and glare from harsh viewing conditions also reduce the macrocontrast of an image, contributing to lower resolutions. While the measurement of flare, film contrast, and glare can be done routinely with inexpensive equipment, the measurement o
14、f blur cannot. A microdensitometer or microphotometer is usually required for blur measurements. Blur can be measured by calculating the distance it takes an edge to make the transition from minimum to maximum luminance. This distance is called the rise distance. Sometimes the luminance difference o
15、ver an edge is divided by the rise distance. This measure of contrast is also used and called the edge slope or edge gradient. For further information, see 2.2, Related publications (Dainty QI = Quality Index; h = height of small characters. Today, four levels of QI are commonly used in evaluating i
16、mages and systems to be read by the human eye as listed in table 1. For example, the lower case “e” in typical reading fonts can be approximately 2 mm (0.078 inch) high. In order to achieve a QI of 5 on film, one would need to resolve the 2.5 Ip/mm group on that film. However, much typeset technical
17、 literature uses 8, 6, and 4-point types which are smaller than 2 mm in height, so smaller patterns must be resolved. Some 4-point and 6-point type are about 0.032 inch (0.8 mm) high, which requires resolving the 6.3 Ip/mm pattern for a QI of 5. ASSOCIATION FOR INFORMATION AND IMAGE MANAGEMENT INTER
18、NATIONAL 9 COPYRIGHT Association for Information QI = Quality Index; h = height of small characters. So, if the 5.0 Ip/mm pattern p was just resolvable on a document, to predict the legibility of an 8-point font of 1 mm character height h, a QI of 5 would be calculated, giving quite legible text. Th
19、e use of QI described in Equation 2 is based on characters which have a high contrast between the ink and paper, such as the IS0 no. 2 target. It does not consider secondary criteria like stroke widths, or isolated details such as the dot on an “i.” Nevertheless, QI serves as an excellent starting p
20、oint from which to fine tune image quality. For additional information see ANSVAIIM MS23 and figure 4, Quality Index graph, for a chart to use in determining the pattern to be resolved for a given character height and QI level. 10 Association for Information and Image Management International COPYRI
21、GHT Association for Information not only is the measuring method different, but so are the units. Digital resolution quantifies the number of features sampled (dots or lines) per unit distance (if the sampling rate is high enough) while photographic resolution quantifies observed feature pairs per u
22、nit distance as in line pairs per millimeter (ip/mm). Dividing digital resolution values in half (1 pair = 1 line + 1 space = 2 dots) will yield units that are equivalent to photographic resolution but whose utility may not be equivalent to photographic resolution because digital resolution does not
23、 involve visual evaluations and are less revealing. Note: throughout the remainder of this technical report, digital resolution is used to refer to the spatial interval between pixels or lines, while resolution implies the actual visual detection of these lines. This is explained in the following se
24、ctions on digital display and digital scanning. 5.2 Display Consider a noise-free display device (.e., a CRT) with a specified digital resolution of 80 dpi. This means that the horizontal and vertical pitch, is 1/80 of an inch and that the display can record a spot at only this interval. In order to
25、 create a 40 Ip/i image (80 dpi / 2 = 40 ip/i), the maximum resolution expected with this ASSOCIATION FOR INFORMATION AND IMAGE MANAGEMENT INTERNATIONAL 13 COPYRIGHT Association for Information they are not for digital resolution. Therefore, choose a printer with slightly higher digital resolution (
26、IO%), say 220 dpi, as an appropriate device for this application or choose a 200 dpi printer for this application, and achieve very legible text quality, but fine details and serifs might be lost. A 300 dpi printer would certainly meet the initial criteria of QI=8. The same criteria can be applied t
27、o scanning documents with a IO-point font. Assume that the input document is of superior or equal quality to the desired output. The level of digital resolution required on a scanner to capture that quality with a QI=8 is calculated with equation 3 with the same result of 220 dpi. However, further a
28、ccommodation needs to be taken for misregistration in the scanner. Adding 50% onto 220 dpi for this effect yields a scanner resolution of 330 dpi. Since printers are not generally available in such digital resolutions, 300 dpi is probably sufficient. Because digital images are represented as numbers
29、, there is no degradation of the image during copying or transmission. A conventional image is degraded (at least slightly) every time it is copied. Unless the degradation is so severe that there is a total loss of data, digital data allows the image to be represented exactly as before, with no loss
30、 of information. 18 Association for Information and Image Management International COPYRIGHT Association for Information & Imaging ManagementLicensed by Information Handling ServicesANSI/AIIM TR26-2000 - TECHNICAL REPORT -RESOLUTION AS IT RELATES TO PHOTOGRAPHIC AND ELECTRONIC IMAGING I 5.5 Language
31、 used with digital resolution This section addresses other interpretations and uses of the terms used with digital resolution that are sometimes used in imaging vocabulary. These terms can be confusing to the user when selecting imaging devices. 5.5.1 Lines vs. line pairs These terms are often used
32、in the specification of digital and classical resolution (see 4.1). Because some prefer to think of white space between black lines of a resolution target as simply a white line, resolution specified in lines is always twice as high as resolution specified in line pairs. While this alone is not conf
33、using, the ways they are communicated can be. For instance, lines per millimeter and line pairs per millimeter are abbreviated identically as Ip/mm. They are also often spoken of as the same. It is common to hear lines per millimeter as a contraction of line pairs per millimeter. The user must be ca
34、reful to question what exactly is meant. Because they differ by only a factor of two, it is even difficult to tell them apart in context. For instance, a printer may be specified as capable of 200 Ip/i. This can mean: -100 line pairs per inch, - 400 lines per inch, - 200 lines per inch, or - 200 lin
35、e pairs per inch. All are equally likely because all are commonly available, and could be used for the same applications. The only way to really tell the difference between a line and a line pair is to test them or request clarification. 5.5.2 Cycles vs. line pairs In photographic (classical) and op
36、tical uses, resolution (spatial frequency) is often specified in cycles per mm (cylmm) rather than line pairs per mm (ip/mm). The difference between these is very subtle and, for most cases, not worth distinguishing. They can be treated interchangeably with little likelihood of error. 5.5.3 Digital
37、resolution vs. image size There is a common practice of referring to digital resolution as simply the total number of pixels within a frame or field, or alternately as the number of pixels in the horizontal and vertical directions. Alone, this information is useful to indicate the amount of data tha
38、t can be handled in a single field at a given moment, but tells nothing about that units ability to resolve spatial information. 5.5.4 Digital halftone resolution Still another chance to confuse the different resolution measures is the use of digital halftone resolution. This term is mentioned becau
39、se of its extensive use with binary digital display devices. Occasionally, when printing grey level imagery to binary (two level) output devices, e.g., xerographic laser writers, to give the visual effect of multiple grey levels with only the two grey levels available, (black=ink, white=paper) a var
40、iety of digital halftoning algorithms are used that mimic the halftone screens of the printing industry. As such, they are specified in lines per inch or sometimes just lines, and are always cited in English, not metric, units. One way of distinguishing between halftone resolution and other resoluti
41、on measures is by magnitude. Most halftone resolutions lie in the 60-120 Ip/i region while most modern binary printers have digital resolutions greater than 200 dpi. ASSOCIATION FOR INFORMATION AND IMAGE MANAGEMENT INTERNATIONAL 19 COPYRIGHT Association for Information & Imaging ManagementLicensed b
42、y Information Handling ServicesANSI/AIIM TR26-2000 -TECHNICAL REPORT-RESOLUTION AS IT RELATES TO PHOTOGRAPHIC AND ELECTRONIC IMAGING While further discussion of halftoning and halftoning resolution is beyond the scope of this document, more information on the subject can be found in 2.2, Related pub
43、lications (see Alexander). 5.5.5 Synthetic digital resolution Because digital imaging lends itself well to mathematical processing, a variety of image processing methods can be employed. One of these is interpolation. Interpolation is the estimation of data between two or more known data points. Som
44、e scanner manufacturers will synthetically create, or interpolate, data between these points after scanning to yield an artificially high digital resolution. A popular and simple interpolation method is averaging adjacent pixel values. Using this method yields twice as much data in a given direction
45、. Hence, an artificial 400 dpi image can be created from a 200 dpi scan. Creating data this way is equivalent to maintaining a spots size while decreasing its pitch, and while there are image quality benefits to this processing, they should not be considered as real gains in resolution. It is diffic
46、ult to tell whether manufacturers are doing this when scanning, so visual evaluations using targets are always the best choice for determining resolution. 5.6 Summary on digital resolution This section on digital resolution started with a definition of the term and how it differed from classical res
47、olution, namely by its units and utility. While translating between units is a simple matter of “two,” translation of utility is not. This is because visual detection determined classical resolution while spot pitch defined digital resolution. However, some guidelines for converting from digital res
48、olution to resolution were given for digital display and scanning processes. These were based on factors that affect contrast and noise in digital imaging like spot width-to-pitch ratios and registration errors in scanning. Scanning processes required higher digital resolutions for the same text qua
49、lity because of registration errors. These guidelines are meant as starting points to further refine imaging hardware selection. Several examples were also given to demonstrate the suitability of the QI for determining digital resolution for a given level of text legibility. A final section was included to alert the reader to a number of confusing units and uses of the term digital resolution. This should help the reader from being misled when selecting digital imaging devices. 6 A comparison between photographic and electronic imaging Figure 5, Typical micrographic image chain from sourc
