AC

REO

Res

earc

h an

d de

velo

pmen

t

in

elec

tron

ics

and

o ptic

s.

Acceptance of Jagged Edges of Letters on Computer Screens A Masters Thesis examined by Göteborg University

Robert Schmid

Acceptance of Jagged Edges of Letters on Computer Screens A Masters Thesis examined by Göteborg University

Robert Schmid

© ACREO AB

Kista and Norrköping

2001-07-20

Report no: acr-xxxxx [Ev rev. nr]

ACREO AB Isafjordsgatan 22

Electrum 236

SE-164 40 Kista

Bredgatan 34

SE-602 21 Norrköping

Abstract The goal of this Masters thesis was to examine when jagged edges of letters were accepted. More specifically, it investigated how participants judged texts presented in different simulated screen resolutions on LCD screens. Such knowledge is important in the area of visual ergonomics, especially for legibility. The idea was to find an acceptance level of jaggedness to use as a requirement in the future TCO 200x standard. Typography and earlier attempts to measure jaggedness and visual ergonomics were investigated, through the use of various literature sources. A subjective judgement experiment was also designed and performed. The experiment was divided into two different viewing distances using two techniques to simulate different measurable screen resolutions. 18 voluntary observers participated in the experiment. The results showed that judgements varied as a function of jaggedness. The category "Clarity" was found to be less sensitive as a judgement scale compared to "Aesthetics, Jaggedness, and Appearance. Finally, there was not possible to combine the results from the two different viewing distances. The recommendation to TCO 200x’s future requirement for jagged edges on LCD displays is suggested to be of a minimum resolution of at least 5 pixels/mm.

i

Contents

1 INTRODUCTION 4

1.1 AREA OF INVESTIGATION 4

1.2 AIM 6

1.3 HYPOTHESIS 7

1.4 TYPOGRAPHY 7

1.5 LEGIBILITY 8

1.6 JAGGED EDGES EXPERIMENT 9

1.7 LEGIBILITY EXPERIMENT 9

1.8 ANTI-ALIASING EXPERIMENTS 10

2 EXPERIMENT I: CLOSE DISTANCE 11

2.1 METHOD 11

2.1.1 APPARATUS 11

2.1.2 LABORATORY ROOM 11

2.1.3 STIMULI AND DESIGN 12

2.1.4 TEST PROCEDURE 13

2.1.5 OBSERVERS 14

2.2 RESULTS 15

3 EXPERIMENT II: FAR DISTANCE 18

3.1 METHOD 18

3.1.1 APPARATUS 18

3.1.2 PERCEPTION LAB 18

3.1.3 STIMULI AND DESIGN 18

3.1.4 TEST PROCEDURE 19

3.1.5 OBSERVERS 19

3.2 RESULTS 20

4 DISCUSSION 22

ii

5 CONCLUSIONS 23

ACKNOWLEDGEMENTS 24

REFERENCES 25

LIST OF APPENDIX 26

iii

1 Introduction

Since the introduction of the personal computer and the more recent increase in the use of on-line sources of information, the option of reading from paper rather than from computer screens has been drastically reduced and in some cases eliminated (Boyarski et. al, 1998). People are very visual animals and use their sense of sight to interpret much of the world around them. Visual perception is probably the most important sensory impression for people when interacting with the environment. The visual system is the most investigated sensory but more knowledge about how people perceive is utterly important because of the increasing amount of information-flow in our society. To handle this we need to develop good Human Machine Interaction interfaces by producing more ergonomic facilities. More knowledge about the human visual system is needed to eliminate sources of disturbance. With more knowledge about how people perceive, the possibilities to present better-formed information are improved. One example of visual comfort is legibility, or how easily information can be read from a computer screen. Lack of visual comfort in this area may give rise to serious complaints such as headache, eyestrain, extreme fatigue or other annoying effects (Cakir et. al, 1980). Visual comfort is largely determined by, but not identical to, perceptual quality. Visual comfort is about layout, fonts, interior lightning, and colour etc. Perceptual quality is about visual attributes like sharpness, brightness, and contrast etc. (Roufs & Boschman, 1997).

1.1 Area of investigation

The amount of time that people spend reading and writing from computer displays is in some cases more then eight hours per day. Having this in mind it is not so hard to understand the importance of legibility. Visual comfort is dependent on many factors for example image stability and legibility (TCO’99). A legibility phenomenon that can be annoying is jagged edges also known as the staircase effect (see Figure 1). The underlying cause to jagged edges is the pixel grid i.e.the computer screens smallest units, which in most cases has a horizontal and vertical pattern. This leads to a problem with representing objects that are diagonal or round. Jagged edges can be observable when an oblique edge, especially a linear edge of very shallow or steep slope, is rendered on a bitmap display. The effect is even more pronounced when a text in a bitmap font is enlarged. The jagged edges effect increases with high edge sharpness and high contrast, e.g. new LCD screens.

Figure 1. Jagged letters

4

Since the use of new LCD based computer screens in offices and at home is growing, the problem with jagged edges becomes more acute as well. The pixels on LCD screens are more distinct compared to the pixels on CRT screens. On a CRT there is often today no such thing as a pixel grid. However n a LCD screen, the pixel grid is distinct. The wiring produces hard edges to each pixel. The screen manufactures will probably solve the problem with jagged edges on LCD-screens in the future using new materials and high resolution to form smaller pixels. There is already some evidence today which level of pixel density is required to ensure that the problem disappears, and that level is suggested to be 35 pixels/mm for a 400 mm viewing distance (Naiman, 1998). Since there isn’t any LCD screens with a pixel density that high today, there are other solutions to the problem with jagged edges. Anti-aliasing is one technique used to solve this problem in texts. Anti-aliased fonts are typically used in headlines and subheads, rather then in long passages of text. Conventional anti-aliasing attempts to smooth out the "jaggers" at the edges of letters, but it doesn’t work so well for small font sizes. The effect on small texts is that the type appears blurred, loses contrast, and legibility is badly affected. Because anti-aliased fonts need extra bits for the grey levels the minimum font size that may be used is 14 pt (Boyarski et. al, 1998). A new colour anti-aliasing filtering technique applicable to displays with individually addressable sub-pixels that can make small fonts more legible has been created. This technique is called RGB-decimation and is used for instance in Microsoft’s ClearType™ system (see Figure 2). Figure 2. A pixel on a typical colour LCD-screen consisting of three addressable sub-pixels:

Red, Green, and Blue (RGB) The filtering in ClearType is based on a perceptual model of human vision that leads to an optimisation technique for finding the best legibility values (Betrisey et. al, 2000). The optimal filters allow the use of 25% smaller fonts with no degradation of font quality (Platt, 2000). The impact of ClearType on the display industry is claimed to be significant. Without changing any manufacturing processes, or adding any cost to the display itself, displays will achieve higher effective resolution (Betrisey et. al, 2000). Another technique to decrease the problem with jagged edges is vector fonts (or outline fonts). Vector fonts can be defined as a set of lines and curves as opposed to bitmap fonts. A vector font (e.g. PostScript, TrueType etc.) can be scaled to any size. It can also be transformed more easily than a bitmap font, and with more attractive results, though it requires massive computational processing.

5

Finally, a feasible but not realistic way to solve the problem with jagged edges in texts could be a typeface that only uses horizontal and vertical lines. Rubinstein (1988) has an example of such a typeface with no case distinction that is designed without any regard to the expectations of the reader and it’s therefore very hard for people to decode. The only way to understand the symbols is through reading in a meaningful context using the extraordinarily good ability of human pattern recognition.

1.2 Aim

The underlying problems to jagged edges are known as well as some solutions to it, and this work intended to analyse the phenomenon in more detail. Another aim was to establish data that could be used to achieve better visual ergonomics in the area of legibility. The main purpose with this work and experiments was to find an acceptance level of jaggedness, to use as a requirement in the new TCO 200x-standard. This work primary investigates and adjusts just one cause to jagged edges, namely resolution. Other important causes like screen-contrast and the pixel-angularity are fixed in our experimental design. Two different techniques to simulate measurable jagged edges on fonts will be used in the experimental design. Two different experimental set-ups were used to solve the problem with screen resolution, which is fixed on the LCD-screen, approx. 3,36 pixels/mm (1024 pixels/305 mm). Both long and short viewing distances, actual and bitmap typefaces made by simulated pixels were used in the experimental design. Category scales were used to measure the observers’ judgement about the different simulated screen resolutions (see Figure 3). The intention with this report is to find out on which resolution level jagged edges can be accepted of users. What is your opinion about the Clarity of the letters? Very unclear Neither nor Very clear

0 1 2 3 4 5 6 7 8 9 10 Figure 3. A judgement scale for the category “Clarity”

6

1.3 Hypothesis

I. If jagged edges were increased on a screen, then the observers’ judgement will vary as a function of the jaggedness.

II. If jagged edges were increased on a screen, then the clarity shouldn’t be

that badly affected compared to jaggedness, aesthetics, and appearance. III. If two different viewing distances were used, then the result would be

comparable, if measured as a function of visual angle.

1.4 Typography

A typeface is a character set that shares a similar appearance and design. The character set includes letters, numbers, and symbols. Traditionally, the term “font” represents a complete set of characters or symbols, which share the same size and style. For example: 12 pt. Arial, 12 pt. Arial Bold, and 12 pt. Arial Italic are all different fonts, but the same typeface is used in each. This report is, for example, written in a Roman typeface (Century Schoolbook) which always include serifs and stroke variations (see Figure 4). This family of typefaces is the most commonly used typeface for a body text (text set 12 pt. or less). Other common examples of Roman typefaces are Times and Georgia. The typeface Arial used in our experiment was a sans serif typeface, which means that it does not have serifs. They are generally more modern looking than traditional Roman typefaces. They also have a uniform stroke width as opposed to Roman typefaces, which has a variable stroke width. Examples of other sans serif typefaces are Geneva and Verdana. A type family includes all the various styles within a typeface. For example: Arial, Arial Bold, Arial Italic, Arial Bold Italic, etc. A point (pt) is the smallest increment of typographic measure used to measure height of type and ruled line thickness. 12 points are equal to one pica, 6 picas are equal to one inch, and finally 72 points are equal to one inch. Common point sizes are 7, 8, 9, 10, 11, 12, 14, 18, 24, 36, 48, 60, 72, and 96 pt. All text set 12 pt. or below is called body text. All text set at or above 14 pt. is called display text. Line space (leading) is the space (measured in points) between rows of text, from baseline to baseline. Line length is measured in picas.

7

serif ascender

x-height Graphic

baseline

descender Figure 4. Description for various parts of letters

1.5 Legibility

The basic task of reading is letter recognition. When people learn to read, they probably start with the alphabet, learning individual letters, but at some point most of them cross a threshold at which the shape of a word like "chair” is imprinted in their memory. From that time on, they never read the letters c-h-a-i-r, they just recognise the whole word. When they become more adept readers, they may read larger units of meaning, like phrases, sentences, and even paragraphs. The science and art of typography has evolved over thousands of years to create word shapes on paper that are instantly recognisable to us. There is no conscious effort involved. Therein lies the first problem as far as computers are concerned. On a computer display one big problem is that of low resolution. The paper printout is therefore much better then the screen picture. The resolution of an average monochrome laser printer is 600 dpi (dpi equals to dots per inch, 1 inch are equal to 25,4 mm) and a computer screen has a resolution that is equivalent to approximately 100 dpi. That makes the resolution of the printout six times better then the screen picture. Legibility depends on several independent factors interacting with each other. The size of the fonts is an important factor in reading (Tullis et. al, 1995), and it interacts as well with line length and spacing (Muter, 1996). Upper-case letters, italics and right justification by inserting blanks result in slower reading (Muter, 1996). People prefer to read text at relatively small sizes. Readable body text in printed paper (for adult readers with no visual impairment) starts around 9 point. Readability improves as text goes up to 10 point, gets even better at 11 point, starts to fall off slightly at 12 point, and gets a lot worse as you move past 14 point (Hellmark, 1994). Serifs are the tiny strokes at the end of main strokes of letters (see Figure 4). They help to tie the letters together to form word gestalts and make them easier to recognise and to lead the eye horizontally along lines of text. Serif typefaces are better for sustained reading than sans serif faces.

8

1.6 Jagged edges experiment

In Naiman’s (1998) study, the stimulus was two parallel oblique edges presented at the same time on a CRT-screen. One of the edges consisted of actual pixels and the other of simulated pixels (see Figure 5). The subject’s task was to decide which of the two edges that was most jagged. Stimuli did only differ in polarity and size of the (simulated) pixels used to compose the edges. The simulated pixels consisted of 4 x 4 block of physical pixels, 1,072 mm a side. Viewing distances ranged from 2 to 18,5 meters. The aim of the study was to determine the human visual system’s ability to detect jaggedness along a straight edge. Because of the finite resolving ability of the human visual system Naiman was interested in identifying the different threshold conditions under which visual artefacts arise.

Figure 5. Model of stimuli used in Naiman´s study. The upper edge is made of simulated pixels. According to Naiman (1998), a pixel density of about 35 pixels/mm was required to ensure that no edges appear jagged at a viewing distance of 400mm. At a pixel density of about 15 pixels/mm the proportion of visibly jagged edges increases to 50%, and at a pixel density of about 10 pixels/mm there is already approximately 100% detection of jagged edges. One may compare these results with the normal pixel density on computer screens of today, which is 4 pixels/mm. For presentations intended for people there is no reason to waste any resources trying to improve something that isn’t detectable (Naiman, 1998).

1.7 Legibility experiment

Boyarski et. al (1998) studied the legibility and subjective preferences of a set of fonts designed for screen display. In the first part of the test binary bitmap fonts presented on a CRT-screen was examined. The stimulus in the first part was a reading test (e.g. the Nelson-Denny Reading Test) with some associated comprehension questions. All text, Georgia (serif) vs. Times (serif), and Georgia (serif) vs. Verdana (sans serif) and finally, Verdana (sans serif) vs. Verdana Italic (sans serif), was set in 10 point size, with an average line length of 10 words. All fonts (except Times) were screen display fonts. Effective reading speed (score/time) was recorded. Two questionnaires to measure subjective impressions of the fonts were used in the first part as well.

9

The first questionnaire asked the participants to indirectly rate a font’s characters (hard to read/easy to read), the image of the characters (fuzzy/sharp) and the shape of the characters (barely legible/very legible) on 9-point scales. The second questionnaire was used to assess whether observers could discriminate directly between fonts when they viewed them side-by-side. The participants was asked to rate which of the two fonts were easiest to read, most pleasing to read, and most fuzzy/least sharp on 9-point scales. Boyarski (1998) showed that a font (10- point size) designed for screen displays Georgia (serif) was preferred by the participants in comparison with a font designed for print Times (serif). In the comparison between Georgia (serif) and Verdana (sans serif) both designed for screen display produced mixed results. Verdana was preferred over Verdana Italic. Tullis, Boynton, and Hersh’s (1995) performed a study of the legibility of fonts in the Windows environment. They tested twelve different size/font variations, Small Font, Arial, MS Sans Serif, and MS Serif ranging in size from 6.0 point to 9.75 point. Their results showed that Arial 7.5 and Small Fonts should be avoided due to their lack of legibility. They used a subjective 4-adjective scale: Poor, Fair, Good, and Excellent. These data led to the suggested use of Arial 9.75 or MS Sans Serif 9.75. As compared with all fonts in point sizes ranging from 6.0 to 9.75, optimal reading rate and accuracy was found for 8.25, 9.0, and 9.75 point fonts – except for MS Sans Serif 8.25. The participants were given the opportunity to rate typeface preferences. The majority of the test observers preferred sans serif typefaces. However serif typefaces were shown to be more legible (i.e. able to be read faster).

1.8 Anti-aliasing experiments

Boyarski (1998) also dealt with greyscale (anti-aliased) fonts presented on the same CRT-screen. The stimuli was a reading test (the Tinker Speed of Reading Test) a thirty-word sentence including a word in the latter half that spoils the meaning of the text. All text was set in 16-point size, 19-point leading, with an average line length of 10 words. The participants had to read as many sentences as possible (out of 60) in seven minutes and speak aloud the word that spoiled the meaning of the sentence. The participants individual score was calculated (Score equals items attempted minus incorrect items). Two questionnaires to measure subjective impressions of the fonts were used in the second part as well. The first questionnaire asked the participants indirectly to rate a font’s characters (hard to read/easy to read), the image of the characters (fuzzy/sharp) and the shape of the characters (barely legible/very legible) on 9-point scales. The second was used to assess whether observers could discriminate directly between fonts when they viewed them side-by-side. The participants were asked to rate, which of the two fonts was easiest to read, most pleasing to read, and most fuzzy/least sharp on 9-point scales.

10

The participants preferred anti-aliased fonts (16 pt. size) to bitmapped fonts. Even if they preferred anti-aliased fonts, the majority of the participants had difficulty telling the difference between anti-aliased- and bitmapped fonts. In other words, this was not a major issue for them but the visual clarity and grace afforded by anti-aliased display fonts cannot be overlooked. Anti-aliasing will absolutely not decrease the preference for a larger point size font (Boyarski et. al, 1998). Booth et. al (1987) investigated the effects of aliasing artefacts (e.g. pixel size) and screen resolution, a two-part experimental design, using both subjective rating and objective measuring (e.g. reaction time) was used. Viewing distance in the test was 1820 mm. The stimuli were objects consisting of cubes. The images were rendered at various resolutions defined by tiling six sizes: 1, 2, 3, 4, 6 & 8 on the (512 x 512) pixel matrix (1 equals 512 x 512, 8 equals 64 x 64). The results showed that at higher resolutions the elimination of aliasing artefacts did not significantly improve subject’s performance. At intermediate and low resolutions the reduction in aliasing artefacts resulted in improved performance. However the subjective ratings showed that for both the highest and lowest resolution the elimination of aliasing artefacts didn’t improve the qualitative experience. 2 Experiment I: Close distance

2.1 Method

2.1.1 Apparatus To perform this part of the experiment a Mitsubishi Diamondcrysta RDT153A, 15” LCD-panel only was used as a light source. (For further details see Apparatus Experiment II, p. 18)

2.1.2 Laboratory room The luminance at the centre of the screen was 170 cd/m², on the frame 20 cd/m², and on the background wall 20 cd/m². The illuminace on the frame was 100 lx (vertical), background 160 lx (vertical), desk 165 lx (horizontal), and finally near the eye 75 lx (vertical). The experimental set-up in the perception lab was made to control the viewing distance (see Appendix C & D, for further details).

11

2.1.3 Stimuli and design Real viewgraph pictures were put in front of the LCD screen (see Figure 6). In this “Close” experimental design the observers’ head was in a fixed position. To ensure the 500 mm viewing distance a facial rest was used, in which the observers could put their chin and forehead (see Figure 6). To simulate different screen resolutions and produce the viewgraph pictures a HP 4500 colour laser printer was used, which has six times better resolution than the LCD-computer screen used in this experiment.

Figure 6. Facial rest used during “Close” (500mm) viewing distance. One of the viewgraph

picture stimuli in front of the LCD-screen is visible in the background. Arial 12 was set as the standard typeface; the font size 12 is a normal setting for reading and writing on computer screens at a viewing distance of approximately 500 mm. A text was written in MS Word (page layout view, zoom 100%) and the height and width was measured on the screen. The same text were then rewritten in Jasc® Paint Shop Pro™, which is a pixel-based program (anti-aliasing were not used). Then the text was cut and pasted as a picture into a new document in MS Word (page layout view, zoom 100%). The text (now treated as a picture) was resized in height and width to make an almost perfect viewgraph printout copy to match the original screen size of the text (see Figure 7).

Figure 7. A bitmap text in Arial 12, shown in the size of the original screen picture in MS Word.

12

This procedure was repeated with the other font sizes 9, 15, 18, and 24. The fonts were resized as well to match the original text in height and width. Five different screen resolutions: 3-, 4-, 5-, 6-, and 8 Simulated OH (SOH) pixels/mm (OH is equal to viewgraph in Swedish) was made in this way (see Appendix A). The SOH pixels were made of five different pixel sizes.

2.1.4 Test Procedure The test procedure started with an evaluation of the observers’ vision on both eyes. The observer could chose, which eye to test first. Thereafter the observers had to answer questions about age and education etc. After that they were given a brief written instruction about the experiment (see Appendix F). The 10 minutes it took for the vision test, questions, and instructions were sufficient for light adaptation of the lab. Before the test the subject had to perform a short practice round. In the test all observers were given the same textual passages to read, though the texts were in different randomised order. After reading each text, they had to fill in a questionnaire with four judgement scales (graded 0 – 10) about the fonts in the presented text. The categories in the questionnaire were in Swedish, meaning Clarity, Aesthetics, Jaggedness, and Appearance (see Appendix J). The top three of the total four judgement scales used in the questionnaires were presented in 6 different orders as well. Randomisation of texts, mixing of presentation order, and counterbalancing the questionnaires were made to prevent any carryover effects. Half of the observers started with Experiment I ("Close" or 500-mm reading distance). The other half started with Experiment II ("Far" or 3000-mm reading distance), presented below in this paper. In the “Close” reading distance each text was shown to the subject during 30 seconds. The evaluation time was limited for the reading part. The observers had, though, unlimited time for answering the questionnaire after each text. The total amount of texts were 25 (five different texts multiplied with five different simulated resolutions). The “Close” part of the experiment lasted for approximately 25 minutes. After the whole session a paper with the former (in the experiment) presented texts were shown to the observer (see Appendix G). In order to find were the observer looked at the presented texts, the observer’s visual point was searched. The observer had to answer the following questions. “Which letters did you especially look at when you evaluated the appearance of the texts?” and “Did you look at anything particular in those letters?”

13

2.1.5 Observers The test population selected for this study was composed of volunteers of students, office-, and research personnel. The sample size were 18 observers (12 males & 6 females), the median age was 32 years (see Table 1) The majority of the observers were employed as engineers at ACREO AB (11), Master’s thesis students at ACREO AB (3), students at other schools (2), and finally employees at other companies’ (2). All observers had a normal or corrected acuity to normal vision. The observers were compensated for their effort with one cinema ticket for participating in both Experiment I and II. Nr Sex Age

1 M 25 3 F 45 4 M 35 5 M 31 6 M 53 7 M 35 8 M 27 9 M 27

10 F 20 11 M 22 12 M 29 13 M 33 14 F 35 15 M 47 16 F 37 17 F 18 18 F 27 19 M 53 Table 1. Chart over the test population

14

2.2 Results

The average value of the “Close” distance test of the experiment shows that the category “Clarity” gets the highest score (see Figure 8). This indicates that the observers regarded the texts relatively easy to read, even if they judged the texts as ugly and jagged. Suppose that a score level of 7 is necessary to reach an acceptance level of visual ergonomics, this corresponds to approx. 4,7 SOH-pixels/mm for the “Clarity” category. The reason for the choice of the score level of 7 is its strategic position on the judgement scale (see Figure 3, p. 6). The “Jaggedness” category needs approx. 6,3 SOH-pixels/mm to reach the same level of acceptance, followed by “Aesthetically” and “Appearance” at approx. 6,5 SOH-pixels/mm.

Average all texts (Close)

3,00

4,00

5,00

6,00

7,00

8,00

9,00

2 3 4 5 6 7 8 9

Simulated OH-pixl./mm

Judg

emen

t ClarityAestheticsJaggednessAppearance

Figure 8. The observers’ average judgement of all texts in all presented simulated resolutions of each category in Experiment I.

One sees that the general form for all the scales is similar, but with different position relative the x-axis. Since three of the scales are close to each other, i.e. “Aesthetics”, “Jaggedness”, and “Appearance”, this may indicate that they are based on the same perceptual process.

15

There is a slight tendency that 5 SOH-pixels/mm is judged equal or even better than 6 SOH-pixels/mm in some of the texts at the “Close” viewing distance (see Figure 9). One reason could be that the text in 5 SOH-pixels/mm appears bolder than the same text in 6 SOH-pixels/mm (see Appendix A). Another reason could be that there is some kind of visual borderline between those simulations where the observers couldn’t chose which one was the better one.

Average Aksjuk (Close)

3,00

4,00

5,00

6,00

7,00

8,00

9,00

2 3 4 5 6 7 8 9

Simulated OH-pixl./mm

Judg

emen

t ClarityAestheticnessJaggednessAppearance

Figure 9. The observers’ average judgement of the text “Aksjuk” in all presented simulated resolutions of each category in Experiment I.

It is also possible for example that the graphs presented in Figure 9 are based on two underlying regression lines, one steep from 3 to 5 SOH pixels/mm and one more flat from 5 to 8 SOH pixels/mm.

16

The average score of the “Jaggedness” category shows that there were no apparent differences between the texts except at for the point of 5 SOH pixels/mm on the x-axis (see Figure 10). At this point two of the texts, i.e. “Aksjuk” and “Hungrig” was judged higher then the other texts. The texts at the 5 SOH pixels/mm level appeared bold (see Appendix A.). The text “Aksjuk” included the word “yrsel.com” in italics, and the text “Hungrig” included the word “(takeaway)” between parentheses (see Appendix G).

Average Jaggedness (Close)

3,00

4,00

5,00

6,00

7,00

8,00

9,00

2 3 4 5 6 7 8 9

Simulated OH-pixl./mm

Judg

emen

t VingenSQAksjukHungrigLinus

Figure 10. The observers’ average judgement of the category “Jaggedness” in all presented simulated resolutions of each text in Experiment II.

17

3 Experiment II: Far distance

3.1 Method

3.1.1 Apparatus The same Mitsubishi Diamondcrysta RDT153A, 15” LCD-panel was used as in Experiment I. But in this “Far” part the LCD screen actually was used to present the stimuli. Resolution was set to 1024 by 768 pixels, True colour, 75 Hz. Contrast and brightness was set to 100 % each. The computer used was a Pentium Pro 200 Mhz with 196 KB RAM. The graphic card used was a Matrox Graphics MGA Millenium (MGA-2064W R1).

3.1.2 Laboratory room Luminance measurements were the same in this set-up as in the former “Close” part of the experiment. There is though one different illuminance measurement in this set-up, and that was nearby the eye 100 lx (vertical) measured at the 3000 mm viewing distance, (for further details see Appendix C & D)

3.1.3 Stimuli and design This experimental design was built on a larger font size (e.g. Arial 72), that leads to a 6 times longer viewing distance (e.g. 3000 mm) in comparison with a common viewing distance of 500 mm (see Figure 11).

Figure 11. A six times longer viewing distance at the “Far” experiment. Because of the longer viewing distance, Arial 72 has the same size (angle) on the retina as Arial 12 at 500 mm. The different stimuli were presented as a slideshow in PowerPoint on the LCD screen. A short text in Arial 72 was written in Jasc® Paint Shop Pro™ (no anti-aliasing), and saved as standard size.

18

Then the same text was written in Arial 36 enlarged 200% and saved as a new picture to match the former standard both in height and width. The procedure was then repeated with the other font sizes of 24 and 18. Four different screen resolutions: 24-, 12-, 8-, and 6 Simulated BMP (SBMP) pixels/mm was made in this way (see Appendix B). The SBMP pixels represented four different screen resolutions (see Figure 12). Figure 12. 24-, 12-, 8-, and 6 S.BMP pixels/mm

3.1.4 Test Procedure The test procedure was the same as presented earlier in the“Close” experiment procedure. In the reading distance test of this experiment each text was shown to the observer during 15 seconds. The evaluation time was limited for the reading part. The observers had unlimited time for answering the questionnaire. Total amount of texts were 16 (four different texts multiplied with four different simulated resolutions). The “Far” part of the experiment lasted for approximately 10 minutes. After the whole session a paper with the former (in the experiment) presented texts were shown to the observer (see Appendix G). In order to find were the observer looked at the presented texts, the observer’s visual point was searched. The observer had to answer the following questions. “Which letters did you especially look at when you evaluated the appearance of the texts?” and “Did you look at anything particular in those letters?”

3.1.5 Observers The observers participating in this part of the experiment were the same as those in the former presented “Close” part (see Table 1).

19

3.2 Results

The average score of Experiment II shows high values for the “Clarity” category (see Figure 13). “Clarity” crosses the criterion level of 7 at approx. 11 SBMP pixels/mm, and the “Jaggedness” and “Appearance” categories need approximately 14 SBMP pixels/mm to reach the same level of acceptance followed by “Aesthetically” at approximately 17 SBMP pixels/mm. The average scores of all texts also show that the “Jaggedness” category was badly affected of low resolution.

Average all texts (Far)

2,00

3,00

4,00

5,00

6,00

7,00

8,00

9,00

4 6 8 10 12 14 16 18 20 22 24 26

Simulated BMP pixl./mm

Judg

emen

t ClarityAestheticnessJaggednessAppearance

Figure 13. The observers’ average judgement of all texts in all presented simulated

resolutions of each category in Experiment II. The same appearance of non-linearity as in “Experiment I” is also evidenced in Figure 13, suggesting similar underlying perceptual or cognitive processes.

20

The average score of the “Jaggedness” category in Experiment II indicates that there were no apparent differences between the texts. There was a strong similarity between the judgement scores for the texts (see Figure 14). This reflects the similarity between the texts used in the “Far” experiment (see Appendix I).

Average Jaggedness (Far)

2,00

3,00

4,00

5,00

6,00

7,00

8,00

9,00

4 6 8 10 12 14 16 18 20 22 24 26

Simulated BMP pixl./mm

Judg

emen

t FilmLarmVinKista

Figure 14. The observers’ average judgement of the category “Jaggedness” in all presented simulated resolutions of each text in Experiment II.

21

4 Discussion

The results from the two different parts of the experiment ought to be discussed at a unifying single scale. The simulated pixels were translated into pixels/degree at the eye at the actual viewing distance (see Table 2). For the complete stimuli translation-list (including formulas), see Appendix E. SOH SBMP Pixels/degree 3 20,8 4 27,7 5 34,6 6 41,6 6 44,0 8 55,3 8 58,6 12 88,0 24 175,9 Table 2. Translation from SOH and SBMP into pixels/degree.

The average value from the category “Clarity” of both distances shows that the observer judged the stimuli at the two different viewing distances independently. It almost seems like the functions are moved parallel on the x-axle (see Figure 15). This could be because that when looking at the text at close distances, the observers performed a reading task, while at the far distance, they performed a picture evaluation task. This may cast some doubt on using long distances, when studying legibility. There were though some artefacts following the resizing procedure of fonts used in this work.

Average Clarity (Close & Far)

4

5

6

7

8

9

20 40 60 80 100 120 140 160 180

Pixels/degree

Judg

emen

t

Clarity "Far"Clarity "Close"

Figure 15. The observers’ average judgement of the category “Clarity” in all presented simulated resolutions of all text in both Experiment I & II.

22

5 Conclusions

Hypothesis I: If jagged edges were increased on a screen, then the observers’ judgement will very as a function of the jaggedness. The first hypothesis was valid. Hypothesis II: If jagged edges were increased on a screen, then the clarity shouldn’t be that badly affected compared to jaggedness, aesthetics, and appearance. The second hypothesis was valid. Hypothesis III: If two different viewing distances were used, then the result would be comparable, if measured as a function of visual angle. The third hypothesis was invalid, since it was not possible to extend the “Close” viewing distance in Experiment I, with the “Far” viewing distance in Experiment II. Recommendation: The recommended lowest resolution for TCO 200x’s future requirement for LCD colour computer screens, based on the results from this work, should be at least 5 pixels/mm.

23

Acknowledgements

This study was financed by TCO Development AB and SEMKO AB and was done with the Perception Group within the Imaging Department at ACREO AB. The Mitsubishi LCD panel used in the experiments was made available by Peter Eninger at Dotpitch Service AB. I gratefully acknowledge the help from Bo Schenkman that have provided me with the statistical analyses and supervising; Börje Andrén and Antti Revonsuo for supervising, and Bengt Jacobsson for technical support, programming and experimental design. I would also like to thank the Perception Group, Imaging Department at ACREO AB for the opportunity to perform this work. Finally, comments made on the experimental design by David Granath were appreciated.

24

References

Betrisey, C., Blinn, J.F., Dresevic, B., Hill, B., Hitchcock, G., Keely, B., Mitchell, D.P., Platt, J.C., Whitted, T. (2000). ”20.4: Displaced Filtering for Patterned Displays”, In Society for Information Display, Digest pp 1-4. Booth, K.S., Bryden, M.P., Cowan, W.B., Morgan, M.F., Plante, B.L. (1987). ”On the Parameters of Human Visual Performance: An Investigation of the Benefits of Antialising”. In CHI + GI 87, pp 13-19. Boyarski, D., Forilizzi, J., Neuwirth, C., Harkness Regli, S. (1998). ”A Study of Fonts for Screen Display”. In CHI 98, April, pp 87-94. Cakir, A., Hart, D.J. & Stewart, T.F.M. (1980). Visual Display Terminals. Norwich: John Wiley & Sons Ltd. Gescheider, G.A. (1985). Psychophysics: Method, Theory, and Application. Hillsdale: LEA. Hellmark, C. (1994). Typografisk handbok. Stockholm: Ordfront & Ytterlids. McCormick, E.J., Sanders, M.S. (1983). Human Factors in Engineering and Design. Singapore: McGraw-Hill. Muter, Paul. (1996). “Interface Design and Optimization of Reading of Continous Text”. In van Oostendorp, H., de Mul, S. Eds.“Cognitive aspects of electronic text processing”. Norwood: Ablex. Naiman, A.C. (1998). ”Jagged Edges: When Is Filtering Needed?”. In ACM Transactions on Graphics, Vol. 17, October, 238-258. Platt, J.C. (2000). ”Optimal Filtering for Patterned Displays”, In IEEE Signal Processing letters, Vol. 7, July, 179-181. Rubinstein, R. (1988). Digital Typography: An Introduction to Type and Composition for Computer System Design. USA: Addison-Wesley. Roufs, J.A.J., Boschman M.C. (1991). “Visual comfort and performance”. In The man-machine interface, Roufs, J.A.J. (ed), Vol. 15 of: Vision and visual dysfunction, Cronly-Dillon, J.R. (gen. ed.), Macmillan Press, London, 24-40. Roufs, J.A.J., Boschman, M.C. (1997). “Text quality metrics for visual display units: I. Methodological aspects”. In Displays, 18, 37-43. Tullis, S., Boynton, J.L., Hersh, H. (1995). ”Readability of Fonts in the Windows Environment”. In CHI’ 95 Mosaic of Creativity, May, 127-128.

25

List of Appendix

Appendix A: Stimuli “Close” p. 27

Appendix B: Stimuli “Far” p. 28

Appendix C: Laboratory design (top view) p. 29

Appendix D: Laboratory design (side view) p. 30

Appendix E: Stimuli translation p. 31

Appendix F: Instructions “Close” p. 32

Appendix G: The observers Visual point “Close” p. 33

Appendix H: Instructions “Far” p. 34

Appendix I: The observers Visual point “Far” p. 35

Appendix J: The Questionnaire p. 36

26

Appendix A. Stimuli “Close” (500 mm)

Arial 9 3 SOH pixels /mm

Arial 12 4 SOH pixels /mm

Arial 15 5 SOH pixels /mm

Arial 18 6 SOH pixels /mm

Arial 24 8 SOH pixels /mm

27

Appendix B. Stimuli “Far” (3000 mm)

(Note: the size isn’t exactly Arial 72)

28

Appendix C. Laboratory design (top view)

Grey background wall 340 mm 190 mm 190 mm LCD screen Table 10 mm

• “Close” (500 mm viewing distance)

• “Far” (3000 mm viewing distance)

29

Appendix D. Laboratory design (side view)

m

m

Facial- rest

n

n

Scree

Scree

“Close” distance

“Far” distance

500 m

3000 m

30

Appendix E. Stimuli translation

SOH p./mm x SBMP p./mm y Pixl./deg. 3 2,38 8,73 20,8 4 3,17 8,73 27,7 5 3,96 8,73 34,6 6 4,76 8,73 41,6 6 0,84 52,36 44,0 8 6,34 8,73 55,3 8 1,12 52,36 58,6 12 1,68 52,36 88,0 24 3,36 52,36 175,9

SOH = Simulated OH-pixels/mm (in front of the screen)

SBMP = Simulated Bitmap pixels/mm (on the screen)

p./mm = Real measured pixel size/mm

x = formula ((2π x 500 mm)/360) ≈ 8,73

y = formula ((2π x 3000 mm)/360) ≈ 52,36

31

Appendix F. Instructions ”Close”

Experiment A

Du skall delta i ett experiment om bokstävers utseende och utformning. Vi är intresserade av några aspekter på hur man uppfattar bokstäver på bildskärmar. Experimentet är uppdelat i två delexperiment: kort respektive långt läsavstånd. På det ena läsavståndet presenteras de olika texterna på OH-bilder framför bildskärmen och på det andra visas texterna direkt på skärmen. Hela experimentet tar ca: 1 timme.

I detta delexperiment kommer vi att visa en text framför bildskärmen. Din uppgift är att först läsa texten och sedan svara på fyra frågor om bokstävernas utseende. Frågorna finns i häftet. Vänd blad när alla fyra frågorna är besvarade. Texten visas under en halv minut. Därefter tas texten bort och du skall besvara de fyra frågorna. Ta den tid du anser dig behöva och när frågorna är besvarade meddelar du mig detta. Därefter visas nästa text under en halv minut o.s.v. Mellan de olika delexperimenten kan du ta en längre paus. Innan experimentet får du öva på några texter. Några frågor? Meddela mig när du är redo att börja experimentet

32

Dina uppgifter kommer att behandlas konfidentiellt. Personuppgifter kommer endast att användas för identifiering av data. Du har också rätt att avbryta experimentet när du så önskar.

33

Appendix G. The observers Visual point ”Close”

34

Appendix H. Instructions ”Far”

Experiment B

Du skall delta i ett experiment om bokstävers utseende och utformning. Vi är intresserade av några aspekter på hur man uppfattar bokstäver på bildskärmar. Experimentet är uppdelat i två delexperiment: kort respektive långt läsavstånd. På det ena läsavståndet presenteras de olika texterna på OH-bilder framför bildskärmen och på det andra visas texterna direkt på skärmen. Hela experimentet tar ca: 1 timme.

I detta delexperiment kommer vi att visa en text på bildskärmen. Din uppgift är att först läsa texten och sedan svara på fyra frågor om bokstävernas utseende. Frågorna finns i häftet. Vänd blad när alla fyra frågorna är besvarade. Texten visas under några sekunder. Därefter tas texten bort och du ska då besvara de fyra frågorna. Ta den tid du anser dig behöva och när frågorna är besvarade meddelar du mig detta. Därefter visas nästa text under några sekunder o.s.v. Mellan de olika delexperimenten kan du ta en längre paus. Innan experimentet får du öva på några texter. Några frågor? Meddela mig när du är redo att börja experimentet

Dina uppgifter kommer att behandlas konfidentiellt. Personuppgifter kommer endast att användas för identifiering av data. Du har också rätt att avbryta experimentet när du så önskar.

35

Appendix I. The observers Visual point ”Far”

36

Appendix J. The Questionnaire Markera med ett streck på den vågräta axeln, det värde som bäst överensstämmer med din uppfattning. (Använd gärna hela skalan) Vad anser du om tydligheten hos bokstäverna? Mycket otydliga Varken eller Mycket tydliga

0 1 2 3 4 5 6 7 8 9 10 Vad anser du om bokstävernas estetiska utseende? Mycket fula Varken eller Mycket tilltalande

0 1 2 3 4 5 6 7 8 9 10 Vad anser du om jämnheten hos bokstävernas yttre konturer? Mycket kantiga Varken eller Helt jämna

0 1 2 3 4 5 6 7 8 9 10 Vad anser du generellt om utformningen av bokstäverna? Mycket dålig Varken eller Mycket bra

0 1 2 3 4 5 6 7 8 9 10 37