Upload
others
View
8
Download
0
Embed Size (px)
Citation preview
Rochester Institute of Technology Rochester Institute of Technology
RIT Scholar Works RIT Scholar Works
Theses
1967
The determination of a set of alphanumeric characters of equal The determination of a set of alphanumeric characters of equal
recognizability recognizability
Keith C. Donaldson
Harry O. Gough
Follow this and additional works at: https://scholarworks.rit.edu/theses
Recommended Citation Recommended Citation Donaldson, Keith C. and Gough, Harry O., "The determination of a set of alphanumeric characters of equal recognizability" (1967). Thesis. Rochester Institute of Technology. Accessed from
This Thesis is brought to you for free and open access by RIT Scholar Works. It has been accepted for inclusion in Theses by an authorized administrator of RIT Scholar Works. For more information, please contact [email protected].
The De,ter::rinat ion of A set:
ReG(jgnizability ,l
,Keith C • . ,
'·r
": i
'"
A thes i..s ,' ~s ent(2d ir_ par \..: requirem2n~s of the Bach~ Rochester Institute oiTe
. ' . • ! ..
Harry O.
Donaldson '
Go(3h o
o r c.: :~e
~"3 6r€:€ .. "., ccr, ~ : ~'{ •
. , .. J.. •
\" .
bart~rget, i.e., 5:1~ dfld ,th~t ~ney and the surrc~~d
,have g. 10,\'; density difference. Such a set is ~.;:sc.r:...bed) .. ,,' ,', ',. ," , " -' ;:,:::;?<:':':";~-,:,>"~':/,,~,
.tbg~th~r 'wit h the experimental technique and da'ca :, -,' ,'.;.:'." '".-",
:\.~f{~lysis used by the present authors, to evaluate , - ,,-, ( ;F:~t< "
'iviH~j~jset:"'Atan alpha risk of 0.10 the set of char-~;~:,~j);;¥~.~~/:"{'_;:', 'Yi~;~'·:::,:,·
a'c:~~rs~\S 23 8 9 was found to have equal recog-"
'nizability under, two,varied viewing conditions.
·.~
INTRO DUCTION
.' f.o"1"'_· e"/;~,:., .. ,." 19' o· J, :'''''a'l -n(l ;..,1-,0'-0 ... ",,' ...... ' ~·"""'ru~:,..·..;ts " " .... ~ _L ["~':'~' ""- ." ill. '- ~"J i~ ~ ~ ... a... .;. •. ,~~. ".'\;: tl
one method i 'i.vi~~ ~;'3E:ist~t employing ::he ftimilia:c .. ',:ri-
bar target ;:;-:.
I""'E----. :3 V/ Ii
!
Figure.l. tv. ...r~\/rW1 Tri~Bar Target
(AIR 5:1)
The t~sttarget consists of·an array of tri-bars, grouped
. in threes, d~creasing'in,s~ze in a systematic pattern. The
figure of merit using thisnlcthod depends on the smaL.est
tri~har pattern imaged by, t:he instrument and recognizable
1:>Y an obse::v;.::r 0 ':'here is s0:netimes disagreement among
ohservers aboDe vinether or not a giVer: set of bars is
available. 1
It has been suggested-by McCamy, and ot~~~£ th~c .Co
alphanumeric characters be subse ituted for .: c.".; -:..:.::.:- :, ... :.
targe:: ~ I f this could' be done, much 0 f t~1e i.~J.. -Ef:i.cul'.:y
a/~soc ... 6.;::ed with the definition of rec~J:!~~:" ~5.;J:'lity wou-:..d
. dJsappear, S ii\ce. the observer could b € Cl"sk:.d :0 iderJ·\. fy
..
1
a 6haracter rather than make a judgement whether three bars
are distinguishable as suet:. This is a genuine problem in
recognition, and the o~se=V2r can b~ s~ored as right or
wrong, a pr0c~du~a no~ ?os$~.hle wi~.h ~~~ ~:~2cnt target.
char",c} er ·carg'"t is that alL
herenlly equal recognizab~i~ty. Constderable wo~k has been
done in the past Ivith this aim in vie\v in the fiel.d 0: 2-6 ophthalmology, where visual acuity is most often teste~
using the familiar. letter c;,art. The usefulness of t:·.e
previ·ous. \Vork to the presen;: research is doubtful; th"y
invariably employed high contrast characters; i.e., black
letters on a white surround; character geometry frequently
varied from worker to worker; curved and oblique contours
Ivere included, for the purpose of detecting errors in
visual refraction. Additiona11y, there are serious diff-
erences of opinion abou:: what constitutes a set of "equally
legible" or "equally difficult" characters. For example,
Hartric.ge and Owen2 claim ·:::1.e set: :. F H N P T .U X Z to
be of;::-.e 4 "sG:me difficulty"; Sloan says the set: C' J H 1<
NOR S V Z is "nearly equal in legibility".
A second consideration is that the characters selected
bear, to som" '::egree, a resemblance to the three-bar array,
since c~arE ~s some desirability in relating information
from the ch~=acter target to data based on the tri-bar target.
Figure 2 sho',;-$ such", set of characters.
2
Since most interest in optical sYHtem evaluation is in
the area of low subject-co~=rnst imaging performance, the
characters should be of .:.,'\, contr~,s::., £i.':.' 0.15 - 0.20
OBJECTIVES
The objective of the present research m;s: "c" ee,s.: the
hypothesis that the characters E 3 S 2 3 6 8 9, 1n ~lock
fo:;:m I'i'ith aspect ratio 5;1, are equally recognizable. A
chi-square test was to be applied to the data to determine
equality; the alpha risk selected was 0.10.
EXPERI~ENTAL PROCEDURE
A. Apparatus: Characters were cut from Colol--Aid paper,
No.3 Gray, reflection densi.ty = .56; the same paper, No.
5A Gray, refle~tion density = .72, was used for the surround.
These p"~et"s. ·,,;(:.-::e selected <-,; having a density difference
,,':.thin ihe r{)..'().':::,~ ori~in<.l1y select.",d; No. 5A has a density
t,bse to th<,;: ,:,£ M\ l87. Gr&y Car,L Ciiaracters measured 25mm
on aside; they were afEixe; to 6-inch squares of No. 5A
paper I'i'hich ,-,areo L.vcd:ad o.bc,wt the circumfe:ce:-.ce of a disc.
Figur\:: .3 sho.ws; a dr8.-..ving of cht:; presel1tatiO~~J ,:'!evice; the i
~isc 1~ rA~~~ed to bring in~ividual characters to the 5~"
square "pCIClclre in the plate bahin" ,,;1.\ e;h t:he disc is
piaced. The front plate was covered with No. SA paper.
B. Viewing Conditions; T':w targ>2t: \,'~,? h l: by twe
Hood lamps in reflectors; illumination "BS. hei.d conSL<>-nt
12 < - .,' ."] b 1· 7,8 1'· d· . a I:: ...J rt-c.. ~ata. are ,aval. . .:1. e Wil~cn 1....-1 locate tnat
3
-----,1 5 .,
. !
at the observer). This relationship is shown in Figure 4.
With the large-group viewing the intention was: to posi-
tion the observer approximat"'::'y by use of the Snellen Chart,
present a short s2ries of ch&rnct0~s1 calcc:~~2 ~he per
·cent. corr~cc ~~sponse, and relocste =hc observer at the
distanc~ which wo~ld result ~~ 50% cor=~ct :esponse. This
prediction t2chn~~~e was ~ns&~isfactory) p~0bahly because
the observeis f~~~ whose daca it was der~ve~ ware expar~en-
eed in ~~e viewing operation and g~oups of observers w~~e
not. C;, this basis, the secc"'c1 of tlvo sets of observations
was used in the &nalysis.
Each observer was given a report form for each series
of characters presented. As each member of the rando~.11y-
o:;:dered series was shOlvn the experimenter called out the
p:;:~sentation order number (this had been found necessary
to avoid co;,::'~s~o~1 in recording), and the observer recorded
his response in ;:,le co:;:respouling space on his form. For the
data analyzed in this pape:;:, ~he presentation series to the
fi:;:st g:..-oup, 11 men, was each character six ci.~es, 48 total;
t:-:.e secon~ men, saw each character ~w~lv0
95 total. A :;:esponse was forced in all cases.
,
DATA ANAL YS IS
Data from observers \vere talm lilted .indi vioua lly showing,
for each character, ea) numbel: of t:.I,les presented, (b) number
of times reported, and (c) nUGhc:: 0;' times :.:-eported correctly.
A typical individual tabulation ;"r; shown in ?igure 5.
~~ !., ~~:: c; t:: 1.-E G :; 2 3
, 8 9 To::al 0
(a) Presented 12 12 ~2 1.2 -! 'I 12 12 12 S5 '. _. (b) Reported 10 10 ::.6 11 12 6 1 ' .0 15 96 (c) Correct 10 5 11 H ' , l.L 3 10 9 70
Per ce'0 t • correct: 72.97,
Figure 5. Tabulation of Data of Typical Obse:.:-vcr
In the data analysis an experimental chi-square value
was calcu1~~cd from the test statistic,
)(' 2 _""24 (O-E) 2 I _ -.~
, .-oG- ~ 1
(1)
v;'":le:ce. 0 l.~ ;:;',e :;'u,'ilber of 'c:Lrnes ti1e event happened - in this
cas~ ~~~ ~~~~er of times 3 chnr~cter was reported - and E
~s .ch~ ~U~Der of times· the event Ivould be expected if the
The arith .. ,etic .is quite ,
strcu.ghl:fo:;:-wc..:d; using the above data to calculate response
'h' . ..,,~o • "-~ ~-;;,qu.;...- ....... .. 2 [,. ~ . ",2 'r-- == ... .Ll)~_L) ~ (2 )
2 2-, <:"0-12) -:- ••• +(15 .. 120112 ='7 .. 1665
Simi~~~ly, c~~~cct response chi sqU&r8: _ 2 ,2 2 ' .
'X2 = ~lO-E) + (5-E) + ••. + (9-E)]/E :, 7,4861, (3)
where the expected va:Llc, E, is the, product of the (total)
per cc:::nt. corrc:ct and the numbo=r. of presentations of a,
6
-
character, 'i.e., E = (12)(.729). In these examnles the
experimentRl values of chi-square are less than the critical
value of chi-square for a sample of eight and an alpha risk
of 0.10, ,i.e., 12.0170. T~us th~ da~a indicate Lhat there
is no difference Ivithin the set of ;:,,12 respeJ-:1ses or the
correct responses.
T\'1O criteria \>le:ce US2C to l:'..:!j2,";:': da'Ca f:corrl individuals.
Those from observers showi~~ lJl12q~a]. respons2 w~~e =ejected;
the reasoning was 'Chnt ariY bias in ~eSrGr.3E; "". v ... h.o..tc·ver the
sponse. Oata were also discarded from observers scoring 100%
correct responses for more than "one character.
A tabulation was made of the; sunmed responses of obser-
vers meeting the above criteria and chi-sc;.uare calcu1E.ted
for response and correct response. Response chi-square was
less than the critical value, indicating that responses were
equal; correct-response chi-square was greater,than the
critical value, which indicated that the responses - and
therefore the set of chariicters - \.;e1:e not equal.
It was noticed that the summed correct responses for
the characters G and 6 were lower than similar val~€s for
the other characters. Individual tabulat~ons were recon-
sidered, eliminating from the calculations all responses
for the suspec~ed characters. Four observers previously '-,
rejected for ucequal responses became accepteble; three
Ivere not helped by this operation, since the inequality of
their responses was due to apparent biases for other char-
acters in which no pattern was noticeable. Also, with the
7
elim'ination of G and 6 re!;pOllSes, the correct-response chi-
square based on thG ten ncc('l''cnble observers, taken indivi-
dually and collect!.vely, W;),; :'esf.. 1:h"n the critical value,
9.23635 for the same 0.10 <: ... 1.'1 :::isk a.nd sar'"r:>le s1.,e of. 6.
shO\vn in Figure 6.
: ; .. '. . , ..
Ca) Presented (b) Reported (c) Correct
Per cent.
E 90 87 70
r,i, 'Ira,'~r-er •. ::!.':":-~~-
" .:> 2 ~
.)
90 90 90 10;~ 89 85 65 74 68'
CO:;:"7LQct: 10.2
8 90 87 51
Chi-square: ?..:.es:)O~"',:,2: 2.LJ.888 6.9531 Co:c::ec~: :
9 Total 90 540 84 534 53 381
Figur~ 6~ Tabula=:Lon of SUffir:.lec1 Data; of" :.';:. ~ ~b ,.~......, ... en u scr\}.::rs i.li3.V~ng .t!.C,ua.L
l{e S ;)011 S(~
A check of internal co;-.3ist,wcy was made Dy considering
eX'Jeri':.lental chi-square \'las lesS ~han the critical V6:~ .. :..:e.
Of the ten observers five i~?rov~d in performance between
the fi:st and second halves of the preseritation series; two
scored che same on both halves, and three scored worse on the
second halE.
It was decicLd to confirm the conclusions drawn from
the preceeding &~~~ysis) that the six-member set of char~
acterswas e~~ally recognizable. It was thought that this
should done so that the character viewed would be degraded
8
• - ---<;I -
imag~-wise, in a manner similar to thac which might be found
in a practical situation. Several schemes were considered;
the one finally adopted was tho: o~ defocusing the projected
image of the charactcro Acccirdingly, ench ch~~=nctcr was
phot o,[;r aphed and the rc s III t :.;~ "E: ga t i VeS werE. S l.:ccle-mounted.
0.15 - 0.16; screen illuminat~on was 210 ft-C. with no slide
in the light path; and 5.3 ft-C with a slide in place
(proj ected background).
'rvlelve slides of each d;m:acter, 72 total, were randomly
presented to six observers, four of whom were replicated.
Tabulaciol1 and sorting were acconplished according to the
previous ly i'.lent i0l'12G. cri teric-.. ..
D .. \TA A:-TALYSIS
The data f:com three obsc:rvers \,,:tere acceptable; t\"O of
these were replicated. Only one of the rejected observers
\'las rejected by reason of unequal response; the others
\·;ererejected for 100/'0 correct r,!sponses for more than one
characte::- ...
Correct-response chi-squares ';'Jere ca lculated from the
i;-,Givicual data and the summed responses; all were less
than the critic~.l. value. Of the tl\'O observers who \.;ere repli-
cated one scored :,e:ter in the secOl1c" series., the other worse.
Figure 7. Ta bt:l. c:~~ ion of Sumrr.ed Data
'" S 2 3 8 ~
:resented 60 60 60 60 60 ;;2;>ortecl 56 62 72 If 7 70 Correct 40 40 51 33 46 Per cent. correct: 67.2%
Chi-square: Respo~lse : Coi:rect:
8.0330 6.6779
of Three Observers
9 Total 60 360 53 360 32 242
.. , I
, '
CONCLUSIONS
The data indicate that, unde~ two widely varied view
ing conditions, there is no diEf:erence i;-, cr.e correct
responses to members of the six-character Bet when they
are presented each an equal number of times.
We therefore conclude Lli1t the original objective
of the experiment has been satisfied; this is a set
of characters of the requi::ed geo .. ,etry which are, inter .§£.,
equally recognizable.
10
ACKN(1'.~LE DGEXENTS
The authors express tki.r thanks to the photographic
Science Faculty for much helpEul consultation and advice;
in particular to Professors Hollis Todd and Albert
Rickmcrs, \-lhose counsel \Vf.iE· illvc:luahlc.
To the students I"ho sex'veo <lS observers goes m.;.
undying gratitude, especi<:J.1y those -"ho served during
the early part of the project. Without their patient
cooperation the research \Vould have been literally
impossible.
11
I I I
. I
x
x
x
10 -';...cO ~o
><
~ ~ ~o
..:- t[)
: 10
i·-,---J,~---1-1 -'·-'l-"-+-·-·-·I·---+~-l--;-:-.J o Q
. N lJ3't:l'dOJ
o lq
'C13c 90l
Q N
FIGURE ~ 4
r-'--I
CHARAC TE RS
USED
r~~-;-"'-"" '--"'-'-'-'-;' . i .. ,,,~--,,,,,-,,,,,~
l'· L_ .. ~ .. ~ ~-..... --.~ .. ~ i
i .
[~--! ,
:: .;.. __ Nt= .. ~..: ..... ,..,.. ......... j
,,-,-. ~-.. l L,., __ ._ .. ·1
n---.-.~--.,-j j L ' ........ _._._... I
i I :~.-~-'- 1 L ..... "'''' ......... ~ .... .hWOJ
B
FIGURE -2
APF'AFtATUS
, , , I .
I. I I
~ -f-..:::.... - -,., .....
/ " , '/
I I,
5~11 E I T:;~ 5.5'~
\
I
~
._ ... --.~-------------l . I • : ' 4/------~. 7:
FIGURE-3
,-..
I I 'I
30
FOOTNOTES
1. HcCamy, C. C.; On the Information in a l~licrophotograph, Appl. Optics, ~:405 (April)1965
2. Hartridge, H. and Owen, H. 8.: Test Types, Brit. J. Ophth., fi:543, 1922
3. Banister, H.: Block Capital Let~ers as Tests of Visual ACdity,. Brit. J. Ophth., 1l:lf9 (Fo.b.) 1927
4. Sloan, L. L. ~ ComparinG;') of Three Types of Test r-;arget for the ~:casur0ment of Visual Acui;;y, Qua;:t. RE:V. Ophth., 12.:4, 1952
5. Hay)? J.: (no ticla)~ Tra~~ Ophth. Soc. U.K. Vol.39) 1919.
cited i:l~ Hartridge and O\vcn, Q12 cit . .
6. Sheard: (n.::. title), (,me;:. J. Phys. Opt., 1:168, 1921 cited ~:~; Baniscer., OD cit.
~--
7. :Sythgoe,~. J,,: The ;''lensurcmen"c of Visual Acuity, S?2cial Repo::::t t\o. l73, London, H~ Sta:ionery Offica"
8. Hecht, S: ::ZclatioTi oet:t:·.1eG:n Visual Acuity and Illuminac ior,. J. Gen. 1'11ysio1., ll: 255, ( • \" ·'28 .JU:-Io) I •. ';1
7 and 8 a;:c cited in: Sloan, L.L.: Meas"c:re:r:ent of Visual .\cuity, a Critical Review, Arch. (',?:~'.:h., 4·5:70l, (June) 1951.
9. ' Craile, K. J. W.: The E~fect of Adaptation U?O~ Visual Acuity, :Orit. J. 1'sychol., 29:252, 1939 Vi sio11 ~;,rl. Yj~SUEl1. Percent ion, Grc:.,~Qm, C. (ed) , John \-IUey & Sons, Ne\>l York, 1965, p. 340
12
BIBLIOGRAPHY
1. Banister, H.: Block Capital Letters as Tests of Visual Acuity, Brit. J. Ophth., 11:49 (Feb.)1927
2. Bl ackl'le 11 , H. R.: Contrast Thresholds, JOSA, 36:624 (1946)
3;,",CoI),an, A.: Test Letters ~lhich Comply vlith the Physiological Requirements of a VisuAl Test Object, Amer. J. Ophth., 11:625 (1928)
4. COIVan, A.: A Visual Test Card; Amer. J. Ophth., Q:301 (1929)
5. Duntley, S. Q., et al.: Visibility, Appl. Opt., 2:563 (May)1964
6. Graham,C. H., et al, eds.: Vision and Visual Perception, New York, John iViley & Sons, 1965
7. Hartridge, H., and Owen, H. B.: Test Types, Brit. J. Ophth., 2,:543, 1922
8. Mauge, '" . / ".
R.: Etude du Pouvoir Resolvant des Emulsions Photographiques, Science et Ind. Phot., 1:297,
9. Sloan, L. L.: Measurement of Visual Acuity, a Critical Review, Arch. Ophth., 45.:704 (June) 1951
10. Sloan, L.L·., et a1.:. Comparison of Three Types of Test Target for the Measul~ement of Visual Acuity, Quart. Rev. Ophth., ~:4, 1952
11. Sloan, L. L.: New Test Charts for the Measurement of Visual Acuity at Far and ",ear Distances, Amer. J. Ophth., 48:807, (Dec.) 1959
1936
12. Sturge, J. M.: Detection and Recognition of Low-Contrast Images, PSE, 6:156 (May-June) 1962
13. Wul feck, J., and Taylor, J., eds.: .EQ.rm Discrimination as Related to Military Problems, Publication 561, Nat. Acad. of Sciences - Nat. Res. Council, l-lashington, D. C., 1957