1、 CIE 95 72 I 9006345 0004480 357 = ISBN 3 900 734 32 1 COMMISSION INTERNATIONALE DE LCLAIRAGE I NT ER NAT I O NA L C O M M I SS 1 O N O N 1 LLU M I NAT I ON INTERNATIONALE BELEUCHTUNGSKOMM ISSION CONTRAST AND VISIBILITY Pub. No. CIE 95 Ist Edition 1992 UDC: 159.931 Descriptor: Vision, psychology 61
2、2.843.63 Capacity of recognition 61 2.843.7 Visual perception, contrast, visibility CIE 95 92 I 9006145 0004481 273 ii This Technical Report has been prepared by CIE Technical Committee 1-1 7 of Division 1 Vision and Colour and has been approved by the Board of Administration of the Commission Inter
3、nationale de IEclairage for study and application. The document reports on current knowledge and experience within the specific field of light and lighting described, and is intended to be used by the CIE membership and other interested parties. It should be noted, however, that the status of this d
4、ocument is advisory and not mandatory. The latest CIE proceedings or CIE NEWS should be consulted regarding possible subsequent amendments. Ce rapport technique a 626 prpar par le Comit Technique CIE 1-17 de la Division 1 Vision et couleur et a t approuv par le Bureau dadministration de la Commissio
5、n Internationale de IEclairage, pour tude et application. Le document traite des connaissances courantes et de lexprience dans ie domaine spcifique indiqu de la lumire et de 16clairage, et il est tabli pour lusage des membres de la CIE et autres groupements intresss. II faut cependant noter que ce d
6、ocument est indicatif et non obligatoire. Pour connaitre dventuels amendements, consulter les plus rcents comptes rendus de la CIE ou le CIE NEWS. Dieser Technische Bericht ist vom CIE-Technischen Komitee 1-17 der Division 1 Sehen und Farbe ausgearbeitet und vom Vorstand der Commission International
7、e de IEclairage gebilligt worden. Das Dokument berichtet we will not consider chromatic discrimination and temporal variation of stimuli. 1.2 Definitions of contrast The contrast of a target is a measure of the difference in luminance between the target and its immediate surround or background. Ther
8、e exist different defini- tions of contrast depending on the shape of the targets. For simple, isolated tar- getson a uniform background, e.g. a disk or a Landolt ring, the (psychometric) contrast is given by the expression I c=t- Lt - L, Lb (1.11 where C = contrast Lt = target luminance Lb = backgr
9、ound luminance. Here the difference of neighbouring luminances is related to the background lu- minance to which the observer should be adapted. Eq. (1 .l) has a physiological meaning for the detection of threshold contrasts in visual psychophysics where the just perceptable luminance increments are
10、 studied. In the Weber-Fechner re- gion C is constant. The contrast definition in Eq. (1.1) leads to contrast values that range from O to infinity for targets brighter than their background and from O to 1 for targets that are darker than their background. The problem of an asymmet- ric range of con
11、trast values is avoided if the difference between neighbouring lu- minances is related to their sum or mean (apart from a factor 2). The relative contrast sensitivity -4- c VL= - c is designated as the visibility level. It would represent the ratio of actual contrast C and threshold contrast c.The t
12、hreshold contrast is usually taken as the contrast when P = 0.5 where P is the probability of the target being seen. For spatialty periodic stimuli like sine gratings, another contrast definition is com- monly used: IL, - L,I m= t Lb (1.3) which leads to contrasts that range from O to 1 for all cond
13、itions. In Eq. (1.3) Lt and Lb have the same meaning as above. The expression (1.3) has a physiological meaning if the adaptation luminance is governed by the mean luminance of target and background as could be possible for spatial periodic stimuli, e.g. a grating target. For such targets the expres
14、sion (1.3) defines also the modulation. Between Eq. (1 .l) and Eq. (1.3) there exist the following relations m = C/(2 + C) for targets brighter than their background, and C = 2m/(l - m) m =C/(2 - C) and C = 2m/(l + m) for targets darker than their background. The contrast of characters on active dis
15、plays, e.g. VDUs, are usually defined by the ratio of two K= - Lb for L, Lb for Lb Lt b K= - L, (1.5) where Lt and Lb are defined as above. So far, there is no accepted specification of the location and spatial extent of the area where the target luminance Lt should be measured. The luminance modula
16、tion of rasters istheoretically and experi- CIE 95 92 U 9006345 0004488 648 -5- mentally analysed by outer and mean contrast of characters by the following expressions: resulting in the definition of local inner, local nr outercontrast Ka = - L Lb =2 innercontrast Ki = - (1.6) (1.7) where Lm = maxim
17、al luminance of a bright character L2, L1 = luminances of two adjacent raster points. The inner contrast of critical detail is one of the determinants for the effect of contrast on visual identification tasks. The size of the areas, where the luminances Lm, L2 and Li should be measured, is only one
18、picture element (pixel). The mean contrast K is defined by Eq. (1.4) and (1 -5) where Lt is the mean luminance of the whole character. Fig. 1 - 1 shows an ex- ample of a raster character. The luminance profile through the middle part of the character M is depicted in the lower half of Fig. 1 .l. Her
19、e the luminances Li, L2, Lm and Lb are localized. The mean contrast K is 15. On the basis of detection, identification and search experiments on VDUs it is shown by that the inner local contrast of characters is the deter- minant for the effect of contrast on visual performance and acceptance. 1.3 D
20、etection probability per glimpse The threshold contrast in Eq. (1.2) has to be measured under reference condit- ions according to CIE Publ. 19.21.1. But one gets a fluctuation of the threshold con- trast e from observation to observation.Therefore one has to assume a probabili- ty P of the target be
21、ing seen. P = 0.5 usually defines the threshold contrast e. For a presentation time of approx. 0.3 sa probability per qlimpse P, can be defined be- cause this presentation time simulates a single fixational pause. A plot of Pg as a function of VL or log VL is usually called a frequency-of-seeinq cur
22、ve. The visibility VL is defined by Eq. (1.2). A great number of such frequency-of-seeing curves have CIE 95 92 9006345 0004489 584 -6- been measured over the years by Blackwell and co-wrkersl.,. as well as by oth- ers. cd ln2 - 60 50 Lo 30 10 Y) O - L L - lmm Fiq. 1.1 : Luminance profile of the com
23、plex details of the character “M“ measu- red on a 1 5“-display. The scanned profile shows a typical configurati: on “pixel of the character-gap-pixel of the character“. The illuminan- ce of the display was 12 Ix (from The important conclusion from all the experiments is discussed with respect to Eq.
24、 (1.10) which is derived in the following from Eq. (1.8). For the mathematical description an equation is selected that has been suggested by several authors (BrirdIey-, Quick1.lo, Robson e Oe (1.15) (1.16) Thus with small numbers of objects and large search fields compared to o, there is a dramatic
25、 effect of extrafoveal guidance while for sufficiently large k values quasi- random search degenerates into perfect random search. 1.5 Modification of the visibility concept for highly suprathreshold stimuli 1.5.1 Visual performance. The term visual performance is used for a variety of performance s
26、cores such as frequency of seeing, detection time, detection dis- tance, reaction time, speed and accuracy of visual work. Although each of these measures may be relevant for a specific task they may be interrelated and a more general quantity would be desirable.*. It is the consensus of the committ
27、ee that a crucial etement of CIE Publ. 19.2 is the introduction of a contrast metric of visibility with the Visibility Level as a unifying stimulus intensity for the threshold zone of vision. Lighting parameters affect ei- ther the contrast or the contrast sensitivity, or both, and can thus be treat
28、ed as modifiers of the Visibility Level. CIE 95 92 9006345 0004496 714 m - 13- At high suprathreshold levels of contrast the “perceived“ visibility of a task and its CIE Visibility Level are not necessarily the same. Evaluations of perceived con- trast, conspicuity, and readability of image quality
29、all indicate that there is an optimum value of contrast above threshold. Studies directly measuring suprathre- shold visual performance have shown that it will increase with both contrast and The model of visual performance (CIE Publ. 19.2) cannot predict the data of which are obtained at a realisti
30、c, reading-writing task. The “refa- tive visual performance (RPV)“ model of al performance will increase only slightly with luminance when the contrast is high and more strongly when the contrast is low. predicts that suprathreshold visu- The RPV model of Rea is based on suprathreshold sensory compr
31、ession, which has the foltowing form, first employed by Naka and Ruhton. In - - R Rma.T In + k“ (1.17) where R = response Rmax = maximum response I = stimulus intensity n = exponent k = stimulus intensity producing half of maximum response. This expression is quite robust in describing visual respon
32、ses to luminous rnodula- tions. The exact values of n, k and R, in Eq. (1.17) will vary with the experiment. With respect to the reading-writing task described in Rea1.* a measure of visual performance VP is defined by the reciprocal of the time taken to read each refer- ence list of a particular co
33、ntrast at a particular adaptation luminance. Eq. (1.17) is rewritten in the following form (1.18) where: AL is the absolute value of a luminous increment or a luminous decrement from the threshold criterion at a given adaptation luminance, and VPmaxthe maxi- m u m pe rf o rm a n ce. CIE 95 92 81 900
34、6345 0004497 650 -14- The relationship between visual performance, which is the reciprocal of the total time to compare two number lists (US) at each reference sheet background lumi- nance (Lo) and contrast (CJ, are shown in Fig. 1.4. Best fitting curves using Eq. (1.18),which is based on Eq. (1.17)
35、, weredrawn through the four datsetsto help illustrate the different trends and relationships. The best estimates of the three parameters VP, n, and k at each adaptation luminance are discussed by Rea -la. CONTRAST Fis. 1.4: Performance plotted as a function of contrast (scaled logarithmically) at t
36、he four background luminances used in the numerical verification task. The reciprocal of the time to compare the reference and response lists (US) is used as the performance measure. The solid lines are best fitting curves using Eq. (1.18) which is based on Eq. (1.17), and the four contrast threshol
37、d values in Real.l* (Table 3). CIE 95 92 B 900bL45 0004498 597 -15- Although the model still requires extension and testing, it seems unlikely that the model is grossly inaccurate in representing suprathreshold visual performance. Visual response compression, which forms the basis of the model, is a
38、 widely ob- served phenomenon. The model makes predictions consistent with documented trends in threshold and suprathreshold behaviour1.18. Because the CIE model of visual performance is not justified by the data obtained by the experiments of Rea some appropriate modifications, e.g. equations simil
39、ar to Eq. (1.17) must be considered. 1.5.2 Complex suprathreshold stimuli. According to the reference conditions in CIE Publ. 19.2 threshold contrast should be measured with isolated simple stimuli on a uniform background as shown in Fig. 1.5a. Here, and in Fig. 1.5b. c, the lumi- nance in the immed
40、iate surround of the target is the same as the luminance La to which the observer is adapted. A more complex stimulus condition is shown in Fig. 1.5b where the two halfs of the target have different luminances Land L + AL, respectively. threshold contrast AVL which is the smallest perceptable differ
41、ence of lumi- nances for the detection of the target details. suggested a method for the evaluation of the Using a model of Adams S L U L Y a E 4.4 .! U VI I 35 .- 3; 1 t I O 10 1 O0 150 Angle off fovea (m rad) I Fis. 2.4: Prediction of sizekontrast thresholds as a function of retinal image position . mean observed data (from Ref. 2.61). - predictions from ORACLE. I 0.08. 4v rn 2 o 2 4 6 8 Boundary width (m rad) Fiq. 2.5: Predictions of the mean effect of image quality on detection thresh o Ids. mean observed data (from Ref. 2.62). - predictions from ORACLE.
