THE GREY SCALE OF A TELEVISION PICTURE

Transcription

1 RESEARCH DEPARTMENT THE VARATON N THE VSBlTY OF NTERFERENCE OVER THE GREY SCALE OF A TELEVSON PCTURE Report No. T-092 ( 1962/27) w. Ko Eo Geddes M. A. Grad 0 Eo Eo (D. Maurce)

2 The Report s the propert1 or the Brtsh Broadcastng Corporaton and aa1 Dot be reproduced n any form wthout tbe wrtten perm18son or the Corporaton.

3 Report No. T-092 THE VARATON N THE VSBLTY OF NTERFERENCE OVER THE GREY SCALE OF A TELEVSON PCTURE Secton Ttle Page SUMMARY NTRODUCTON " DESCRPTON OF THE EXPERMENTS RESULTS DSCUSSON OF RESULTS. 7 5 CONCLUSONS REFERENCES. " 8

4 July 1962 Report No. T-092 ( 1962/27) THE VARATON N THE VSBLTY OF NTERFERENCE OVER THE GREY SCALE OF A TELEVSON PCTURE SUMMARY Subjectve measurements have been made n order to establsh the relatve vsblty at dfferent ponts on the grey scale of nterference supermposed on a televson pcture. Results are compared and contrasted for a number of dstnct types of nterferng sgnal. 1. NTRODUCTON t s sometmes requred to know the relatve subjectve vsblty of nose or other nterference at dfferent grey-levels* of a televson pcture gven the objectve magntude of the nterference. A relatonshp of ths type was recently establshed 1 for an nterferng sgnal that produced a statonary coherent pattern; a dsplayed test card had numerals supermposed at a barely vsble level of contrast on the fve plan squares of the step wedge. By measurng the values of contrast at whch the numerals could just be dentfed t was possble to establsh the relatonshp between the vsblty of a pattern of ths type and ts mean grey-level. However when attempts were made to apply ths relatonshp to problems nvolvng the vsblty of random nose dscrepances were found suggestng that the relatonshp was not unque as had been assumed but depended on the nature of the supermposed sgnal. Ths report descrbes an nvestgaton of ths dependence n whch the vsbltes of several types of supermposed sgnal were measured as a functon of grey level. 2. DESCRPTON OF THE EXPERMENTS Fg. 1 shows two typcal dsplays used n the experments. Each conssts of a normal test card nto whch two comparatvely small test areas have been electroncally "nlad"; consequently the observer's eye was mantaned n a state of adaptaton approprate to the vewng of televson pctures. The test procedure wll be descrbed for the case llustrated n Fg. l(a) n whch random nose was dsplayed n both squares. One of the team of seven observers that took part n the tests sat n front of a 21 n (53 cm) montor at a dstance from the screen equal to fve tmes the heght of the dsplayed pcture. The engneer supervsng the test could control the mean grey-level of each square test-area and the ampltude of the nose waveform dsplayed n the left-hand square. *The qualltatlte term 'grey letel' s used here n order to descrlbe the brghtness or any part of a plcture relatte to the peak-whte brghtness.

5 2 ( a) Cb) Fg. - Forms of dsplay used n the tests (ab) Ran dom no se n test squ ares () Coherent pattern n test squares

6 3 An attenuator controllng the ampltude of the nose waveform dsplayed n the rghthand square was operated by the observer whose task was to adjust ths attenuator untl he judged the nose to be equally vsble n the two squares. For the purposes of ths nvestgaton t has been found convenent to descrbe grey levels n terms of a percentage scale of vdeo voltage havng the voltage at blankng level as zero and that at peak whte as 100 per cent. The quantty so defned wll be referred to as the 'pcture-sgnal level'. Throughout the test the mean pcture-sgnal level wthn the left-hand square was mantaned at 50 per cent and the supermposed nose-voltage was mantaned at a predetermned ampltude. The mean pcture-sgnal level of the rght-hand square was successvely set to a number of values and for each value the observer adjusted hs attenuator takng hs own tme to do so; the attenuator settng was then recorded by the supervsng engneer. Hgh and low values of pcture-sgnal level were presented n a random sequence n order to mnmze any tendency for the observer to be nfluenced by hs prevous decsons. Each test ncluded a control comparson n whch the mean pcture-sgnal level was the same for the two squares thus provdng a check of equpment calbraton. The contrast range of the dsplay was eghty to one bolmded by brghtnesses of 20 ft-l and 0 25 ft-le The relatonshp between brghtness and pcture-sgnal level for the partcular montor used s shown n Fg. 2. The settng of the brghtness control was such that no scannng lnes were vsble n areas scanned at blankng level. The brghtness of these areas was 0 25 ft-l although the ambent lghtng caused the screen to reflect a brghtness of only 0 13 ft-l when the montor was swtched off; ths dscrepancy s smlar to that encountered durng the nvestgaton descrbed n Reference 1 and s thought to be due to flare n the tube face. The ambent lghtng caused a brghtness of 0 35 ft-l to be reflected from a whte card held n front of the tube face. t was found necessary to restrct the pcture sze n order to prevent the llumnated part of the raster from extendng beyond the edges of the screen because exctaton of the phosphor on the sdes of the tube ncreased the brghtness n dark parts of the pcture wth consequent loss of contrast. Three types of random-nose waveform were used n the tests a low-pass flter beng used n each case n order to restrct the bandwdth to 3 Mc/s. () 'Flat' nose. Ths was the output from a nose generator* desgned by the B.B.C. Desgns Department. t utlzes an llumnated photo-multpler as ts nose-generatng element and s desgned to produce an output havng a unform energy-spectrum over the range of frequences extendng from 100 kc/s to hgher than 3 Mc/s. () 'Trangular' nose. Ths nose was derved from the flat nose by passng t through a crcut Whose gan was proportonal to frequency. () 'Recever' nose. Ths was the output of a vestgal-sdeband televson recever fed wth a c.w. sgnal at ts vson-carrer frequency. *TJP. 684/501 P.

7 4 ;1 f----~+_. =t:. -r- ~_-----L-_~. --.t~ '----~ / ~2-3 > «...J Q. '" o 0 5 ~~~~~~~+---/f/~~---~--- -J- f------c------; ~/----~--~---- L_V r ~-~--l r oj -~-"'-----' RGH 1 NESS =mr- ZERO PCTURE-SG\AL LEVEL = 0-25 FOOT-LAMBERTS ZERO (&LANKNG LEVEL; PCTURE-SGNAL LEVEL (PER CENT) t PEAK WHTE Fg. 2 - Contrast 1 aw of man tor used The rato between the maxmum peak-to-peak value and the r.m.s. value was measured for each type of nose. The peak-to-peak value was measured by observng a waveform montor and the r.m.s. value was measured by means of a thermocouple; a snusodal waveform was used for mutual calbraton of the two measurng nstruments. Fbr all three types of nose the rato was measured as 18 db. All the tests nvolvng nose were carred out at the same degree of vsbl ty. The reference sgnal defnng ths vsblty was flat nose supermposed on a mean pcture-sgnal level of 50 per cent and havng an r.m.s. value 35 db below the black-to-whte ampltude of the pcture waveform; ether ths reference sgnal or a sgnal adjusted to have the same vsblty was dsplayed n the lefthand square for all tests. n addton to tests of the type already descrbed n whch vsblty was related to grey level for a partcular type of nose the nvestgaton ncluded tests to establsh the relatve vsbltes of dfferent forms of nose at dentcal values of grey level. A test of ths type was carred out at an early stage of the nvestgaton n order to determne whether there was any measurable dfference n vsblty between flat nose and recever nose. Flat nose was dsplayed n the left-hand square and recever nose n the rght-hand square. The two

8 5 squares were mantaned at a common value of mean pcture-sgnal level whch was successvely set to be 10 per cent 50 per cent and 100 per cent. To mantan the nose n the left-hand square at the standard vsblty ts ampltude was approprately adjusted for the tests at pcture-sgnal levels of 10 per cent and 100 per cent. The result of ths test establshed that there was no measurable dfference n vsblty between the two types of nose at any of these three grey levels and recever nose was not used n any further tests t beng assumed that the results for flat nose were also vald for recever nose. A further comparson establshed the dfference n vsblty between flat nose and trangular nose. Ths comparson whch was made wth both squares at a mean pcture-sgnal level of 50 per cent showed that the r.m.s. value of trangular nose should be 6 db hgher than the reference value of flat nose n order to be equally vsble. Thus n the subsequent measurement of the relatonshp between vsblty and grey level for trangular nose the reference sgnal conssted of trangular nose supermposed on a mean pcture-sgnal level of 50 per cent and havng an r.m.s. value 29 db (35 db - 6 db) below the black-to-whte ampltude of the pcture waveform. n the measurements nvolvng statonary coherent patterns two forms of test sgnal were used. The frst of these produced a dsplay of the form llustrated n Fg. l(b) and conssted of a square wave havng a fundamental frequency of about 250 kc/s sutably locked to the lne-scannng frequency. The bandwdth of ths sgnal was lmted to 3 Mc/s by means of a low-pass flter. The other form of test sgnal was a sne wave havng a frequency of 2 Mc/s; ths sgnal also was locked to the lne-scannng frequency so that vertcal strpes were dsplayed n the test squares. When these test sgnals were used the peak-to-peak ampltude of the reference sgnal dsplayed n the left-hand square at a mean pcture-sgnal level of 50 per cent was 1 per cent for the 250 kc/s waveform and 2~ per cent for the 2 Mc/s waveform. The respectve vsbltes of the patterns produced by these two reference sgnals were smlar to each other and to that of the atandard nose sgnal already specfed. 3. RESULTS Fg. 3 shows the results that were obtaned for flat nose trangular nose and the two patterns just descrbed. The abscssae represent values of the me~pcture-sgnal level and the ordnates represent n decbels the relatve ampltudes of the test sgnal that were found to be necessary for the achevement of constant vsblty. The vertcal relatonshp between the curves has been chosen n order to show ther smlartes and dfferences; the absolute ampltudes represented by them may be derved from the arbtrary scale of decbels by referrng to the table n the lower rght-hand corner of the fgure. Each curve represents the mean of results obtaned from the team of seven observers all of whom were engneers accustomed to makng crtcal apprasal of televson pctures. The standard devaton of ndvdual results from the mean result had an average value of about 1 db for mean pcture-sgnal levels of 25 per cent or more but ncreased to an average value of about ~ db for lower values of mean pcture-sgnal level.

9 1 \ -r T\-l '----~-----'---7--;.-L T-YPE OF DSPLAY AS OdS' RELATVE -6 \ ' / TO PEAK PCTURE- SGNAL A.WLmJ)~- -7 ' 1 './ / -----FLAT NOSE ----~- - '-...</ TRANGULAR NOSE -2SdS.' R.M.S. "t--_ - 8 r ~--' ;-=-.".--"'---- :_PA-ATTTTE"R"N.-c~r--~---~-+-~&l<-eh---P.OH-ERl'l--- -~ PEAC-TO- 2Mc/. PATTERN - 26dS J PEAK - 9 O~ --'---:-O:---'---=-20::----'----:30:'-:-----:4-=0----:SC::0---60~----:7:-' ~ MEAN PCTURE-SGNAL LEVEL. 120 Fg. 3 - Ampl tudes of nterference producng a constant vsbl ty as a functon of the mean grey-level of the background t wll be seen that a sngle curve has been drawn for the two patterns descrbed and that only at values of mean pcture-sgnal level below 20 per cent s the curve for trangular nose shown as dfferng from the curve for flat nose. Where two curves have been assumed to be dentcal the approxmaton nvolves a maxmum error of 1 db and a mean error of abou~ ~ db. The grey level at whch flat nose s most vsble has been measured under threshold condtons of vsblty by Klvngton et al. 2 Maxmum vsblty was found to occur for a mean brghtness equal to 8'6 per cent of the maxmum brghtness whch corresponds to a pcture-sgnal level of 35 per cent n Fg. 3; ths result s therefore n good agreement wth the results of the present nvestgaton. A dsplay of random nose dffers from a statonary pattern n beng ncoherent n both space and tme and random-nose waveforms are al30 characterzed by hgher crest factors than those of sne waves and square waves. n an attempt to dscover whether anyone of these propertes s prmarly responsble for the dfference n shape between the 'nose' and 'patterns' curves of Fg. 3 tests were carred out usng waveforms that ncluded only two of these three propertes. A sgnal havng a crest factor of about 6 db was produced from the flat-nose sgnal

10 7 by peak-clppng. The results obtaned wth ths test sgnal when expressed n the form adopted n Fg. 3 defned a curve whose shape was very lke that of the curve for flat nose whch had a crest factor of 12 db; thus the value of the crest factor does not appear to be sgnfcant. Another test sgnal was produced by means of a televson camera from a transparency of enlarged photographc gran. Ths sgnal had a crest factor smlar to that of random nose and produced a spatally ncoherent dsplay whch however dffered from that of random nose n beng statonary. When the results obtaned wth ths test sgnal were expressed n the form adopted n Fg. 3 the curve they defned could be made to le wthn ~ db of the 'patterns' curve for values of the mean pcture-sgnal level of up to 50 per cent. However f ths was done the curves dverged for hgher values of mean pcture-sgnal level; that for the statonary pseudo-nose was then parallel to the 'nose' curve. 4. DSCUSSON OF RESUDTS When the results of the tests are expressed n the form shown n Fg. 3 they are applcable only to the partcular montor used and to a partcular settng of ts brghtness control. n Fg. 4 the results have been expressed n a more SO '-25 f :q) :::>... ~... > ;: c... a: o \': ' '.':' ':' ' ~_~~ -\+~-: q ' : ' : \ 1 \ \ \{. ~ 1 ~ "J 1"--< :: \ LJ ' ' ' :~ ": " ' T-+~1'" ~ '--+- \ [ : 1 r--"-=.::..-_ : or;;; ::::'-:;~.L -r--t-l. : \ BRGHTNESS B AS PERCENTAGE OF PEAK BRGHTHESS ~ l 1 T 1 11 f T T 1 : 1 1 : 1 1 \ Fg. ~-Varaton wth background grey-level of the fractonal brghtness-dfference producng a constant vsbl ty of nterference Vsblty of numerals (Ref_l) --- Statonary coherent patterns Ran dom no se

11 8 generally applcable form by transformng the varables of Fg. 3; ths has been done by makng use of the relatonshp between the brghtness of the montor screen and the pcture level whch s shown n Fg. 2. The abscssae of Fg. 3 have been transformed to values of mean brghtness and the ordnates to relatve values of fractonal brghtness-change. Nether transformaton s vald when the supermposed test-waveform contans very large excursons and for ths reason all the results measured at a mean pcture-sgnal level of 10 per cent have been omtted from Fg. 4 as has the result for trangular nose at a mean pcture-sgnal level of 15 per cent. The fgure also ncludes a curve plotted from the results of the nvestgaton nto the vsblty of numerals (Reference 1) and ths curve may be seen to be n good agreement wth the curve for statonary coherent patterns. All of the curves exhbt a horzontal porton at hgh values of mean brghtness ndcatng that the Weber-Fechner law s obeyed under these condtons; at lower levels of brghtness there s a progressve ncrease n the fractonal brghtness-change requred for the achevement of constant vsblty. Ths concluson s n general agreement wth that reached by Bllard3 who measured thresholds of vsblty for patterns under somewhat dfferent expermental condtons. t may be seen from Fg. 4 that the vsblty of statonary coherent patterns obeys the Weber-Fechner law for all values of brghtness exceedng about 10 per cent of the maxmum but that t s only for values of brghtness exceedng about 30 per cent of the maxmum that the vsblty of random nose vares n ths way. The 'patterns' curves of Fgs. 3 and 4 have been found to be equally applcable to ether of two statonary coherent patterns derved from vdeo sgnals whose respectve fundamental frequences were n the rato of eght to one; smlarly for brghtnesses exceedng 3 per cent of the maxmum the 'nose' curves have been found to be applcable to the dsplay correspondng to ether of two waveforms one of whch was the tme dervatve of the other. t therefore seems unlkely that the general shape of the vdeo spectrum above 100 kc/s affects the shape of the relatonshp between vsblty and grey level. t would appear that at low greylevels the 'patterns' curves are applcable to any statonary dsplay even f t s spatally ncoherent whereas at hgh grey-levels the 'nose' curves are applcable to any spatally ncoherent dsplay even f t s statonary. 5. CONCLUSONS t has been found that under condtons of vsual adaptaton approprate to the vewng of televson pctures the functon relatng the vsblty of random nose to ts mean grey-level s markedly dfferent from the correspondng functon for statonary coherent patterns. The dfference s such that the departure from the Weber-Fechner law at low grey-levels s greater for nose than for patterns and extends over a larger part of the grey scale. 6. REFERENCES 1. "The Vsblty of Small Brghtness Dfferences n Televson Dsplays" Research Department Report No. T-083 Seral No. 1962/6.

12 9 2. Klvngton T. Judd D.L. and Meat yard L.R. "An nvestgaton of the Vsblty of Nose on Televson Pctures" Post Offce Rado Report No (1954). 3. Bllard P. "Les Seuls Dfferentals de Lumnance en Televson" Acta Electronca Vol. 2 Nos.1 and 2 p or Optca Acta Vol. 5 No. Hors sere p. 283 Jan BRH