SUBJECTIVE SHARPNESS OF IMAGES 575 



the vertical direction. Inasmuch as the effect is fairly small, and 

 found only with the less sharp images, we shall leave it as another 

 problem in physiological optics. 



With an actual television image this small skewness would probably 

 be reversed by the attendant coarsening of the scanning line structure. 

 We do not know how much to allow for annoyance caused by visibility 

 of the line structure. Taking our best estimate * of the height of the 

 figure of confusion which would be equivalent in vertical resolution to 

 a just noticeable pattern of scanning lines, we may say that for the 

 uppermost curve in Fig. 5 the scanning line structure would not be 

 noticeable except possibly for the shape marked 2/5. For the central 

 curve the line structure would be noticeable for all shapes except 

 possibly the one marked 5/2. It appears that the skewness and the 

 line structure vanish together as the sharpness is increased. 



Figure 5 demonstrates that equality of horizontal and vertical 

 resolutions is a very uncritical requirement on the sharpness of an 

 image, especially of a fairly sharp one. An image somewhat better 

 than present television grade, exemplified by the uppermost curve in 

 Fig. 5, shows a remarkably wide tolerance in this respect. Its figure 

 of confusion could be three times as high as wide, or three times as wide 

 as high, yet any intermediate shape between those two extremes would 

 yield an equally sharp image to within one liminal unit. Under the 

 ordinary conditions of television viewing the difference would be even 

 less marked than that. This would imply that if the square figure of 

 confusion simulates a television image of say 500 lines, then the 

 number of lines could be changed to any value from about 300 to about 

 850 without altering the sharpness by as much as one liminal unit, 

 under the condition, of course, that all the other pertinent factors, such 

 as frequency band width and number of frames per second, remain 

 unchanged. 



The curves in Fig. 5 represent the averaged responses of fifteen 

 observers each viewing five different motion picture scenes. Each one 

 of the five selected shapes of figure of confusion was shown with each 

 other one as a pair, a total of ten pairs. The observer was asked to 

 identify which member of each pair he judged to yield the sharper 

 image, or to report "no choice" if he judged them to be equally sharp. 

 The pairs were scheduled in irregular order, and the observer could 

 have the aperture shape shifted at will. The observers were instructed 

 to consider the whole image area without undue regard for some fea- 

 tures to the neglect of others. 



* Engstrom ^ estimates that the scanning line structure becomes just noticeable 

 when the spacing of the lines subtends an angle of 2 minutes at the observer's eye. 

 In section 5 we show that the equivalent figure of confusion has a height 1.9 times as 

 great as the spacing of the scanning lines. 



