m > | P - se B iem cmn = re pi rare rene ^ Apa SA 
AGED LBS S M ELA PAR i on) CORSA n M SERES ER za 
errors may vary during observations on one photograph, although it is 
known that such errors may vary unexpectedly.  Trinder (1972) proposed 
a method of observation which would eliminate such errors. Roger and 
Mikhail (1969) have shown that the subjective centre of a target is 
significantly influenced by non-homogeneity of the background, and 
therefore asymmetric backgrounds must be avoided when placing ground 
targets. For the purpose of deriving relative variances of coordinate 
observations as given in this paper, pointing precisions only need be 
considered. However, the possibility of systematic errors occurring 
in pointing observations should be recognized, and further research 
should be encouraged to determine their full extent, and whether 
corrective measures must be taken. 
2.2 Procedure for Investigating Relative Variances 
Precisions of pointing to blurred targets (Trinder, 1971) were 
expressed in terms of target size and slope of the target density 
profile. Such results were theoretically related to ground target 
sizes and the imaging quality of the photogrammetric system, by the 
computation of convolutions of target luminance profiles and point 
spread functions of various sizes, representing the imaging quality of 
photogrammetric systems. The point spread functions adopted were 
Gaussian Functions, this choice being made for two reasons. Firstly, 
spread functions of the observed blurred targets were Gaussian 
Functions. Secondly, there is ample evidence (Wolfe and Tuccio, 1960), 
(Hempenius, 1969) (later confirmed by Trinder (1974)) that the shape 
of the combined Modulation Transfer Functions (abbreviated to MTF) of 
photogrammetric systems normally closely approximates the Fourier 
Transform of a Gaussian Point Spread Function (itself a Gaussian 
Function). It was considered that this study should be based on spread 
functions which are representative of those occurring in practice and 
therefore the Gaussian Function was the necessary choice. 
Before progressing further it should be pointed out that the work 
of Martucci (1972) has not been passed unnoticed. However, Martucci's 
study to derive appropriate relative variances across an image plane 
was based on measurements of widths of bar targets. It is believed the 
results of pointing observations to circular targets will be more 
relevant to this paper than bar target measurements, and therefore 
Martucci's results have not been included. 
Data derived in Trinder (1973) were primarily the optimum sizes of 
ground targets which would result in the highest pointing precision 
for a given image quality. Image quality parameters used were the 
Spread Function width (0) and the Frequency Limit (abbreviated to FL) - 
the intersection of the MTF of the system and the Modulation Sensiti- 
vity of the visual system. The second parameter was used in order 
that estimates of the effects of granularity could be included in the 
study, based on the technique developed by Hempenius (1964). Hempenius 
proposed that granularity reduced the Modulation Sensitivity of the 
visual system as a function of the object contrast and also the