PEE ES ara 4
20.
and also as a general measure of image quality, since the FL would
normally occur in the range of significance quoted by him. Hufnagel
also successfully applied the Wiener Spectrum in his studies, for
modifying visual performance due to the effects of granularity. As
stated earlier the Wiener Spectrum is being studied in a current
research programme for combining granularity effects and MTF's.
Scott (1968) in an extension of Hufnagel's work found that a 5%
difference in the width of the spread functions of two photographs
can be detected when the photographs are of the same scene, while
this factor increases to 10% for photographs of different scenes.
These figures approach the measuring precision of MTF's of photo-
grammetric systems, and thus demonstrate the importance of accurate 6G
determinations of MTF's for future image quality assessment. i
With Scott's data, it was also possible (Trinder, 1973) to prove
that the FL for grainless and grainy photographs are approximately
equal (see sec 2.4). This adds weight to the earlier findings of
Hufnagel, regarding the significant section of the MTF curve for
purposes of interpreting a photograph. Though attempts have been
made by some researchers to include the full shape of the MTF curve
in image quality assessment, significant agreement has been obtained
in several works through the FL.
4.4 Conclusions
(i) An understanding of factors affecting interpretability
involves many parameters» few of which have been investigated to date.
Some knowledge has been gained of the photographic object sizes for
detection and recognition, (ii) For ranking of photographs in terms
of quality, it appears that the high frequencies are most significant
in determining which photographs are considered of higher quality than
others, (iii) In addition, observers are sensitive to variations in
MTF's which approach the measuring precisions of MIFs.
5. Final Remarks
Current research experiments on the significant parameters of
image quality reveal that there are as yet many unknown factors to be
investigated. Generally attempts have been made to use a single
parameter e.g. Frequency limit or spread function width, to express
performance in terms of image quality. A more thorough parameter may
be more suitable, which describes the shape of the MTF curve and the
effects of granularity. Finally this parameter must be related to
various visual tasks in a general manner so that the effects of using
different components can be estimated.