having more information contents, too,may be added
into processed result in after-processing.
T.QUANTITATIVE EVALUATION OF SYSTEM ENHANCEMENT
EFFECT FOR GEOLOGICAL LINEAMENT.
According to the Feedback Dynamic Recognition
Pattern, this paper has calculated respective IL
valus of various functions in common use for
enhancing lineament. On the basis of total
statistics, these values were obtained from 1500
sheet of images concerned with various physical
features. Due to lacking statistics to some
specific areas, thus in statistics this experiment
did not carry out area classification with variety
of physical feature. The grade I of signal Y, or
selected maximal grade in a number of X signal
grades responding to same target,was obtained from
respective X signals of 10 functions ( or series
connection programs). Statistic number of each
function was more than ten times number of I(C)J
distribution state, about above 140 strips of
lineaments.It can be known from examples stated by
section 5, that required statistic number need not
reach such large value in practice. The percentile
IL values of various functions obtained by this
experiment can be written in big or small sequence,
as follows (percent symbol does not be listed):
HIST NORMALIZE ( 43) ; CONVOLVE( 38) ; LOCAL! ENHANCE 35) ;
HIST!EQ(33) ;PLIM(32); SCALE(32) ; SELF-MULTIPLY ( 31) ;
KL-PRINCIPAL COMPONENT(30); HADMARD(29);FFT1D(28);
LOGARITHM( 27 ) ; KL’ TRANSFORM( 23) ;RATIO( 19) ; CLUSTER( 1
71) ; MSS! BAND-7 (26) ;MSS! BAND-5 (22) ;MSS' BAND-4(18).
These IL values concerned with original single
band images reflect visual recognizable and
responsive ability to image signals, namely, the
visual interpretation is considered as function
transform, these differences among IL values are
consistent with physical mechanism analysis
concerned with band property and geological
environment. The IL values of function-CONVOLVE
and LOCAL'ENHANCE belong to high level in system,
that is concerned with respective property of
themselves,and is consistent with most qualitative
evaluation conclusions. These quantitative IL va-
lues have also revealed difference between theory
and practice of some functions, such as function
-FFT1D or KL'TRANSFORM.
Putting various functions to participate in
parallel connection operation in big or small
order of their respective IL value, can obtain
parallel connection system (IL)p values from
formula (5), as shown in List 7-1.
LIST 7-1, parallel connection system CIL)p values
(function).
The parallel 1] 3] 31 4] 5] 6$] 1] v] 9] 10
function number
CILdp % 43 | 65 | 77 | 85 | 90 [ 93 | 95 | 96 | 97 | 97
The formula (5) has larger practicality. In
practice, people can usually obtain satisfactory
results for lineament only with 4-5 functions, from
List 7-1 , it can be known that corresponds to
having tapped 90% system gain potential; for some
images which are processed difficultly, that often
want over five functions to participate in
operation, but incremental gain signals are less,
that is consistent with retarded appearance of
(IL)p increasing velocity in this range shown as
List 7-1.
For those series connection programs through
practical check, according to name of function
66
participating in main-processing, whose IL values
can be arranged in order of big or small, as
follows: :
COMPOSITION OF KL'COMPONENT WITH ORIGINAL IMAGES
(45); CONVOLVE(44); NEGATIVE(42); KL PRINCIPAL
COMPONENT ( 41 ) ; LOCAL! ENHANCE ( 37 ) ;KL' TRANSFORM( 34) ;
SELF-MULTIPLY (34); RATIO(31).
These respective IL values of above-mentioned main-
processing functions, due to their having
participated in series connection operation, all
have respective improvement over IL value obtained
singly by oneself. If these programs sequentially
participate in parallel connection operation in
big or small order of their respective IL value ,
those obtained (IL)p values can be listed in List
7-2:
List 7-2 parallel connection system CIL)p values
(program).
The parallel 1] 2| 31.415 6| 718
program number
CILdp 96 | 45 | 69 | 82 | 89 | 93 | 96 | 97 | 98
That can be known from above, if we put only 4
programs to participate in parallel connection
operation, we can tap 9094 system gain potential.
Series connection programs and functions can also
participate in parallel connection operation
together.
8. CONCLUSION
For evaluating IPOS enhancement effect, the
qualitative evaluation was putted first
previously. Except that subjectivity and
localization in its weakness , this evaluation
hardly considered the actualities information of
original image, and only regarded "GOOD" and "BAD"
or "HAVING" and "NOTHING" of enhanced results as
standard, that is perplexed by surface phenomenon
easily. From these composite signals I(C)J, we can
carry out quantitative and qualitative evaluation
accurately. On the statistic basis of 14 kinds of
composite probability distribution results, we can
evaluate enhanced effect by single quantitative
parameter IL, and can consider all possible cases
overall and really.
The IL value represents function or system self
enhancement feature, thus can be used as a means
checking new functions or programs. Some functions
could be a series connection program composed of
various procedures, if strict measure leads to
increment of procedure number, as formula (6) its
enhancement effect could be just the opposite to
what one wished.
To use the series-parallel connection programs,can
give full play to system enhancement potential.For
image enhancement of designated target,we need not
employ all system functions, with reference to
those respective IL values of various functions or
programs which is concerned with designated target,
we. can compose a subsystem with definite functions,
whose effect can represent enhancement potential
of the whole system.
To counter the geological lineament,this paper has
recounted some quantitative parameters concerned
with system and functions, which have definite
directive action in the image processing dynamic
policy-making course,from which we can obtain many
lineaments enhanced results quickly and
effectively.
+ References from Books
Zhon Shuo-Yu, 1988. System Science guide.
soso. Press .Peking-China,PP.41-44,PP.50-58,
PP.96-98.
Each
are vc
volum
very 1
There
(head
recor(
adjust
are th
{il |
Scene
longit
( scan
Four
corres
Scene
betwe
merid
Angle
Scene