QUANTITATIVE METHOD BY INFORMATION THEORY FOR
EVALUATING IMAGE ENHANCEMENT BENEFIT OF IPOS
Hu Tinghui
(high level engineer)
Han XiQin
(engineer)
Institute of Seismology, State Seismological Bureau.
(Xiao Hong Shan, Wuhan, China)
Commission lll
ABSTRACT:
According to the method of Informatiom Theory and many practical statistical image enhancement examples
of the geological lineament, this paper has made quantitative analysis to the enhancement effect of the
Image Processing Operation System and its functions, thus solving a current problem that enhancement
effect can only be evaluated subjectively and qualitatively. Quantitative parameter calculated by the
Information Theory, or Information Level, has definite directive action
in the dynamic policy-making
course of image processing, and can decide correctly the practical effect of new functions or methods.On
the basis of quantitative analysis, the IPOS can make the most of image enhancement benefit quickly and
effectively by series-parallel connection programs.
Information, Processing, Enhancement, Signal, Evaluation, Entropy, Function, Probability.
KEY WORDS:
1 . INTRODUCTION
Photogrammetry and Remote Sensing are concerned
with informaton input-output and its being
processed, in this image information domain,
Information theory should have certain applied
potential. This paper first applies Information
Theory analysis method for evaluating enhancement
effect of the digital image processing.
To counter wanted fixed enhanced contents f
OL
image, this paper regards IPOS function as a
information translator, analyzes original image
actualities signal Y at input end, processed image
gain signal X at output end ( Fig. 1 ) and both
composite signal C. On the basis of statistics, 14
kinds of composite probabilities are calculated
from respective composite signals. To apply the
Channel Mutual Information Theory , we can
calculate that function Information Level IL from
composite probabilities, which can reflect Y
signals in X ones.
--] Functiom or F-—— — —
Translator Signal X
Signal Y
Fig.L Image information transform
Original image , or total of processed images, can
be regarded as information source producing
information or signal sequence. Owing to the
variety of physical features and whose seasonal
dynamic variation , the display degree of
actualities signal Y should have that randomness,
after being transformed by function, which should
also be existential with respect to gain signal X,
because of that image processing itself is a
random experiment. The randomness of input-output
signals is a theoretical premise of calculating II,
IL of function, or probability of its obtaining
information, can be used for evaluating degree of
its actually reflecting the orginal image
information. The calculation of IL is concerned
with statistics, in order to attaining the firm
statistics purpose, a Feedback Dynamic Recognition
Pattern by progressively calculating is applied.In
the circulative calculation process,the statistics
is incremental. Owing to the mutual causality
between input statistics and output calculated
value, through observing the dynamic variation of
calculated value IL in various process, the IL
63
value can be obtained with higher accuracy in
stable state of variation.
IPOS enhancement potential is reflected by
carrying out series-parallel connection programs
of functions.This paper presents IL budget formula
for series-parallel connection, which can be used
for calculating and analyzing IL value of system.
under conditions of applying this budget formula,
various function IL values related to enhanced
contents can be used for determining series-
parallel connection programs, fixed functions and
applicable function number participating in
programs, that helps system to obtain deserved
information contents with 100/100 reliability
approximately.
2. THE PROBABILITY DISTRIBUTION COMPOSITION FORM
OF COMPOSITE SIGNALS I(C)J
The information received or processed by men
generally is fuzzy one for the most part. After
processing to the designated target (such as
lineament) with some one system function , the
enhanced result in image, or gain signal X, may be
decided fuzzily according to the following display
grades:
1(it represents distinct ); 2(obvious); 3(darkish);
4 (obscure); O(not have result).
On the image processed by a certain function, a
result represented by signal I(C)J is obtained, in
which :
J is divided into above-mentioned 1,2,3,4,0 grade,
it represents the grade of gain signal X enhanced
by this function;
I is divided into above-mentioned 1,2,3,4 grade.It
represents the grade of actualities signal Y. With
respect to same target concerned with signal X,the
original image should contain hidden actualities
signal Y which can be enhanced by system. I is the
attainable enhanced grade processed by system,
namely is the maximal grade selected in a number
of X signals processed by various system functions
and responding to same target. I(C)J is composed
of signal Y and X.every designated target (such as
lineament)should have a result I(C)J, according to
a certain number of targets, to count the
respective number of various I(C)J produced by
this function, can obtain following I(P)J
probability distribution composition form composed
of various I(P)J:
