A RELIABLE METHOD FOR ASSESSING THE IMAGE MEASUREMENT
QUALITY UNDER THE INFLUENCE OF IMAGE COMPRESSION
Zeng — Bo. Qian-
Ze — Xun. Geng
Zhengzhou Institute of Surveying & Mapping
66 Longhai Zhonglu
Zhengzhou ,450052
China
Comission 1,Working Group 1
KEY WORDS :Photogrammetry ,Remote—Sensing ,Softcopy Image Match , Quality.
ABSTRACT:
Up to now there is not any practical and available method to assess the image measurement quality under
the influence of image compression. This paper provides a reliable method which not only agrees with the
assessing criterion of the visual effect of decompressed images , but also can quantitatively point out the
geometric position distortion of pixels in decompressed images .
It is , therefore , suitable for high preci-
sion digital photogrammetry and other high precision digital image processing areas.
1 INTRODUCTION
The huge volume of digital image has been a seri-
ous obstacle for applications concered with storage
and transmission of image. So,it is very important
to compress image data. In a lossy compression, the
reconstructed images (or decompressed images)
differ from the originals. Measuring differences (or
distortions) between decompressed images and o-
riginals is very significant. There have been suc-
cessful methods for assessing visual quality of de-
compressed images known as subjective and objec-
tive criterions. These criterions are acceptable in
some applications, such as television and visual
like digital pho-
togrammetry and computer vision, these criterions
telephone. But in other areas,
are not acceptable, because they can not quantita-
tively point out the degree of geometric position
distortion of pixels in decompressed images in con-
trast to homologous pixels in original images. This
distortion, called geometric distortion, is most im-
portant in these areas, because it changes locations
of image points and, consequently, influences the
accuracy of objects reconstructed by photogram-
metric procedures. But ,up to now,how to measure
the degree of geometric distortions, or how to as-
sess the measurement quality of decompressed im-
160
ages, there is not any practical and scientific
method. T his paper provides a method for quantita-
tively assessing the measurement quality of decom-
pressed images. It not only agrees with the objec-
tive criterion for assessing visual quality of decom-
pressed imaegs, but also can quantitatively point
out the degree of geometric distortions of decom-
pressed images. It is, therefore, suitable for high
precision digital photogrammetry and other digital
image processing areas. The experimental results
also verify this conclusion.
2. ASSESSING METHOD
2. 1 LS image match. Least square (LS) image
match method is one which minimizes distance, in
mean while, it takes gecmetric distortions and ra-
diometric differences into account. It predicts dis-
tortional parameters according to the principle that
the square sum of grey value differences of homolo-
gous pixels in search whindows is minimal, makes
modification of parameters and finds out matched
points in search windows (Zhizhuo, w. 1990).
Assume the grey values of same object point is f
(24590) in target windows in original image and g
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996
Ça
res
If:
are
wh
hoi
of
tio:
gol
Sio.
C:
Th
fol
wh
Bu
low
wh
ran
pet
ter
wh
vat
do
