International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
Color R ; Regi ,
Preprocessing egion Contour
Image Á Y » segmentation > Coding
The base color Region
»| approximation of
the base color
Texture y Compressed
Coding » ( ) Image
A
The Subordinate colors Subordinate
cC
olor encoding
Figure 1 - Block Diagram of the proposed encoder
3. PROGRESSIVE TRANSMISSION
For slow communication channels, such as aerial wireless links,
it is possible to improve the efficiency of the transmission if the
information is sent progressively. In such a process we reduce
the size of the regions gradually. We start with the transmission
of large regions and then sub-divide the regions to improve the
resolution of the received image. The first compressed image
has only large regions. In each consecutive stage, new and
smaller regions are added to produce a more refined image.
It should be noted that the statistics of the contour's chains may
be different in each stage. The length of the chains decreases in
later steps, so that the code for end-of-chain appears more
frequently. This implies that each stage may need different
Huffman codes. In this work the selected code is the one that
minimizes the number of bit per direction. From the gathered
statistics, 16 different Huffman codes were found. Thus only
the index of the code is transmitted to the receiver.
4. COMPRESSION RESULTS
The algorithm for aerial imaging was implemented with the
Green as a base-color. Each region of the base color was
approximated using a second order polynomial function. The
subordinate colors were approximated using a first order
polynomial expansion of the base color. For comparison,
images compressed by the proposed method and by the JPEG
algorithm are shown in Figures 2 and 4. Results of progressive
transmission are shown in Figures 3 and 5.
5. CONCLUSIONS
This technique achieves a high compression ratio with higher
image quality compared to the JPEG algorithm, especially for
images of aerial mapping, where the blockiness effect of JPEG
is very noticeable. Unlike the arbitrary 8x8 blocks of JPEG,
here the regions are based on the natural segmentation of the
mapping, providing an efficient tool for high quality content-
based compression. The quality of the image can be controlled
using progressive transmission — in each stage a more refined
706
image is obtained. Results of the compression for various
images have been shown, and specific properties of the
algorithm have been analyzed and discussed. Our conclusion is
that a regional color correlation approach could be superior to
the traditional decorrelation methods, especially for aerial
imaging.
ACKNOWLEDGMENTS
This research was supported in part by the HASSIP Research
Program HPRN-CT-2002-00285 of the European Commission,
by the Technion VPR Maas Fund, and by the Ollendorff
Minerva Center. Minerva is funded through the BMBF.
REFERENCES
H. Freeman, “On the encoding of arbitrary geometric
configuration”, IRE Elec. Comp, EC-10, pp. 260-268, 1961.
L. Goffman and M. Porat, “Color image compression using
intercolor correlation", IEEE ICIP 2002, Rochester, NY, 2002.
M. Kocher and M. Kunt, *A contour-texture approach to picture
coding", in Proc ICASSP-82, Paris, France, May 1982.
H. Kotera and K. Kanamori, “A novel coding algorithm for
representing full color image by a single color image”, J. of
Imaging Technology, Vol.16, pp. 142-152, Aug. 1990.
J. O. Limb and C.B Rubinistein, “Statistical dependence
between components of a differentially quantized color signal”,
IEEE Transaction on Communication, Vol. COM-20, pp. 890-
899, Oct. 1971.
H.Yamaguchi, “Efficient encoding of colored pictures in R,G,B
components”, IEEE Trans. on Communication, Vol.32, pp.
1201-1209, Nov. 1984.
D. C. C. Wang and Anthony H. Vaganucci, “Gradient Inverse
Weighted Smoothing Scheme and Evaluation of Performance",
Computer Graphics and Image Processing, Vol. 15, Feb 1981.
Intern
Figur
