The Wavelet Digest Homepage
Return to the homepage
Search the complete Wavelet Digest database
Help about the Wavelet Digest mailing list
About the Wavelet Digest
The Digest The Community
 Latest Issue  Back Issues  Events  Gallery
The Wavelet Digest
   -> Volume 7, Issue 9


Thesis: Monochromatic Still Image Compression
 
images/spacer.gifimages/spacer.gif Reply into Digest
Previous :: Next  
Author Message
zhl (zhl@ns.glc.cn.net)
Guest





PostPosted: Thu Aug 27, 1998 2:09 am    
Subject: Thesis: Monochromatic Still Image Compression
Reply with quote

#6 Thesis: Monochromatic Still Image Compression

Research On Monochromatic Still Image Compression

Author: Li Xuejun Email:li_xj@hotmail.com

Li Xuejun was born on September 20,1965 in China. He was awarded
B.S. in E.E. Beijing Fang Hua College, July 1988, M.S. in Computer
Science, March 1993, Beijing Univ. of Astro & Aero., and Ph.D. in
Computer Science, December 1996, Beijing Univ. Of Astro & Aero..His
research interests include :Inverse Problem, Wavelet Transform, Dct
Coding, Image and Signal Processing, Artificial Life, Genetic
Algorithms, Pattern Recognition, Image Compression. He is good at
signal processing and pattern recognition. Now he wishes to be a
post-doctor or a visiting scholar in the field of signal/image
processing.<li_xj@hotmail.com>

Keywords: Image Compression, Synthetic Aperture Radar, Remote Sensing
Image, Wavelet Transformation, Fractal, JPEG, Image fidelity Evaluation,
Filter

Abstract of the thesis:

The principles and schemes of monochromatic still image data
compression are studied in this dissertation, it includes four parts
as follows. First, the principles of image compression are analyzed,
the relation between the complexity and redundancy of a image is
studied, two new kinds of correlative analyses method based on
deviation and histogram are presented, using this method images are
classified into three classes: first class is those data which are no
correlation, such as Synthetic Aperture Radar(SAR)raw data; second
class is those data which have moderate correlation, such as remote
sensing images; the last class is those data which have strong
correlation, such as common figures and scenes. The shortcoming of
traditional error evaluation criteria, the Root Mean Square
Error(RMSE), is analyzed. Three new criteria to evaluate image quality
are proposed, which are Local Maximum Distance(LMD), Laplace Mean
Square Error(LMSE)and distribution of error histogram. The above
mentioned new criteria can compensate the weaknesses of RMSE. In the
second part of this dissertation, lossless compression techniques
(Huffman coding and Arithmetic coding)and lossy compression
techniques(JPEG, Fractal and Wavelet)are analyzed and evaluated. The
features and applicable area of these methods are presented. In the
third part, according to the different purpose of compression three
new kinds of image compression methods are developed. For SAR raw
data, a Multi-level Blocked Adaptive Quantification(MBAQ)compression
method based on saving each pixel information is presented; for remote
sensing images, a Multi-model Quantification Wavelet
Coding(MQWC)method is presented; for common figures, a Partial Zero
Tree Quantification(PZTQ)coding method is presented. All of three
methods are implemented on computer and the compression results are
compared with JPEG's. The results indicated that our methods could get
better fidelity than those of JPEG's, and come up to advanced world
standards in the same compression ratios. In the last part, the
international still image compression standard JPEG is deeply studied
in the dissertation, and the principle and quantification distortion
of JPEG are analyzed, and the limitation of JPEG is presented. An
improved JPEG compression method, the Low Frequency prior Coding
(LFPC) compression method is presented. The LFPC preserves all of the
advantages of JPEG's and overcomes the shortages of JPEGís. In the
new method, correlation removing is expanded to the whole image, so
that the fidelity of reconstructed images is increased by a big
margin, but the algorithm complexity is no changed. The Peak Signal
Noise Ratios(PSNR)of LFPC are 1-6dB higher than JPEGís in the same
compression ratios, these results can be competitive with wavelet
compression's.

28 Figures, 152 pages.
All times are GMT + 1 Hour
Page 1 of 1

 
Jump to: 
 


disclaimer - webmaster@wavelet.org
Powered by phpBB

This page was created in 0.027098 seconds : 18 queries executed : GZIP compression disabled