Qass
所属分类:图形图像处理
开发工具:matlab
文件大小:54KB
下载次数:36
上传日期:2007-10-04 05:04:49
上 传 者:
fx_328
说明: 源码用于计算不同的重建图像方法得到图像的质量评估检测
(Source used to calculate the different image reconstruction methods to assess the quality of the image detection)
文件列表:
GPL (15145, 2003-05-23)
image_show.m (2241, 2006-02-16)
iq_measures.m (3063, 2006-03-25)
LICENSE (811, 2006-03-25)
Pqs (0, 2006-03-25)
Pqs\COPYRIGHT (1775, 1996-08-06)
Pqs\PQS.EXE (81408, 2004-11-13)
pqs.m (2175, 2005-02-27)
****** What is PQS?
The Picture Quality Scale (PQS) is a still image quality metric for
the evaluation of images resulting from common encoding techniques.
It correlates well with the subjective mean opinion score (MOS), as
determined by human observers. For more information on its design and
behavior, see the accompanying paper "Objective Quality Scale (PQS)
for Image Coding" by M. Miyahara, K. Kotani and V. R. Algazi, which is
under review for publication in the IEEE Transactions on
Communications and can be found in the doc subdirectory. It can also
be found at http://info.cipic.ucdavis.edu/scripts/reportPage?96-12.
****** An implementation of PQS:
In the src subdirectory, you will find an implementation of PQS. This
code can also be found at
http://info.cipic.ucdavis.edu/scripts/reportPage?96-12 and in the
ftp://info.cipic.ucdavis.edu/pub/cipic/code/pqs directory. It
faithfully implements the methodology described in the paper, but must
not be used blindly. In the next few paragraphs we outline
the limitations of this implementation and caution the user as to
potential pitfalls. We hope that the software will be useful and will
stimulate interest in image quality metrics, but at the same time
don't want users to be overly optimistic as to its capabilities.
If you use this code, please refer to it as CIPIC PQS version 1. See the
accompanying COPYRIGHT file for more details.
The PQS methodology appears to have fairly wide applicability to
quality assessment in image processing, but the specific algorithm has
definite limitations. This is principally due to the use of multiple
regression to determine a single metric based on several distortion
factors. The relative importance or weights of these factors to a single
measure is dependent both on the image and coding algorithm.
1. PQS is a quality metric for picture coding, and is designed to
provide results independent of the image, and to a lesser extent of
the coding technique.
2. PQS was designed and tested on 256 x 256 images. Although some
care has been exercised in making the factors scale with image size
and resolution, its use with other than 256 x 256 images at 4 times
picture height is shaky. We plan to carefully reexamine the factor
computations and release a revision of the PQS code that will
explicitly allow for changes in image size and viewing distance. This
is not as simple as it seems since it may require calibrated MOS test
data for a new test set.
3. The regression coefficients used in PQS restricts its applicability
to data within the range of the test set. That means that if any of
the weighted distortion factor values are outside the range of the MOS
scale [0,5], the PQS evaluation has little meaning. However, a clear
statement can generally be made about the quality of the image when
weighted distortion factor values exceed their limits, i.e. the image
quality is poor. Warnings are issued when the weighted contributions
are beyond the range of values obtained during the design of PQS.
4. The MOS subjective quality scale was designed for fairly high
quality images and video. The MOS scale does provide enough
differentiation of quality at the low end of the scale and will
uniformly assign very low scores. The problem of limited range
subjective quality scales is now being addressed at CIPIC
(http://info.cipic.ucdavis.edu).
5. Some work has also been done on a color version of PQS, and the
corresponding code will be published. See
http://info.cipic.ucdavis.edu/scripts/reportPage?95-11.
6. We encourage comments and feedback on the paper and the algorithm,
and will respond if possible. We also foresee constructing a FAQ, as
necessary. Please report any problems with the code, so we can
improve the quality of the distribution.
7. This is PQS v1.0. If you so indicate by e-mail, we will keep you
posted when new versions are released.
8. Related activities can be found in our home page
(http://info.cipic.ucdavis.edu) and in the home page of HIQNET
(http://www.jaist.ac.jp/~kan/HIQNET_English.html), a international
work-group in the area of high quality imaging. A CIPIC HIQNET Web page is
also in preparation.
A conference on Very High Quality and Resolution Imaging, as
part of the SPIE Electronic Imaging Symposium in San Jose, will be
held in February 1997.
***** More information:
For more information on PQS and related activities at CIPIC, please
see our Web site: http://info.cipic.ucdavis.edu, or contact us directly:
Robert Estes, estes@cipic.ucdavis.edu
V. Ralph Algazi, vralgazi@ucdavis.edu
近期下载者:
相关文件:
收藏者: