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：

评论：[我要评论] [举报此文件]

收藏者：