Knowledge base: Warsaw University of Technology

Settings and your account

Back

Metoda detekcji i redukcji usterek związanek z kompresją wprocesie wytwarzania płyt DVD-Video

Rafał Kłoda

Abstract

The initial part of the thesis comprises the analysis of the DVD-Video manufacturing process, classification of image distortion caused by video lossy compression and the state of knowledge in the field of objective and subjective quality evaluation of digital motion pictures. The capability of quality analyzers was examined: the characteristics of the video sequences were determined for the wide range of coding parameters. Their usefulness as a tool for monitoring the level of defects in compressed video was evaluated. Then a concept of compression artifacts detection and reduction was presented. The general idea of the method is to divide the movie into segments of similar duration, which are described with a set of numerical measures characterizing it. They state an input to the forecasting algorithm, which shows segments of the video, where the quality is poorer than required. This list, along with a set of numerical measures characterizing the individual segments, is displayed. Then the system operator, after seeing fragments of dubious quality, can decide which of them require a re-encode with a larger bitrate. The library of video test sequences and the measurement station was prepared. The measurement station enables for MOS (Mean Opinion Score) determination along with a set of numeric measures for video before and after compression. The measurement station provides playback of source material and the compressed video in a way consistent with the quality evaluation method. Obtaining scores from viewers was realized with paper forms, which were digitized and then processed. Developed software automatically determines over a dozen measures used for video quality evaluation. Created library enables for preparing test material that covers the typical range of compression parameters used for DVD-Video. For each video sequence a set of parameters was computed. Then the experiment with DCR (Degradation Category Rating) method was conducted and MOS was calculated. The effect of the added stimuli (reference video) on the quality distinction and the range of scale used were examined as well as the influence of a number of viewers. A set of indicators was computed for numerous video sequences of a different character and a way of coding. The objective parameters were used for examining their correlation with MOS acquired with DCR method. The efficient algorithm to search for the optimal model was developed. It creates k-element combinations of variables from the n-element set for the multiple regression equation and automatic analysis of established relationships with assumptive criteria. To obtain a good statistical predictive model of video quality, the issue of a number of variables that should be incorporated into the regression equation and which of them should be used. The significance of the model and of individual explanatory variables was examined, the accuracy of the model fit to the experimental data was assessed, the ratings of the model predictions for a different set of sequences was checked and finally a thorough analysis of the assumptions required for the multiple regression was conducted. The concept of using the proposed method of impairments detection for video control was presented. Integrating the diagnostics into the production process is carried out by a dedicated station. The basic feature of the proposed solution is that the process of quality control is disconnected from the encoding process and enables for optimal use of the coding system capabilities. Automation of the selected segments re-encoding process was realized. Experimental research was conducted, which confirmed the effectiveness of proposed impaired segments detection method.
Record ID
WUT82934eaa8c224620a3d8c8a4c3192b13
Diploma type
Doctor of Philosophy
Author
Title in Polish
Metoda detekcji i redukcji usterek związanek z kompresją wprocesie wytwarzania płyt DVD-Video
Language
(pl) Polish
Certifying Unit
Faculty of Mechatronics (FM)
Discipline
automation and robotics / (technology domain) / (technological sciences)
Status
Finished
Defense Date
10-12-2014
Title date
17-12-2014
Supervisor
Sabina Żebrowska-Łucyk Sabina Żebrowska-Łucyk,, The Institute of Metrology and Biomedical Engineering (FM/IMBE)Faculty of Mechatronics (FM)
Internal reviewers
External reviewers
Marian Wysocki, prof., Polit. Rzeszowska Marian Wysocki, prof., Polit. Rzeszowska,, Undefined Affiliation
Pages
111
Keywords in English
classification of image distortion caused by video lossy compression
Abstract in English
The initial part of the thesis comprises the analysis of the DVD-Video manufacturing process, classification of image distortion caused by video lossy compression and the state of knowledge in the field of objective and subjective quality evaluation of digital motion pictures. The capability of quality analyzers was examined: the characteristics of the video sequences were determined for the wide range of coding parameters. Their usefulness as a tool for monitoring the level of defects in compressed video was evaluated. Then a concept of compression artifacts detection and reduction was presented. The general idea of the method is to divide the movie into segments of similar duration, which are described with a set of numerical measures characterizing it. They state an input to the forecasting algorithm, which shows segments of the video, where the quality is poorer than required. This list, along with a set of numerical measures characterizing the individual segments, is displayed. Then the system operator, after seeing fragments of dubious quality, can decide which of them require a re-encode with a larger bitrate. The library of video test sequences and the measurement station was prepared. The measurement station enables for MOS (Mean Opinion Score) determination along with a set of numeric measures for video before and after compression. The measurement station provides playback of source material and the compressed video in a way consistent with the quality evaluation method. Obtaining scores from viewers was realized with paper forms, which were digitized and then processed. Developed software automatically determines over a dozen measures used for video quality evaluation. Created library enables for preparing test material that covers the typical range of compression parameters used for DVD-Video. For each video sequence a set of parameters was computed. Then the experiment with DCR (Degradation Category Rating) method was conducted and MOS was calculated. The effect of the added stimuli (reference video) on the quality distinction and the range of scale used were examined as well as the influence of a number of viewers. A set of indicators was computed for numerous video sequences of a different character and a way of coding. The objective parameters were used for examining their correlation with MOS acquired with DCR method. The efficient algorithm to search for the optimal model was developed. It creates k-element combinations of variables from the n-element set for the multiple regression equation and automatic analysis of established relationships with assumptive criteria. To obtain a good statistical predictive model of video quality, the issue of a number of variables that should be incorporated into the regression equation and which of them should be used. The significance of the model and of individual explanatory variables was examined, the accuracy of the model fit to the experimental data was assessed, the ratings of the model predictions for a different set of sequences was checked and finally a thorough analysis of the assumptions required for the multiple regression was conducted. The concept of using the proposed method of impairments detection for video control was presented. Integrating the diagnostics into the production process is carried out by a dedicated station. The basic feature of the proposed solution is that the process of quality control is disconnected from the encoding process and enables for optimal use of the coding system capabilities. Automation of the selected segments re-encoding process was realized. Experimental research was conducted, which confirmed the effectiveness of proposed impaired segments detection method.
Thesis file
  • File: 1
    Rozprawa_doktorska_Rafal_Kloda.pdf
Request a WCAG compliant version
Citation count
2

Uniform Resource Identifier
https://repo.pw.edu.pl/info/phd/WUT82934eaa8c224620a3d8c8a4c3192b13/
URN
urn:pw-repo:WUT82934eaa8c224620a3d8c8a4c3192b13

Confirmation
Are you sure?
Report incorrect data on this page