ITU-R BT 1788-2007 Methodology for the subjective assessment of video quality in multimedia applications《多媒体应用中视频质量主观评定的方法论》.pdf

《ITU-R BT 1788-2007 Methodology for the subjective assessment of video quality in multimedia applications《多媒体应用中视频质量主观评定的方法论》.pdf》由会员分享，可在线阅读，更多相关《ITU-R BT 1788-2007 Methodology for the subjective assessment of video quality in multimedia applications《多媒体应用中视频质量主观评定的方法论》.pdf（13页珍藏版）》请在麦多课文档分享上搜索。

1、 Rec. ITU-R BT.1788 1 RECOMMENDATION ITU-R BT.1788 Methodology for the subjective assessment of video quality in multimedia applications (Question ITU-R 102/6) (2007) Scope Digital broadcasting systems will permit the delivery of multimedia and data broadcasting applications comprising video, audio,

2、 still-picture, text and graphics. This Recommendation specifies non-interactive subjective assessment methods for evaluating the video quality of multimedia applications. The ITU Radiocommunication Assembly, considering a) that digital broadcasting systems are being introduced in many countries; b)

3、 that multimedia and data broadcasting services which comprise video, audio, still-picture, text, graphics, etc., have been introduced or are planned to be introduced using digital broadcasting systems; c) that multimedia services will involve a broadcasting infrastructure characterized by the possi

4、ble use of fixed and mobile receivers, fixed and variable frame rates, different image formats, advanced video codecs, packet loss, etc.; d) that it will be necessary to specify performance requirements and to verify the suitability of technical solutions considered for each service with the perform

5、ance requirements of that service; e) that such verification will principally involve subjective assessment of video quality under controlled conditions; f) that the subjective assessment methodologies specified in Recommendation ITU-R BT.500 may be used for multimedia applications; g) that subjecti

6、ve assessment methodologies other than those specified in Recommendation ITU-R BT.500 may also be used; h) that the adoption of standardized methods is of importance in the exchange of information between various laboratories, recommends 1 that the general methods of test, i.e. the grading scales an

7、d the viewing conditions for the assessment of picture quality described in Annex 1, should be used for laboratory experiments and whenever possible for operational assessments in multimedia applications; 2 that the full descriptions of test configurations, test materials, observers and methods shou

8、ld be provided in all test reports; 3 that, in order to facilitate the exchange of information between different laboratories, the collected data should be processed in accordance with the statistical techniques detailed in Annex 2. 2 Rec. ITU-R BT.1788 NOTE 1 The development of a library of video m

9、aterial appropriate for the subjective assessment of video quality in multimedia applications needs to be further pursued by Radiocommunication Study Group 6. Annex 1 Description of assessment methods 1 Introduction Many countries have begun deploying digital broadcasting systems that will permit th

10、e delivery of multimedia and data broadcasting applications comprising video, audio, still-picture, text and graphics. Standardized subjective assessment methods are needed to specify performance requirements and to verify the suitability of technical solutions considered for each application. Subje

11、ctive methodologies are necessary because they provide measurements that allow industry to more directly anticipate the reactions of end users. The broadcasting system needed to deliver multimedia applications is markedly different from the one currently in use: information is accessed through fixed

12、 and/or mobile receivers; the frame rate can be fixed or variable; the possible image size has a large range (i.e. SQCIF to HDTV); the video is typically associated with embedded audio, text and/or sound; the video may be processed with advanced video codecs; and the preferred viewing distance is hi

13、ghly dependent on the application. The subjective assessment methods specified in Recommendation ITU-R BT.500 should be applied in this new context. In addition, investigations of multimedia systems might be carried out with new methodologies to meet the user requirements of the characteristics of t

14、he multimedia domain. This Recommendation describes non-interactive subjective assessment methods for evaluating the video quality of multimedia applications. These methods can be applied for different purposes including, but not limited to: selection of algorithms, ranking of audiovisual system per

15、formance and evaluation of the video quality level during an audiovisual connection. Terms and definitions relating to this Recommendation can be found in Appendix 3 to Annex 1. 2 Common features 2.1 Viewing conditions Recommended viewing conditions are listed in Table 1. The size and the type of di

16、splay used should be appropriate for the application under investigation. Since several display technologies are to be used in multimedia applications, all relevant information concerning the display (e.g. manufacturer, model and specifications), used in the assessment should be reported. When PC-ba

17、sed systems are used to present the sequences, the characteristics of the systems (e.g. video display card) should also be reported. Table 2 shows an example of the data record for the configuration of multimedia system under test. If the test images are obtained using a specific decoder-player comb

18、ination, the images must be separated from the proprietary skin to get an anonymous display. This is necessary to ensure that the quality assessment is not influenced by the knowledge of the originating environment. Rec. ITU-R BT.1788 3 When the systems assessed in a test use reduced picture format,

19、 such as CIF, SIF or QCIF, etc., the sequences should be displayed on a window of the display screen. The colour of the background on the screen should be 50% grey. TABLE 1 Recommended viewing conditions as used in multimedia quality assessment Parameter Setting Viewing distance(1)Constrained: 1-8 H

20、 Unconstrained: based on viewers preference Peak luminance of the screen 70-250 cd/m2Ratio of luminance of inactive screen to peak luminance 0.05 Ratio of the luminance of the screen, when displaying only black level in a completely dark room, to that corresponding to peak white 0.1 Ratio of luminan

21、ce of background behind picture monitor to peak luminance of picture(2) 0.2 Chromaticity of background(3)D65Background room illumination(2) 20 lux (1)Viewing distance in general depends on the application. (2)This value indicates a setting allowing maximum detectability of distortions, for some appl

22、ications higher values are allowed or they are determined by the application. (3)For PC monitors, the chromaticity of background should approximate as much as possible the chromaticity of “white point” of the display. TABLE 2 Configuration of the multimedia system under test Parameter Specification

23、Type of display Display size Video display card Manufacturer Model Image information 2.2 Source signals The source signal provides the reference picture directly and the input for the system under test. The quality of the source sequences should be as high as possible. As a guideline, the video sign

24、al should be recorded in multimedia files using YUV (4:2:2, 4:4:4 formats) or RGB (24 or 32 bits). When the experimenter is interested in comparing results from different laboratories, it is necessary to use a common set of source sequences to eliminate a further source of variation. 4 Rec. ITU-R BT

25、.1788 2.3 Selection of test materials The number and type of test scenes are critical for the interpretation of the results of the subjective assessment. Some processes may give rise to a similar magnitude of impairment for most sequences. In such cases, results obtained with a small number of seque

26、nces (e.g. two) should provide a meaningful evaluation. However, new systems frequently have an impact that depends heavily on the scene or sequence content. In such cases, the number and type of test scenes should be selected so as to provide a reasonable generalization to normal programming. Furth

27、ermore, the material should be chosen to be “critical but not unduly so” for the system under test. The phrase “not unduly so” implies that the scene could still conceivably form part of normal television programming content. A useful indication of the complexity of a scene might be provided by its

28、spatial and temporal perceptual characteristics. Measurements of spatial and temporal perceptual characteristics are presented in more detail in Appendix 1 to Annex 1. 2.4 Range of conditions and anchoring Because most of the assessment methods are sensitive to variations in the range and distributi

29、on of conditions seen, judgment sessions should include the full ranges of the factors varied. However, this may be approximated with a more restricted range, by also presenting some conditions that would fall at the extremes of the scales. These may be represented as examples and identified as most

30、 extreme (direct anchoring) or distributed throughout the session and not identified as most extreme (indirect anchoring). If possible, a large quality range should be used. 2.5 Observers The number of observers after screening should be a least 15. They should be non-expert, in the sense that they

31、are not directly concerned with picture quality as part of their normal work and are not experienced assessors. Prior to a session, the observers should be screened for (corrected to) normal visual acuity on the Snellen or Landolt chart and for normal colour vision using specially selected charts (e

32、.g. Ishihara). The number of assessors needed depends upon the sensitivity and reliability of the test procedure adopted and upon the anticipated size of the effect sought. Experimenters should include as many details as possible on the characteristics of their assessment panels to facilitate furthe

33、r investigation of this factor. Suggested data to be provided could include: occupation category (e.g. broadcast organization employee, university student, office worker), gender and age range. 2.6 Instructions for the assessment Assessors should be carefully introduced to the method of assessment,

34、the types of impairment or quality factors likely to occur, the grading scale, timing, etc. Training sequences demonstrating the range and the type of the impairments to be assessed should be used with scenes other than those used in the test, but of comparable sensitivity. 2.7 Experimental design I

35、t is left to the experimenter to select the experimental design in order to meet specific cost and accuracy objectives. It is preferable to include at least two replications (i.e. repetitions of identical conditions) in the experiment. Replications make it possible to calculate individual reliabilit

36、y and, if necessary, to discard unreliable results from some subjects. In addition, replications ensure that learning effects within a test are to some extent balanced out. A further improvement in the handling of learning effects is obtained by including a few “dummy presentations” at the beginning

37、 of each test session. These conditions should be representative of the presentations to be shown later Rec. ITU-R BT.1788 5 during the session. The preliminary presentations are not to be taken into account in the statistical analysis of the test results. A session, that is a series of presentation

38、s, should not last more than half an hour. When multiple scenes or algorithms are tested, the order of presentation of the scenes or algorithms should be randomized. The random order might be amended to ensure that the same scenes or same algorithms are not presented in close temporal proximity (i.e

39、. consecutively). 3 Assessment methods The video performance of multimedia systems can be examined using Recommendation ITU-R BT.500 methodologies. A list of selected methods is provided in 3.1. In 3.2 is a description of an additional methodology, called SAMVIQ, that takes advantage of the characte

40、ristics of multimedia domain and can be used for the assessment of the performance of multimedia systems. 3.1 Recommendation ITU-R BT.500 methodologies The following Recommendation ITU-R BT.500 methodologies should be used for the assessment of video quality in multimedia systems: Double stimulus im

41、pairment scale (DSIS) method as described in Recommendation ITU-R BT.500, Annex 1, 4. Double stimulus continuous quality scale (DSCQS) method as described in Recommendation ITU-R BT.500, Annex 1, 5. Single-stimulus (SS) methods as described in Recommendation ITU-R BT.500, Annex 1, 6.1. Stimulus-comp

42、arison (SC) methods as described in Recommendation ITU-R BT.500, Annex 1, 6.2. Single stimulus continuous quality evaluation (SSCQE) method as described in Recommendation ITU-R BT.500, Annex 1, 6.3. 3.2 Subjective Assessment of Multimedia VIdeo Quality (SAMVIQ) In this method, the viewer is given ac

43、cess to several versions of a sequence. When all versions have been rated by the viewer, the following sequence content can be then accessed. The different versions are selectable randomly by the viewer through a computer graphic interface. The viewer can stop, review and modify the score of each ve

44、rsion of a sequence as desired. This method includes an explicit reference (i.e. unprocessed) sequence as well as several versions of the same sequence that include both processed and unprocessed (i.e. a hidden reference) sequences. Each version of a sequence is displayed singly and rated using a co

45、ntinuous quality scale similar to the one used in the DSCQS method. Thus, the method is functionally very much akin to a single stimulus method with random access, but an observer can view the explicit reference whenever observer wants, making this method similar to one that uses a reference. The SA

46、MVIQ quality evaluation method uses a continuous quality scale to provide a measurement of the intrinsic quality of video sequences. Each observer moves a slider on a continuous scale graded from 0 to 100 annotated by 5 quality items linearly arranged (excellent, good, fair, poor, bad). Quality eval

47、uation is carried out scene by scene (see Fig. 1) including an explicit reference, a hidden reference and various algorithms. 6 Rec. ITU-R BT.1788 To get a better understanding of the method, the following specific words are defined below: Scene: audio-visual content Sequence: scene with combined pr

48、ocessing or without processing Algorithm: one or several image processing techniques. 3.2.1 Explicit, hidden reference and algorithms An evaluation method commonly includes quality anchors to stabilize the results. Two high quality anchors are considered in the SAMVIQ method for the following reason

49、s. Several tests have been carried out that indicate minimized standard deviations of scores by using an explicit reference rather than a hidden or no reference. Particularly to evaluate codec performance, it is better to use an explicit reference to get the maximum reliability of results. A hidden reference is also added to evaluate intrinsic quality of the reference, instead of the explicit reference, because the presentation is anonymous as well as processed sequences. The explicit name “reference” has an influence on about 30% of observers. These observer