Field of the Invention The invention relates to a method and apparatus for assessing the quality of a video signal during encoding or compressing of the video signal. The video signal may be a non- interactive streaming video or an interactive streaming video.

Background of the Invention

Among the numerous TV distribution services, IPTV (Internet protocol TV) is becoming increasingly important and is more and more replacing analogue or non packet based transmission methods. It is a major responsibility of the broadcast provider towards both content provider and customer to maintain the quality of its service. In large IPTV networks only fully automated quality monitoring probes can fulfil this requirement.

In order to achieve a high degree of satisfaction of the user of video services such as non- interactive streaming video (IPTV, VoD) or static video (DVD), the perceived video quality of those services need to be estimated.

To this aim, video quality models are developed which provide estimates of the video quality as perceived by the user. Those models can for instance output the degree of similarity between the video received at the user side and the original non-degraded video. In addition, and more sophistically, the Human Visual System (HVS) can be modelled. At last, the model can map the results of extensive subjective quality tests.

Video quality models and thus measurement systems are generally classified as follow:

Quality model types

· Full Reference (FR): a reference signal is required. • Reduced-Reference (RR): partial information extracted from the source signal is required.

• No-Reference (NR): no reference signal is required. Input parameters types

• signal/media-based: the decoded image (pixel-information) is required.

• parameter-based: bitstream-level information is required. Information can range from packet-header information, requiring parsing of the packet-headers but not (full- or partial) decoding of the bitstream to the complete decoding of the bitstream.

Type of application

• Network Planning: the model or measurement system is used before the implementation of the planning in order to plan the best possible implementation.

• Service Monitoring: the model is used during service operation.

Related information of the types of video quality models can be found in references [1], [2], or [3].

In the context of MPEG-based video services, one of the parameters influencing the video perceived quality is the GOP-Structure (GOP = Group of Pictures), including the GOP-length, i.e., the distance between frames which do not require previous or further frames to be decoded, the so-called 'key-frames' or "I-frames". One Group of Picture covers one I-frame and all frames till the next I-frame of the video sequence. The GOP-structure - and thus GOP-length - is generally chosen as a trade-off between encoding efficiency and error-propagation (see, e.g., references [4], [5], [6]). In these references, the authors provide guidelines for selecting the most appropriate GOP structure for MPEG. Some models take as input parameters GOP-related parameters but only under packet loss conditions, as in references [2], [7], [8], [9], or [10]. However, they consider only fixed GOP lengths and examine the impact on quality based on the temporal distance of the frame where the packet loss occurs to the next key frame. The quality impact of the GOP-length on encoding is not taken into account. Quality estimation methods commonly support a distinguished estimation of the quality related to the coding (compression, Qcod) of the video signal and the quality due to packet loss during transmission (Qtrans). Quality estimation methods commonly use one of two approaches to combine an estimation concerning the quality of the compression and the transmission quality. Equation (1) and (2) illustrate the two different approaches

Q = Q0 - Qcod - Qtrans , Q0, Qx 0 ... 100 (1) Q = Q0 * Qcod * Qtrans , Q0, Qx 0 ... 1 (2), in which Q0 represents the base quality or a function of the base quality.

Summary of the Invention

According to a first aspect, the invention provides a method for assessing the quality of a video signal during encoding or compressing of the video signal, the method comprising the steps of:

a) estimating the quality of the encoded or compressed video signal using one or more parameters; and

b) using the length of the Group of Pictures, GOP, and/or GOP -length-related parameters as additional parameter(s) to adjust the estimated video signal quality.

Thus, the method of the invention focuses on the quality estimation of the term characterizing the compression efficiency Qcod. The method of the invention may be combined with different methods for quality estimation of packet loss in video streams.

The method of the invention is a parameter-based video quality model with light-weight parameters, and thus provides a video quality model suitable both for network planning and service monitoring. In case of network planning, values of the parameters are assumed by the network planner, based on knowledge of previously developed similar networks. In case of service monitoring, the model takes as input parameters extracted from the bitstream. In principle, the measurement system in which the method of the invention is embedded can be placed at different locations in the network. However, the closer the probe to the user device is, the more representative of the actual quality at the user side the predicted quality is, when considering packet loss. In case of service monitoring, the parameters do not require access to the payload, and therefore do not require either a partial- or full-decoding of the bistream. A light-parsing of the packet headers is sufficient for accessing the parameters to be sent to the model, i.e., method. Note that if deeper parsing is allowed, as with un-encrypted data, the parameters can also be used with additional parameters extracted from the bit-stream by partial or full-decoding.

The method of the invention considers the parameters of the encoding process in detail. Known parametric quality estimation methods use the bit-rate, the frame-rate, the video resolution, the codec type and the content type to estimate the quality of a compressed video stream. The invention goes beyond these conventional methods by using the GOP-length or GOP-length-related parameters as an additional parameter for adjusting the estimated video quality. This parameter has a direct impact on the number of bits per frame, and thus on the perceived video quality. As a consequence, the invention considers this parameter as input of the model in addition to the above parameters. With the method of the invention a much more accurate estimation of the related perceived quality can be achieved.

Again, the method of the invention takes as input parameters such as the video resolution, the codec type, the content type, the bit-rate, the frame-rate and the key-frame-rate, and output an estimated video quality (Qcod) based on those parameters. This can be written as in equation (3):

Qcod =/ (br,fr, cod, cont, res, kfr, I, G) (3) in which

br bitrate, the number of bits per second

r frame-rate, the number of video frames per second

cod employed codec (for instance H.264 main profile)

cont content type (for instance TVnews, sport, movie, etc.)

res video resolution (for instance High Definition, 1920x1080 pixels progressive) kfr ... key-frame-rate, the number of key-frames per second

/ number of bits in an I-frame

G number of bits in a GOP Note that for some of the parameters their values might not be used directly into the function, but as a switch for selecting the appropriate coefficients values of the function. For instance, a possible model for Qcod in equation (1) for IPTV in which the frame-rate is considered constant is shown in equation (4):

Qcod = a * Qxp(b * br) + c (4) In equation (4), the values of the a, b and c coefficients depend on the employed video codec, on the video resolution and on the content type. Those coefficients are preferably obtained in a least-square-error curve fitting procedure using the ratings of perception tests as target values. Following the invention, and still using ratings of perception tests, the key-frame-rate is used as additional parameter into equation (4), yielding equation (5):

Qcod = (al * kfr + a2) * exp(b * br) + c (5) The coefficients al, a2 and b may be obtained in a least-square-error curve fitting procedure using the ratings of perception tests as target values, al and a2 depend on the given encore settings and content type.

The key-frame-rate may be obtained by computing the ratio of the frame-rate over the GOP length (see Fig.2).

In an alternative embodiment of the invention, the coefficients al and a2 can be explicitly dependent on additional information extracted from the bit-stream or packet headers. For example, al= I / G * al ' and a2= I / G * a2 where al ' and a2 'are curve-fitting parameters that represent given encoder settings, but excluding a part of the variability which can directly be measured using I / G. Thus, the impact of the GOP-length and GOP-length-related parameters is modulated by the key-frame size in bytes normalized by the number of bytes in the GOP. Following this alternative embodiment, we obtain equation (6): Qcod =— * (aYkfr * +a2') * exp(b * br)+ c (6)

G

Indeed the influence of the key-frame rate also depends on how the information bits are spread over the GOP. If the information bits are equally spread over the frames of the GOP, and in the considered case of no packet-loss, the key-frame-rate has no influence on the quality. If all information bits are in the I-frame of the GOP, the influence of the key-frame- rate is maximal. This impact of the bit distribution over the GOP is captured by the ratio I G, where G is the number of bits in a GOP and / is the number of bits in an I-frame. As explained above, it is preferred according to the invention that the one or more parameter used in step a) of the method is selected from the set comprising: bit rate, frame rate, video resolution, codec type, content type.

The values of the parameters used in step a) may be computed from the packet header information extracted from the bit stream of the video signal and/or derived from side information.

Other aspects, features, and advantages will be apparent from the summary above, as well as from the description that follows, including the figures and the claims.

Brief Description of the Drawings

Fig. 1 shows a block diagram of the general framework of the estimation of quality related to compression; and

Fig. 2 shows a detailed view of the computation of the key- frame-rate. Description of Detailed Embodiments of the Invention

Fig. 1 shows the general framework for estimating the perceived video quality in case of compression degradation only (Qcod) either in case of network planning or in case of service monitoring.

In the case of network planning (block 100a), values of the parameters (block 200) to be sent to the video quality model (block 300) are estimated by the network planner. In the case of service monitoring (block 100b), the parameters to be sent to the video quality model are computed from the packet header information extracted from the bit-stream.

Fig. 2 shows a detailed view of the computation of the key-frame-rate. As shown in Fig. 2, one GOP covers one I-frame and all frames till the next I-frame. If n is the number of frames on a time window of t or more seconds, then fr = n/t is the frame rate of the video sequence. If d is the number of frames between two I-frames, then kfr =fr/d is the key- frame-rate of the video sequence. While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive. It will be understood that changes and modifications may be made by those of ordinary skill within the scope of the following claims. In particular, the present invention covers further embodiments with any combination of features from different embodiments described above and below.

Furthermore, in the claims the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single unit may fulfil the functions of several features recited in the claims. The terms "essentially", "about", "approximately" and the like in connection with an attribute or a value particularly also define exactly the attribute or exactly the value, respectively. Any reference signs in the claims should not be construed as limiting the scope.

[1] A. Takahashi, D. Hands, and V. Barriac, "Standardization Activities in the ITU for a QoE Assessment of IPTV," in IEEE Communication Magazine, 2008.

[2] S. Winkler and P. Mohandas, "The Evolution of Video Quality Measurement: From PSNR to Hybrid Metrics," in IEEE Trans. Broadcasting, 2008.

[3] A. Raake, M.N. Garcia, S. Moeller, J. Berger, F. Kling, P. List, J. Johann, and C. Heidemann, "T-V-MODEL: Parameter-based prediction of IPTV quality," in Proc. of ICASSP, 2008.

[4] Huahui Wu, Mark Claypool, and Robert Kinicki, GUIDELINES FOR SELECTING PRACTICAL MPEG GROUP OF PICTURES", In Proceedings of LASTED International Conference on Internet and Multimedia Systems and Applications (EuroIMSA), Innsbruck, Austria, February 2006 http://web.cs.wpi.edu/~clavpool/papers/practical-gop/vGOP.pd f

[5] Arpad Huszak and Sandor Imre, "ANALYSIS GOP STRUCTURE AND PACKET LOSS EFFECTS ON ERROR PROPAGATION IN MPEG-4 VIDEO STREAMS", in Proceedings of the 4 ^th International Symposium on Communications, Control and Signal Processing 2010 (ISCCP 2010), Limassol, Cyprus, 3-5 March 2010

[6] Tektronix, "An Analysis of MPEG Encoding Techniques on Picture Quality, June 1998.

[7] A. R. Reibman and D. Poole. Characterizing packet loss impairments

in compressed video. IEEE ICIP, Sept 2007.

[8] A. R. Reibman et al. Predicting packet-loss visibility using scene characteristics. Packet Video, pages 308-317, Sept 2007.

[9] JP002006033722AA, NTT, "Image quality control method and image quality control system"

[10]

http://ww.telchemv.com/appnotes/aJnderstanding%20IP%20Vid eo%20Qualitv%20Metri Edf