Classifications

• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N17/00—Diagnosis, testing or measuring for television systems or their details
  • H04N17/004—Diagnosis, testing or measuring for television systems or their details for digital television systems
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N17/00—Diagnosis, testing or measuring for television systems or their details
  • H04N17/04—Diagnosis, testing or measuring for television systems or their details for receivers

Definitions

• a method of evaluating at least one quality defect in a data signal, device and associated computer program is a method of evaluating at least one quality defect in a data signal, device and associated computer program.
• the field of the invention is that of quality evaluation of audiovisual data signals comprising at least one video component and one audio component.
• the invention applies in particular to audio-visual data signals that have been degraded during transmission over telecommunication networks or during transcoding operations.
• the defects of the video component often materialize by saccades or frosts of images.
• the defects on the audio component are manifested by a loss of sound.
• the duration of these two degradations fluctuates from a few milliseconds to a few hours.
• the frequency of occurrence of this type of event may also vary over time.
• the defects may as well have a low occurrence as sustained on a time scale.
• the deployed equipment ensuring the automatic control of the audio-visual quality make measurements on the audio and video components of signals at the output of the terminal or samples at the decoder level without taking into account the reference signal, since these are rendered inaccessible. technically or legally. They use models classified as "without reference” and produce fluctuating notes on a quality scale. The conclusions of the expertise of the "no reference” prediction devices have shown that the performances must be greatly improved with respect to the human feeling.
• Document WO2006 / 103327 discloses a method for evaluating a quality degradation of an audiovisual signal based on a model for predicting the impact of the discontinuity of the fluidity of the images on the subjective quality.
• a schematic diagram of the evaluation method in question is presented. Discontinuity measurements are conducted in parallel on the video and audio components of the signal to be evaluated. Different audio and video metrics are used to qualify different types of defects (for example, image freezes, the presence of blurs or blocks in the images for the video component, sound losses for the audio component).
• a video decision module takes into account the results of the evaluations performed using the different metrics and assigns an overall quality score to the video component, using a video interaction model.
• An audio decision module does the same for the audio component and assigns an overall quality score to the audio component based on an audio interaction model.
• An audio / video weighting module takes into account both the overall video quality score and the overall audio quality score to assign an overall quality score to the audiovisual signal. It is a so-called cognitive approach because the model reproduces the mechanism of human judgment by producing notes on a quality scale.
• the prior art calculates the degradation of the perceived quality caused by each break in fluidity, and noise cutoff. The calculations are done in a disjointed way on the audio and video components and without interaction a priori or a posteriori of the results obtained. This approach fails in the face of artificially created discontinuities and when they are not considered perceptible by human judgment. In these cases, interpretation errors appear and generate false alarms.
• the invention improves the situation by means of a method for evaluating at least one quality defect of a signal carrying data intended to be returned to a recipient, said signal comprising at least one video component and a component audio.
• Such a method comprises the following steps, implemented on at least one sample of said signal:
• a fault class is assigned to it.
• the method according to the invention obtains or not the results of a search for defects on the audio component, with a view to confirming or invalidating the detected fault on the video component.
• the method according to the invention makes it possible to improve the evaluation of the quality of a degraded video signal by resorting to a fault detection on the audio component, which enables it to remove an indeterminacy on the video degradation detection.
• the method according to the invention therefore enriches the video quality evaluation model directly to the video component. help degrade detection results on the audio component.
• the invention thus proposes an entirely new and inventive approach to the evaluation of the impairments undergone by a signal, which is based on the enrichment of the video model used to evaluate the video component of this signal using the results. of the evaluation of the audio component.
• the detection step comprises, following the detection of a defect on the video component, a sub-step of rejecting the detected fault if its duration associated is less than a first predetermined perceptual detection threshold.
• the decision step comprises a sub-step of cumulation of the durations associated with the image defects of the same class detected, a substep of rejection of said defects of the same class, if their cumulative duration is less than a second predetermined perceptual detection threshold. In this way, only the image defects of a particular class that are actually perceived by the user are retained for evaluating the quality of the audiovisual signal.
• the detection step is intended to detect image-free image defects, in that it comprises a step of calculating a value of similarity between images of the sample. and a step of comparing the calculated similarity value with a predetermined threshold.
• the detection of defects on the video component relates to image-free defects.
• An image freeze is detected when several images of the sequence are sufficiently similar to each other.
• the detected image freeze class belongs to a group comprising at least:
• a macro-gel At a detected image freeze is assigned a gel class from a set of predetermined gel-like defect classes.
• the step of obtaining a search result of a defect in the audio component is implemented for an image-free defect of which the class belongs to the group comprising less:
• the step of obtaining a result of searching for a defect in the audio component is not implemented for a defect of the image freeze type whose class belongs to the group comprising at least:
• the result of the detection on the video component is considered reliable enough so that it is not necessary to resort to the search for damage on the audio component.
• the detection step is intended to detect block effect type defects, and it comprises a step of calculating a block rate value per image of the sample and a step comparing the calculated value to a predetermined threshold.
• Block effects are another type of defects that are likely to be detected on a video component.
• a blockiness type defect is detected when several images of the sequence have a sufficiently high block rate.
• the invention also relates to a device for evaluating a quality defect of a signal, able to implement the method just described.
• Such a device can be integrated with an audiovisual signal receiver or with any other equipment node of a telecommunications network, such as an audiovisual contribution or transcoding system.
• the invention therefore also relates to a receiver of a video signal transmitted in a telecommunications network, said signal comprising at least one video component and one audio component.
• a receiver 1 comprises a device for evaluating quality defects according to the invention.
• It also relates to a node equipment of a telecommunications network comprising a quality evaluation device according to the invention.
• the invention also relates to a system for distributing an audiovisual signal to at least one user terminal, said system comprising a plurality of node equipment connected to at least one telecommunications network, a node equipment being able to receive said signal and retransmit it to a neighbor node equipment.
• at least one of the group comprising the plurality of node equipment and the receiver equipment comprises the quality defect evaluation device that has just been presented.
• the invention also relates to a computer program comprising instructions for the implementation of a quality evaluation method of an audiovisual signal as described above, when this program is executed by a processor.
• a program can use any programming language. It can be downloaded from a communication network and / or saved on a computer-readable medium.
• the invention relates to a storage medium, readable by a processor, storing a computer program implementing the evaluation method of at least one quality defect that has just been described.
• FIG. 1 illustrates the evaluation scheme of a degradation of quality of an audiovisual signal according to the prior art
• FIG. 2 schematically shows an example of a distribution system of an audiovisual signal according to the invention
• FIG. 3 schematically shows the structure of a device for evaluating quality defects of an audiovisual signal according to the invention.
• FIG. 4 schematically shows the method for evaluating a quality defect of an audiovisual signal according to the invention
• FIG. 5 schematically shows the steps of the method for evaluating quality defects of an audiovisual signal according to one embodiment of the invention, according to which image-gel defects are evaluated;
• FIG. 6 schematically shows a graphical representation of an image of the video component of an audiovisual signal to be evaluated
• Figure 7 schematically illustrates an example of sound loss in an audio component of an audiovisual stream
• FIG. 8 schematically shows classes of image-free-image defects taken into account in one embodiment of the invention.
• Figure 9 shows schematically the profile of a micro-gel type defect
• Figure 10 schematically shows a variation curve of a similarity rate measure for an image gel belonging to the postproduction effects class.
• an audiovisual signal S comprising at least one video component Cv and at least one audio component Ca is considered.
• Such a signal S can be intercepted at any stage of an audiovisual transmission chain, between transmitter equipment, telecommunication network node equipment or customer receiving equipment.
• a distribution system of an audiovisual signal S comprises an audiovisual distribution server S Serv.
• a server may be for example a video on demand server or a live audiovisual content broadcasting server.
• the signal S is transmitted via a telecommunications network 1, which comprises a plurality of equipment nodes EN1, EN2.
• the nodes equipment EN1, EN2 are able to receive the signal S from a neighboring node and to retransmit it to another node of the network 1.
• the signal S is finally received by at least one destination equipment, for example the receiving equipment E , which can be a client terminal of the mobile terminal type or TV decoder for example.
• the invention can be implemented in any of these equipment Serv, EN1, EN2, ER which form the distribution chain of the audiovisual signal S.
• the device comprises a memory 110 comprising a buffer memory M, a processing unit 120, equipped for example with a processor P, and driven by a computer program Pg 130, implementing the evaluation method of quality according to the invention.
• the code instructions of the computer program 130 are for example loaded into a RAM memory before being executed by the processor of the processing unit 120.
• the processing unit 120 receives as input at least a sample of an audiovisual signal to be evaluated.
• the processor of the processing unit 120 implements the steps of the quality evaluation method described above, according to the instructions of the
• the device comprises, in addition to the memory 110, means for obtaining a detection of at least one quality defect on the video component of the sample, means for allocating a defect class audit at least one detected defect at least according to its associated duration, means for obtaining a search result of at least one defect on the audio component at the instants corresponding to the sample according to the assigned fault class and decision means to confirm the detection of a quality defect on the sequence at least according to the result of the fault search on the audio component. These means are controlled by the processor of the processing unit 120.
• such a device 100 may be integrated with at least one of the equipment Serv, EN1, EN2 and E of the distribution system of an audiovisual signal according to the invention.
• the method according to the invention applies to such a sample e, of the signal.
• a sample ei of the video component Cv includes Nbl images, with non-zero integer Nbl.
• a step E 1 at least one image defect Dv j is detected on a sample e, of the video component Cv of the signal S.
• J defects Dvl to DvJ detected in the sample ei with a nonzero integer J Obviously, if no fault is detected, it is not necessary to implement the method according to the invention.
• the aforementioned detection is obtained by analysis of the video component on the sample e ,.
• An example of an analysis method will be detailed later.
• the detected fault Dvj is then identified during a step E 2 , at least according to the j calculated duration. It is to assign an image defect class among several predetermined fault classes Cil to CIK, with K nonzero integer.
• a step E 3 of obtaining an analysis result of the audio component Ca on a sample eai corresponding to the sample evi, is triggered or not. It will be understood that for certain classes of image defect, the analysis of the video component alone is not in itself considered sufficient to allow reliable identification of the detected defect.
• the analysis of the audio component which has been, for example, carried out in parallel with that of the video component, can lead to the detection of possible audio defects Dal to DaL, with L nonzero integer, on the component S signal audio.
• An example of an audio component analysis method will be presented later.
• the class Clk considered it is either to verify that an audio defect was detected at the same instants as the current fault Dvj, or on the contrary to verify that there is no audio defect corresponding to the defect Dvj to to confirm. Depending on the result of this verification, the defect in question is therefore retained or rejected.
• the decision step E 4 comprises a sub-step E 4 , 2 of cumulation of the durations associated with image defects of the same class and of comparisons of the cumulative duration obtained at a given time. threshold. Only the defects of sufficient cumulative duration are retained and participate in the evaluation of the quality of the signal perceived by the user.
• Evaluation of image freeze defects In relation with FIG. 5, the implementation of the method according to the invention for the detection of image-free defects is now considered as an illustrative and nonlimiting example.
• the video component Cv of an audiovisual signal S is cut into samples evi, i integer ranging from 0 to I, each of duration for example between 5 and 10 ms.
• the method according to the invention is implemented for at least one evi sample.
• a step ⁇ image freeze defects are detected on an evi sample.
• An image-free defect is characterized by a break in fluidity between the image k and the neighboring image P prec of the video sequence.
• a break in fluidity is evaluated by calculating the luminance and chrominance derivative as follows:
• Yk (i, j), Uk (i, j), Vk (i, j) and Yprec (i, j), Uprec (i, j) and Vprec (i, j) respectively the components Y, U and V of Pk (i, j) and Pprec (i, j).
• Y represents the luminance of the pixel, U the component unlike the blue and V the component unlike the red pixel.
• ref refe [0, Nbl-2], Vke [1, Nbl-1], V ie [0, Vheight] and V je [0, Vwidth] if :
• ⁇ YMax, ⁇ UMax and ôVMax are integer and positive parameters allowing to adjust the level of similarity between Pk (i, j) and Pprec (i, j).
• APMax is a positive integer that represents the number of max differences accepted to consider that two consecutive images of a video are identical.
• the duration of an image freeze is calculated as follows. For a sequence of x consecutive images of the sample ev, of the video component, consider the preceding image precf such that prece [0, Nbl-x -1], images k, such that ke [prec +1, prec + x], X the number of images of the gel sequence and FPS the display frequency of the video.
• the duration of the image freeze corresponds to: t gel FPS
• the duration of the detected image gel can be compared to a first perceptual threshold ⁇ ⁇ 1 , typically of the order of 40 ms, and rejected according to the following decision criterion. :
• the duration of freezing t ge is greater than 40 ms then the gel is considered potentially perceptible. If this condition is not satisfied, the rupture of fluidity is not taken into account.
• a defect class of gel type is assigned to each defect detected, at least according to its duration, among a group of classes comprising the following defects:
• micro gel In the context of the invention, they are defined as follows: The micro gel:
• micro gel freeze time if and only if: a. There is no audio mute on the audio component (s) associated with the video component (it is detected using the audio signal) for the duration
• tgel ⁇ [imin 'Îmax] ⁇ tgel- & ⁇ ⁇ is a real in the interval [0, 1], it represents the cyclic ratio of the pseudo period of a gel below which the gel is considered as micro gel.
• ⁇ is the pseudo period of the micro-gel or the time between two starts of freezing.
• tgel represents the duration of the video freeze as defined previously,
• ⁇ represents the minimum duration of the micro gel ( ⁇ 76ms)
• a video freeze of time t fre e Z e is considered as a post-production effect of the "interpub" type and therefore as not inconvenient by the observer, if and only if: c.
• a sound loss is associated with freezing of image and the duration of the sound loss satisfies the following relation: T ⁇ X * ⁇ r
• T Mute is the duration of the mute detected, a is a percentage generally greater than or equal to 80% (it allows to take into account the audio / video shift), tjrgeze G [T MinF ⁇ T MaxFF l J ⁇ ⁇ 200ms and T MaxFF ⁇ ls;
• FIG. 10 there is shown an example of an image freeze having video similarity rates considered representative of a post-production effect.
• xrnoy is the arithmetic mean of the similarity rates of the analyzed gel images.
• An image freeze will be directly considered perceptible by itself if the freezing time exceeds the duration t min M F- In this case, we are dealing with a macro freeze and only the video component is considered to make a decision . We do not use an analysis of the audio component.
• step E ' 2 the identification of an image-free defect class among the predetermined plurality of classes is made at least from the calculated duration and possibly from functions. other measures.
• the micro-gel relies on the measurement of a pseudo period of the detected gel and the fact that the calculated duration of the gel is much lower than this pseudo period (duty cycle low).
• the conventional gel is identified by its duration which must be greater than a predetermined threshold.
• the macro gel is also identified using its duration which must be greater than a predetermined threshold.
• an audio error detection result is obtained.
• a loss-of-sound type audio defect is detected when the audio signal is attenuated with an energy level of the audio signal below a certain energy threshold over a given period of time. Each detected fault is therefore associated with a level of energy and a duration.
• an exemplary energy profile of an audio signal presenting an audio gel or sound loss is considered.
• the first step is to measure an energy loss of the audio signal over time.
• the second step is to verify that for times when a power loss has been measured, the energy of the audio signal does not exceed a certain energy threshold.
• the measurement is used on all the audio channels of a chain to be analyzed.
• the first step is to find out if there is a falling edge.
• the falling edge at time n is taken into account if and only if E " ⁇ and E n -i> ⁇ ' ⁇ with In energy of the audio signal at time n, ⁇ a threshold of energy max and ⁇ ' a threshold of energy min.
• the duration of the sound loss is then calculated as follows:
• the signal has a sound loss for a duration t PS if and only if this state begins with a falling edge of mute and that the energy level remains lower than ⁇ during the duration of the Mute state.
• s N Ech x- - (4)
• N Ech is the number of samples meeting the above conditions.
• F Ech is the sampling frequency of the audio signal.
• step E ' 4 will now be described in more detail for this particular embodiment. At the end of the steps of detecting at least one image gel on a sample ei and analyzing the audio component, this step E ' 4 ensures the production of the results on a Nbl analysis time window corresponding to the sample ei. It is first for each defect Dvj identified confirm or deny the class that has been assigned.
• the class will be confirmed if a sound loss has been identified and if it has a duration that satisfies the condition previously exposed (slightly lower);
• the class is confirmed if a sound loss has been identified and if it has a duration that satisfies the condition previously exposed (slightly lower);
• step E'2 For a macro-free fault, the class has been confirmed since step E'2 and the result of the audio analysis is not used.
• the decision module determines for each sample ei of an audiovisual stream if the defects confirmed by E'4,1 are annoying or not for the user. For each confirmed image defect, we first consider its class. If it is a post-production effect, it is rejected because it is considered as non-annoying for the user.
• M C i k is the number of confirmed defects for class Cl k .
• the decision module according to the invention therefore provides at the output a reduced number of image defects compared to the number of candidates from step E'1, corresponding to those who have passed the tests of substeps E'41 and E'42.
• the field of application of the invention is not limited to the evaluation of image-like image defects, but concerns any other type of image defect introduced by the transmission of an audiovisual stream on a In fact, an alteration of the multiplexed stream has an impact on the audio and video components at the same time instants.
• image defects of the block effect type are considered.
• the detection of block effects (“blockiness effects") can be associated with the results of an evaluation of sound losses on the audio component associated with the video component.
• the block-type effects may have different characteristics: in particular, the block effect times may vary between 20 ms to a few seconds.
• a block effect can be limited to a small area of the image or conversely relate to the entire image.
• the detection does not pose any particular difficulty and does not require the use of the method according to the invention.
• the method according to the invention can be applied to defects of the block effect type.
• a previous detection of gel-type defects over a common analysis period can be validly exploited by the method according to the invention to confirm or invalidate the presence of troublesome block-effect type defects in an audiovisual stream.
• the invention that has just been presented can be applied to any audiovisual signal at any stage of its distribution chain. It makes it possible to more effectively evaluate the quality defects caused by its transmission or transcoding step, according to perceptual criteria.

Landscapes

• Engineering & Computer Science (AREA)
• Health & Medical Sciences (AREA)
• Biomedical Technology (AREA)
• General Health & Medical Sciences (AREA)
• Multimedia (AREA)
• Signal Processing (AREA)
• Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)

Applications Claiming Priority (2)

Publications (1)

Family

ID=47263448

Family Applications (1)

Country Status (4)

Family Cites Families (28)

• 2011
  • 2011-11-07 FR FR1160118A patent/FR2982449A1/fr not_active Withdrawn
• 2012
  • 2012-10-30 US US14/356,787 patent/US9143775B2/en not_active Expired - Fee Related
  • 2012-10-30 EP EP12794413.0A patent/EP2777292A1/de not_active Withdrawn
  • 2012-10-30 WO PCT/FR2012/052512 patent/WO2013068676A1/fr active Application Filing

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events