WO2007022803A1

WO2007022803A1 - Method for automatically conducting objective technical quality assessment of the content of audio and/or video material and for filtering out quality-critical sequences from this material

Info

Publication number: WO2007022803A1
Application number: PCT/EP2006/003140
Authority: WO
Inventors: Olaf Lamprecht; Torsten Schulz
Original assignee: Deutsche Telekom Ag
Priority date: 2005-08-25
Filing date: 2006-04-06
Publication date: 2007-03-01
Also published as: DE102005040301A1

Abstract

The invention relates to a method for automatically conducting objective technical quality assessment of the content of audio and/or video material and for filtering out quality-critical sequences, comprising the following steps: A) recording a series of measurements of a number of measured values of a quality feature of the audio and/or video content and recording at least one additional series of measurements of measured values of at least one other quality feature of the audio and/or video content, each measured value having a time relationship to the examined audio and/or video content; b) applying previously established conditions, which are specific to each individual quality feature series of measurements, to the measurement values obtained in step a) in order to filter out such time segments of the analyzed material, which conform to said conditions, one condition representing a comparison of a measured value with an upper limit value established for its measured quantity and/or with a correspondingly established lower limit value; c) conducting a logical association of the filter results of all quality feature, and; d) indexing the quality-critical audio and/or video sequences according to the result of the logical association.

Description

Method for the automatic objective technical quality assessment of the content of audio and / or video material and for filtering quality-critical sequences from this material

The invention relates generally to methods for assessing the technical quality of audio / video content. In particular, the invention relates to a method for the automated objective quality measurement of audio / video content with the aim of an automated and objectified filtering out such sections of the audio / video content, which must be fed to a subjective re-evaluation. That is, the invention in its core relates to a method for the automatic objective technical quality assessment of the content of audio and / or video material and for filtering quality critical sequences from this material. Moreover, the invention relates to a device for carrying out this method.

In the prior art, the quality of audio / video content is usually checked by subjective, manual full control. This is very time consuming and error prone. Objectively and automatically, the quality of multimedia content of the prior art can not be measured with satisfactory accuracy, since the prior art methods and apparatus are incapable of producing quality defects of desired styling means such as e.g. False colors, intentional jerkiness and the like. To distinguish.

In the prior art, there are various, partly standardized quality metrics that make statements about properties of audio / video content. These metrics are determined and issued by commercially available software and hardware solutions. The metrics known from the prior art can be used to significantly reduce the effort of subjective quality measurement. In many cases, averaged values are output as a result of the measurement over the entire playing time of the content. However, this information is not sufficient for an assessment of the technical quality of the content, since short excerpts with greatly reduced quality are not reflected in these values, but are quite an exclusion criterion for the use of the content can. Therefore, in the prior art, a subsequent evaluation of these sections by human, ie subjective, evaluation is essential.

The technical quality of audio / video content perceived or perceived by humans can be determined by subjective and objective measurement methods.

A well-known subjective rating method is Double Stimulus Continuous Quality Scale (DSCQS). This is a procedure recommended by the International Telecommunication Union (ITU) (see Recommendation ITU-R BT.500-9: "Methodology for the Subjective Assessment of the Quality of Television Pictures", 1998 and ITU-T Recommendation P.910: Subjective video quality assessment methods for multimedia applications, 1999). The advantage of this method is a low context-dependent influence on the result. A disadvantage of this method is the high cost of its implementation and the amount of associated costs.

In the field of objective evaluation, there is currently no standard obliging a procedure to assess the quality of audio / video content. However, several techniques have been developed in the prior art which attempt to map the human visual system (HVS) as perfectly as possible. The HVS is a model that describes visual perception by humans. The latter include Webster's approach (see AA Webster et al .: An Objective Video Quality Assessment System Based on Human Perception). In: Human Vision, Visual Processing, and Digital Display IV, San Jose. CA), 1993, pages 15 to 26.), Moving Picture Quality Metrics (MPQM) by Van den Branden Lambrecht (see CJ van den Branden Lambrecht: "A working spatio-temporal model of the human visual system for image restoration and Quality assessment applications ".- In: Proc. IEEE Int. Conf. Acoust, Speech, and Signal Processing, 1996, pages 2291 to 2294.) and the group of so-called JND (Just Noticable Difference) based methods.

Other prior art methods for the objective quality assessment of audio / video content include the National Telecommunications and Information Administration's (NTIA) Fully Automated Video Quality Model (VQM) and Tektronix's Objective Perceptual Video Quality Measurement Technique based on JND. However, all the above-mentioned methods for objectively assessing the technical quality of audio / video content have in common that they do not ensure high reliability in detecting technical defects because they are unable to distinguish between actual technical defects and artistically used stylistic devices the audio / video production.

The invention is therefore based on the object of providing a method for the automatic objective technical quality assessment of the content of audio and / or video material and for filtering out quality-critical sequences which is better able than the methods known from the prior art. between actual technical defects and artistic stylistic devices used in audio / video production. Furthermore, the invention has the object to provide a device for carrying out the method according to the invention.

According to the invention this object is achieved by a method according to claim 1 and by a device according to claim 8.

The significantly improved performance of the method according to the invention over the methods known from the prior art results essentially from the fact that not only a single quality feature is subjected to a time series analysis, but that time series analyzes are carried out for several quality features and finally even a logical combination of these separate time series takes place.

Advantageous and preferred developments of the method according to the invention are the subject matter of claims 2 to 7.

Advantageous and preferred embodiments of the invention are explained below with reference to figures. It shows:

Fig. 1 shows schematically an extraction of significant scenes according to an embodiment of a method according to the invention and

Fig. 2 shows schematically an embodiment of a working using the method according to the invention quality control station. A method according to the invention uses the result of a performed objective quality assessment of audio / video material. This is a series of measurements consisting of n discrete measured values of one quality feature or several quality features, each measured value being temporally related to the examined audio / video content (see FIG. 1).

(X ₁ , tι), ..., (x _m t ") e A

(Measurement series with a measured variable x and n measured values)

(xi, y \, z \, h), ..., (x _n , y ", z _n , t") € A (series of measurements with three measured quantities x, y, z and n measured values in each case)

It is expressly pointed out that the above-mentioned series of measurements with three measured quantities is only a very specific embodiment. The invention may e.g. be carried out with a series of measurements, which comprises two measured variables or four measured variables or even more measured variables, each measured variable represents a quality feature. Such a multi-dimensional series of measurements is, of course, nothing more than a total of two, three, four or more series of measurements, each representing a separate measured variable, the measured values of the individual measured variables being recorded synchronously, as described in the three-dimensional series of measurements shown in the preceding paragraph an embodiment is shown.

Possible measured variables (quality features) in the video area include:

PSNR (Peak Signal to Noise Ratio) - MSE (Mean Square Error),

Blurriness - Degree of "Blurriness", "Blurred", Blockiness - Degree of training of block artifacts, mainly caused by compression methods, and

Jerkiness - measure for jerking video presentation, not constant refresh rate, mainly caused by transfer problems.

These and other video quality features as such are described in detail in:

Winkler, Stefan: Digital Video Quality - Vision Models and Metrics.- John Wiley & Sons

Ltd.- Chichester.- 2005.- ISBN 0-470-02404-6. Possible measured variables (quality features) in the audio domain include:

SNR (Signal to Noise Ratio) - Signal to Noise Ratio

Max Peak - maximum audio level,

Analog over - Display of overdrive when processing analog audio signals 5 Ie,

Digital Clipping - Display of clipping when processing digital audio data,

Digital Zero - digital zeros and

Block repeating - block repeats caused when reading in CDs. 0

These audio quality features as such are familiar to the person skilled in the art, so that a more detailed discussion on this point is unnecessary.

Fixed conditions (rules R) are applied to the set of measurements to elicit the temporal sections of the material being analyzed that conform to the conditions.

A condition represents the comparison of a measured value with the limit value (s) set for its measured variable, the condition being met precisely if a lower or .0 exceeding of the limit values is equivalent to a lower audio / video quality.

Corresponding limit values are defined for each measured variable.

Rl: hc _u <x <Ix ₀

R3: lz _u ≤ z <lz ₀

lx _u lower limit for the characteristic x ho upper limit for the characteristic x 0

In order to suppress the effects of exceeding or exceeding limit values due to measurement errors or spontaneous outliers, which are not perceived by human observation, the conditions can be extended to several temporally successive measured values. Rl: (Ix ₁₁ <x _m <Ix ₀ ) Λ (Ix ₁₁ <x _m + ₁ <Ix ₀ ) A ... Λ (lx _u <x _{m + tw} <Ix ₀ ) tw defined time window for the feature x

In the case of several measured variables included in the evaluation, the respectively defined conditions are logically linked together.

R = R1 V R2 V R3

After the conditions have been applied, a set of temporal coordinates is available, each representing the sections that conform to the conditions and are therefore considered poor in technical terms.

(x _a , t _a ), ..., ( _z , t _z ) e B (filter result with a measured variable x)

(x _a , y _a , z _a , t _a ), ..., (x _z , y _z , z _z , t _z ) € B (filter result with three measured quantities x, y, z)

Based on this result set, the identified sections of the audio / video material can be addressed directly on the basis of the time information and thus automatically forwarded to a person for the purpose of subjective follow-up.

For example, the disclosed invention enables the simplification of a quality control process for audio / video content. Such a control process can, as shown in Fig. 2, be largely automated. The system specifically presents the quality supervisor with scenes that have been identified as deficient. This considerably reduces the time required compared to a manual full control. In addition, the influence of subjective inaccuracies such as fatigue or fluctuating attention, which plays a significant role during a prior art manual full control, is greatly reduced. An example of an application of a device according to the invention in the context of an automated quality control station is shown schematically in FIG. 2 and will be described below.

5 The audio / video content to be checked for its technical quality is stored in a directory monitored by a control system of the device. The analysis then begins automatically using the appropriate analysis tools built into the system. When this analysis is completed, the system evaluates the results according to the procedure described above. This can be done by setting appropriate limits in the system

10 are defined. These can be managed as rating profiles within the application. On the one hand, the results are presented visually as a chart in which limit value violations are prominently displayed, and on the other presented as a scene list. The scenes can be selected by the quality inspector, viewed directly via an integrated playback facility (eg media player) and displayed on an online form.

15 evaluate and document.

In another embodiment of a device according to the invention is an automatically monitored input directory for the reception of video files ready. The delivery of a video file automatically creates an objective quality analysis.

20 lyse, which as a result provides a file with the values of the individual measured variables - in the specific embodiment, blurriness, blockiness and jerkiness. Following the quality analysis, the measured value file is automatically evaluated. In this case, the measured values of the individual measured variables are compared with the threshold values defined for the respective measured variable, with additionally defined time windows acting as a filter criterion.

> 5

Subsequently, a diagram is generated for each measured variable, which represents the time course of the measured variable. By means of the method according to the invention according to the previously defined rules found quality critical scenes are clearly highlighted. An automatically generated table will all be considered as quality-critical.

SO graded scenes with associated timecode listed. It marks the quality characteristic that was decisive for the classification as "quality-critical" and the extent to which a threshold value was undershot or exceeded.Each listed scene can be started and played by a reference, while a form is provided for the selected scene. which allows a rating according to the subjective analysis. Already manually checked scenes are optically marked in the list. By means of a symbolism it can be recognized whether this scene was also judged to be quality critical in the manual, ie subjective, evaluation.

Claims

claims

1. A method for the automatic objective technical quality assessment of the content of audio and / or video material and for filtering out quality-critical sequences comprising the following steps: a) recording a series of measurements of a plurality of measured values of a quality feature of the audio and / or video content and Recording at least one more series of measurements of at least one other quality characteristic of the audio and / or video content, each sample having a temporal reference to the audio and / or video content being examined; b) using predetermined, for each individual quality characteristic series of measurements of specific conditions on the measured values obtained in step a) in order to filter out those time segments of the material analyzed which are in conformity with the conditions mentioned, one condition being a comparison of a measured value with an upper limit value fixed for the measured variable and / or with a correspondingly stipulated under c) exports a logical combination of the filter results of all quality features and d) indexes the quality-critical audio and / or video sequences as a function of the result of the logical combination.

2. A method according to claim 1, characterized in that one of the following quality features is used as a quality characteristic for the objective technical quality assessment: blurriness, blockiness, jerkiness, MSE, PSNR, digital zeros, digital overload, block repetitions, analog overload, signaling

Noise distance, maximum level.

3. The method according to any one of the preceding claims, characterized in that at least one of the conditions according to step b) is treated as fulfilled only if the upper or the lower limit is violated at least for a predetermined contiguous time.

4. The method according to any one of the preceding claims, characterized in that in step a) measured exactly three quality characteristics and the filtering procedure according to steps b) and c) are based.

5. The method according to claim 2, characterized in that two or more of the quality features mentioned in claim 2 are used for the objective technical quality assessment.

6. The method according to any one of the preceding claims, characterized in that the indexed in step d) quality-critical audio and / or video sequences are addressed directly and reproduced on a playback device.

7. The method according to any one of the preceding claims, characterized in that the indexed in step d) quality-critical audio and / or video sequences are automatically visualized as a chart and / or as a scene list.

8. Apparatus for carrying out the method according to one of the preceding claims, characterized in that

- The device is designed so that it can accept the audio and / or video content to be evaluated and processed automatically by the above method and that

- The device is further designed so that it allows a definition and technical management of different sets of thresholds as evaluation profiles for different audio and / or video content.