WO2004100558A1

WO2004100558A1 - Method for protecting against errors in a sequence of video images

Info

Publication number: WO2004100558A1
Application number: PCT/FR2004/001046
Authority: WO
Inventors: François Marx; Joumana Farah-Francis
Original assignee: France Telecom; Centre National De La Recherche Scientifique; Institut National Polytechnique De Grenoble
Priority date: 2003-05-05
Filing date: 2004-04-29
Publication date: 2004-11-18
Also published as: FR2854755A1

Abstract

The invention relates to a method (1) for protecting a flux of compressed video images from errors occurring during the transmission of said flux towards a non-reliable link, redundancy bits being transmitted with the compressed video images. Said method is carried out by a video emitter comprising a video encoder, a processing unit and at least one channel encoder, all of the compressed video images originating from an input video sequence (2). The inventive method comprises the following steps: the processing unit is used to calculate a redundancy bit distribution function (4), said distribution, which depends on the position of the compressed video images in a video sub-sequence, being calculated such that an average distortion of the transmitted sequence is minimised; and for each compressed video image, one of the channel encoders is used to add (5) redundancy bits to the data encoding said image in such a way that the number of redundancy bits transmitted varies according to the distribution function. The invention also relates to an error correction method carried out by a video receptor.

Description

PROTECTION AGAINST ERRORS OF A SEQUENCE OF VIDEO IMAGES

Field of the invention

The invention relates to the field of telecommunications and, in this field, to methods of coding video images and more particularly to methods of protection against errors occurring during transmission of video images over an unreliable link. An unreliable link is a means of transmission on which a so-called digital error-sensitive communication is carried out. In general, this means is a radio link. In the rest of the document, radio link and unreliable link mean the same thing. The invention further relates to reception error correction methods, to a transmitter and a receiver implementing a protection or correction method as well as to a program implementing one of these methods and a support for recording of this program.

In the rest of the document, the terms frame I and frame I designate an image resulting from a video sequence which is compressed without using other reference images. In practice, this means that this image can be decompressed independently of the other images. In this document frame and image means the same thing.

In the remainder of the document, the terms frame P and image P denote an image resulting from a video sequence, which is compressed using other reference images. In practice, this means that only the difference between two successive images is coded. These coded images are less bulky, therefore less demanding in transmission rate, but they require good reception of the previous images in order to be able to be correctly decoded. P and I define the type of the image. In the field of digital communications, a channel coder is a device which adds additional bits to a video sequence to be transmitted, known as redundancy bits, with the aim of enabling a channel decoder to be able to correct transmission errors by exploiting these bits redundancy during reception. The turbo codes, the convolutional codes, the Reed-Solomon codes all correspond to techniques implemented by a channel coder for the aforementioned purpose. Prior art Currently, radio networks such as GSM or UMTS do not offer reliable communications since the transmitted signal is marred by many transmission errors. These errors come in particular from additive noise on reception and the frequency selectivity of the propagation channel (multi-path channel).

To reduce the transmission rate of a video stream, a video encoder is commonly used in transmission which implements video image compression techniques.

When transmitting compressed video images, errors are very penalizing. The more compressed the transmitted data, the more they are susceptible to errors. Due to the use of compression techniques, each bit contains a significant amount of information; the images are coded relative to each other. Thus, an error on the first image leads to a succession of errors on the rest of the sequence. Therefore the data must be imperatively protected.

To combat errors, the known coding methods, before video transmission over an unreliable link, consist in adding information to the video data supplied by the video coder and this before transmission.

A first known technique is the ARQ mechanism, an abbreviation of the English terms Automatic Repeat Request which consists in repeating the erroneous packets.

This technique produces a good quality video stream, however it is sensitive to the delay in the transmission network and can hardly be used in a service with real-time constraints such as a videophone service on a mobile phone. A second known technique which can be combined with the previous one, consists in using an error correcting coder. The latter adds redundancy in the data to be transmitted to allow the reconstruction of the video stream on arrival with as few errors as possible. Currently, the video coder divides the video sequence into sub-sequences consisting of an image I and a given number of P images. The P images are coded with respect to previous images. The error correcting coder or channel coder conventionally codes each video image in the same way by adding to each image a number of redundancy bits proportional to the size of the image. This conventional method has the advantage of requiring a single channel coder with a single coding rate. But this method lacks effectiveness against an error which disturbs an image P located relatively close to an image I: this error continues until the next image I.

As a result of this technique, more elaborate methods have been developed and are known by the acronym UEP, abbreviation of the English terms Unequal Error Protection which means unequal error protection. They consist in dividing the video stream into blocks of different relative importance. One of these methods is the subject of the article "An FEC-Based Error Control Scheme for Wireless MPEG-4 Video Transmission" whose authors are J.Cai, Q.Zhang, W.Zhu and CWChen and which has for IEEE reference WCNC2000, Chicago, Sept. 2000. An example of the application of this method is as follows. Each frame is divided into several different blocks which are as follows:

- a first block contains the image header and general parameters such as the image number, its type (LP, ...),

- a second block contains the motion vectors which describe the similarities between parts of the current image compared to parts of a previous image,

- a third block contains the coefficients which improve the reconstruction of the image after the translations of the motion vectors.

According to this method, the headers of the frames are the most protected since they contain the most important bits, that is to say the information on the format of the frames. For example, they are repeated an odd number of times. The data containing the movement information has a good level of protection, for example due to a 1/3 efficiency turbo-coding. Texture data, on the other hand, have the lowest level of protection, for example due to a simple convolutional coding of yield Vz. It is known that even in the absence of this data, the video decoder can carry out the motion compensation step so as to reconstruct the frame without too much degradation. The aim of this method is therefore to keep the transmission rate constant by sacrificing the coefficients in favor of the header bits which are considered more important for having a good quality video. However, this method has a certain number of drawbacks which are as follows: video quality not necessarily improved, partitioning of the video stream often difficult because the three blocks are not at fixed positions, a necessary reorganization of the video stream in transmission as in reception so as to group the data resulting from a first video coding into several different blocks which increases the complexity of the video coder / decoder, particularly in the case of real-time videophone applications.

An object of the invention is to propose a method of protection against errors which occur during the transmission of data corresponding to a stream of video images on an unreliable link, the set of video images constituting a sequence, which puts implement a channel coder more efficiently than the conventional method and which is simpler than the known UEP methods.

Statement of the invention

To this end, the subject of the invention is a method of protecting a stream of compressed video images against errors which occur during the transmission of this stream over an unreliable link consisting in transmitting redundancy bits with the images compressed video, implemented by a video transmitter comprising a video coder, a processing unit and at least one channel coder, all of the compressed video images being from an input video sequence. The method comprises the steps which consist in: using the video coder to share the input video sequence into output sub-sequences each consisting of a compressed video image I followed by a determined number of compressed video images P, each compressed video image being coded in the form of data, - to calculate by the processing unit a redundancy bit distribution function, this distribution which is a function of the position of the compressed video images in a video sub-sequence is calculated so to minimize an average distortion on the transmitted sequence, - for each compressed video image, to be added by means of one of the channel coders redundancy bits to the data coding this image so that the number of redundancy bits transmitted varies according to the distribution function. Thus, channel coders add redundancy bits to the data coding each compressed video image to be transmitted. The number of redundancy bits associated with a compressed video image varies according to a distribution which is a function of the position of the image in a video sub-sequence. The distribution function is calculated by the processing unit so as to minimize an average distortion on the transmitted sequence.

By calculating a distribution of the redundancy bits as a function of the position of the compressed video image in a video sub-sequence which minimizes the distortion average over the transmitted sequence, the method determines a non-uniform distribution of the redundancy.

Advantageously, the method allows a gain on the video quality by improving the protection by means of a very redundant correction code on the first images of the sequence. Although the last images of each sub-sequence are less protected, the overall transmission rate being assumed to be constant, this is not too penalizing because the propagation error depends on the position of the distorted image in the sub-sequence . The last images being by definition at the end of the sub-sequence, the influence of their distortion is weak on this sub-sequence. The distribution of the redundancy bits is such that the ratio composed of the number of transmitted redundancy bits associated with a compressed video image P divided by the size in bits of this compressed video image P decreases throughout each subsequence limited to l 'all compressed video images P. This illustrates the fact that the protection is greater on the first images of the sub-sequence and decreases throughout the sub-sequence.

According to a particular embodiment of a method according to the invention, the method inserts a parameter defining the distribution of the number of redundancy bits associated with this image. This insertion is carried out by means of the channel coders in a header of each compressed video image to be transmitted. This advantageously makes it possible to know the distribution of the redundancy bits after transmission of the stream, by extracting the parameter during reception.

According to an alternative to this embodiment, the method inserts parameters defining the given distribution of the number of redundancy bits transmitted and associated with a compressed video image. This insertion is carried out by means of the channel coders in the data to be transmitted during a preliminary step of establishing the transmission of the stream. This alternative makes it possible to avoid transmitting the redundancy distribution to each image.

The invention further relates to a video transmitter for transmitting a stream of compressed video images over an unreliable link, all of the compressed video images corresponding to an input video sequence. The transmitter comprises: a video coder suitable for dividing the input video sequence into output sub-sequences each consisting of a compressed video image I followed by a determined number of compressed video images P, each compressed video image being encoded as data, at least one channel coder suitable for adding redundancy bits to the data coding each compressed video image to be transmitted and for varying the number of redundancy bits to be transmitted according to a distribution of the number of transmitted and associated redundancy bits to a compressed video image, a processing unit adapted to control each channel coder and to calculate the redundancy distribution which is a function of the position of the compressed video image in a video sub-sequence.

The invention further relates to an error correction method implemented by a video receiver, these errors occurring during the transmission of a stream of compressed video images on an unreliable link, all of the images Compressed video resulting from a video sequence divided into sub-sequences made up of a given number of compressed video images. The method consists in correcting the transmission errors by means of redundancy bits present in the stream and the number of which varies as a function of a distribution determined during transmission. The method includes the step of: determining by means of a channel decoder the distribution of the redundancy bits defined in a header of the image or at the start of the transmission, this distribution of the number of redundancy bits transmitted and associated with a compressed video image being a function of the position of this image in a subsequence. The invention further relates to a video receiver for receiving a stream of compressed video images after transmission over an unreliable link, all of the compressed video images being from a video sequence shared in sub-sequences. consist of a given number of compressed video images. The receiver includes a channel decoder to correct transmission errors by means of redundancy bits present in the stream. This channel decoder is suitable for determining the distribution of the redundancy bits, this distribution of the number of redundancy bits transmitted and associated with a compressed video image being a function of the position of this image in a sub-sequence and being defined in an head of the image or at the start of the transmission. Thus, by knowing the distribution of the redundancy the channel decoder can correct the errors in the transmitted stream.

The invention further relates to a mobile terminal comprising a video transmitter according to a previous object. This terminal is typically a device equipped with a telephone function such as a telephone, or a personal assistant known by the acronym PDA abbreviation of the Anglo-Saxon terms Personal Digital Assistant. The invention further relates to a mobile terminal comprising a video receiver according to a previous object. This terminal is typically a device equipped with a telephone function such as a telephone, or a personal assistant known by the acronym PDA abbreviation of the Anglo-Saxon terms Personal Digital Assistant. The invention further relates to a computer program product which can be loaded directly into the internal memory of a digital video transmitter, comprising portions of software code for executing the steps of a method for protecting a data stream. video images compressed against errors according to a previous object of the invention, when the program is executed on the digital video transmitter. The invention further relates to a medium usable in a digital video transmitter and on which is recorded a computer program product according to the preceding object. This support can be an electronic component such as a memory, a microprocessor installed for example on a SIM card or permanently in the transmitter.

The invention further relates to a computer program product loadable directly into the internal memory of a digital video receiver, comprising portions of software code for the execution of the steps of the error correction method according to a previous object. of the invention, when said program is executed on the digital video receiver.

The invention further relates to a medium usable in a digital video receiver and on which is recorded a computer program product according to the preceding object. This support can be an electronic component such as a memory, a microprocessor installed for example on a SIM card or permanently in the receiver.

It should be noted that in order to compare techniques for protecting the video stream with one another, one must reason at a constant transmission rate.

Other characteristics and advantages of the invention will become apparent from the following description of particular and nonlimiting examples of embodiment of the invention; these examples being described with reference to the following appended figures given by way of illustration. Brief description of the figures

FIG. 1 is a flow diagram of a method of protection against errors according to a first object of the invention.

FIG. 2 is a diagram of a video sequence divided into sub-sequences.

FIG. 3 is a flow diagram of a particular embodiment of a method of protection against errors according to a first object of the invention. Figure 4a is a diagram of a video transmitter according to a second object of the invention.

FIG. 4b is a diagram of a video receiver according to a third object of the invention. Figure 5 is a flow diagram of an error correction method according to a fourth object of the invention.

Detailed description of embodiments of the invention A method of protection against errors according to the invention takes place in several stages illustrated by FIG. 1. The method is implemented by a video transmitter comprising a video encoder, a video unit processing and at least one channel encoder.

This method 1 is a method of protection against errors which occur during the transmission of data corresponding to a stream of video images on an unreliable link. Process 1 has as input data a video sequence 2 from which all the compressed video images to be transmitted are derived. Method 1 comprises a first step 3 which consists in sharing by means of a video transmitter video coder the video sequence 2 in sub-sequences each consisting of a compressed video image I followed by a determined number of compressed video images P. This sharing of a video sequence 2 into sub-sequences 2.1, 2.2 is illustrated in FIG. 2. According to a particular embodiment, the determined number of images P is twenty-four for an image I, which corresponds to a rate of twenty-five images per second. Each compressed video image is encoded as data by the video encoder.

The method comprises a second step 4 which consists in calculating the redundancy provided to each image. The video stream is protected by the redundancy bits from the channel encoder. For each image, the number of redundancy bits is calculated according to its position in the sub-sequence and so as to minimize the distortion of the sub-sequence. The distribution of the redundancy bits is such that the composite ratio of the number of transmitted redundancy bits associated with a compressed video image divided by the bit size of the compressed video image decreases throughout each sub-sequence. In each sub-sequence, only the P images are considered for the calculation of the distribution.

According to a particular embodiment of the second step of the method, the calculation 4 of the distribution of. bits of redundancy by the processing unit consists in acquiring by the processing unit a parameter R whose value indicates a number of time boundaries within a sub-sequence as well as a set of channel encoder types or than a set of coding rates in the case of a single channel coder with punching. The boundaries determine the type of channel coder or the coding rate among those of the game to be used to add redundancy bits in the domains to be transmitted corresponding to each P image. The calculation also consists in determining by the processing unit the values of the time boundaries t _r which make it possible to minimize the distortion due to transmission errors of the video images transmitted while respecting a bit rate constraint.

The distortion is expressed by the following formula:

SE _ch

and the flow constraint is expressed by the following formula

R

∑C _r (t _r -t _r _ ₁ ) -C _m T = 0. r = l

These formulas are such as:

- R is the number of types of channel encoder or channel coding rate considered, C _r , redundancy, is the inverse of the performance of the channel encoder r, - C _m T characterizes the transmission rate,

- P _e (r) is the probability of error for a given Signal to Noise ratio, a given channel and a type of channel coder or a coding rate,

- σ ² characterizes the impact of an error on the frame,

- f _Sat a function which characterizes the effect of the error, - γ is the attenuation of the error over time,

- t _r , r = 0..R are the limits of the regions, t _r -t _r-1 is the number of frames in the region r,

- to = 0,

- t _R = T, length of the subsequence. An example of calculation of a distribution of the redundancy bits is given below.

According to this example, the number of channel coding rates in the transmitted stream is limited to three, R = 3. Thus, each sub-sequence limited to the set of images P is split into three parts to which correspond three channel coding rates. All three rates form a game. Let T be the number of images in a sub-sequence, ti and t ₂ of the time boundaries in a sub-sequence such that 0 <t _x <t ₂ <T.

The image I at the start of the sub-sequence corresponds to the time t = 0 and the last image P corresponds to t = T. A subsequence is thus formed of an image I and of T images P.

Image I is assigned a channel coding rate called average redundancy, the choice of which is conventionally made. Thus this rate can be the result of the division between the number of data bits to be transmitted and the available bit rate. For example, 128kbits, number of bits to be transmitted, divided by 512kbits, available bit rate, which gives a VA yield, ie a redundancy equal to 4. The redundancy is the inverse of the code yield. In the following, only the P images are considered.

The images P located between 1 and X _\ are assigned a first channel coding rate Ai.

The images P situated between tr + 1 and t ₂ are assigned a second rate of channel coding A ₂ .

The images P located between t ₂ + 1 and T are assigned a third rate of channel coding A ₃ .

The different rates Ai, A ₂ , A ₃ of channel coding are determined with the following constraint: - whatever the set (A _ls A ₂ , A ₃ ) of channel coding rate, the transmission rate is constant.

To determine the game which allows the best video quality in reception, it is necessary to find a better compromise between the number of over-protected images and the number of under-protected images. This compromise is found when the average distortion or mean global distortion according to English terminology for the entire video sequence is minimized. This distortion is expressed by the mean square error or Mean Square Error, MSE, according to Anglo-Saxon terminology or the Peak Signal to Noise Ratio, or PSNR, according to Anglo-Saxon terminology. The MSE is used to quantify video degradation due to video compression, transmission errors and video decoding. It is calculated between the decoded sequence S _d and the original sequence S ₀ . The decoded sequence Sa is that transmitted via an unreliable link and then decoded at the output of the video decoder of the video receiver. The original sequence S ₀ is that at the input of the video encoder of the video transmitter. For a sequence of size NxM and length L, the global MSE is expressed in the form:

In the case of an eight-bit image coding, the PSNR can be evaluated by the expression:

Let S _{e be} the sequence decoded after a transmission via an ideal link. The distortion of S _e is only due to the coding on transmission and the decoding on reception, that is to say is only linked to the quantization. Under these conditions the distortion due to video compression is expressed in the form: SE _quml =

The global MSE, (coding + transmission + decoding) is expressed in the form:

MSE _g = MSE _quant + MSE _ch

MSE _ch is the distortion introduced during transmission and due to the unreliable link.

MSE _quant is independent of the channel coding scheme and distribution. It only depends on the sequence, the structure of the video encoder and the structure of the video decoder. It is considered constant for the invention.

Under these conditions, to minimize the overall MSE _g distortion, it suffices to minimize the MSE _ch distortion introduced during transmission. This distortion MSE _ch can be expressed as the sum of the distortions of each subsequence of length T + l:

MSE _ch = ∑MSE _ch {t)

(= 0

To solve the minimization problem, the MSE transmission distortion _Ch (t) can be modeled by the expression:

MSE _ch {t) = σ ² - ^ - l + γt with σ ² the intensity of an error which occurs on the first image of a subsequence, γ = 0 is the worst case for which there is no attenuation of the error over time, γ = ∞ is the case for which there is no propagation of error over time; the images of a sub-sequence are coded independently of the previous images, σ and γ are determined experimentally for a given video sequence and are considered as initial parameters.

The method comprises a third step 5 which consists in adding redundancy bits in the data to be transmitted by means of a channel coder of the video transmitter using the information determined in step 4 on the amount of redundancy to be provided. . The redundancy bits are obtained by known coding techniques which implement a particular type of coding, the best known of which are the following: turbo-codes, convolutional codes, Reed-Solomon codes. According to a particular embodiment, the addition of redundancy bits is constant whatever the type I or P of the image. In this case, only one channel encoder is required. The variation in the number of transmitted redundancy bits is obtained by punching the bits at the output of the channel coder by a technique known as punching, or puncturing according to English terminology, implemented by the channel coder. Such punching can be carried out by means of a turbo-encoder specified in the UMTS standard in the document 3GPP, "Technical Specification Group: Multiplexing and channel coding (TDD)" TS25.222 V3.2.0, March 2000, with for example yields 1/3, 2/5 and ^l A in the case where R is equal to three. Although the standard only specifies a 1/3 efficiency coder, the other efficiencies can be obtained using the mechanism known as rate adaptation or rate-matching according to English terminology. For a sub-sequence made up of an image I and 10 images P, an example of distribution is illustrated by the following table which gives the coding rate or output, according to the number of the image:

The following example is given by way of illustration. This example relates to a videophone service within the framework of 3rd generation ^telephony . This example corresponds to a particular embodiment in which the addition of redundancy bits is constant whatever the type of image, the variation is obtained by removing redundancy bits from the video stream before transmission by a technique called punching or puncturing according to Anglo-Saxon terminology. Such punching can be performed by means of a turbo-encoder specified in the UMTS standard in the 3GPP document previously cited with yields 1/3, 2/5 and Vz for an R equal to three. The basic encoder has an efficiency of 1/3, Cι = 3. The efficiency of 2/5, C ₂ = 5/2, is obtained by punching one bit out of 6 of the 1/3 coder. The efficiency of Vz, C ₃ = 2, is obtained by punching one bit out of 3 of the 1/3 coder. In this example, the punching technique makes it possible to have the equivalent of three coders of different types with a single so-called basic channel coder with which three different outputs are associated: 1/3, 2/5 and Vz.

The video sub-sequence considered consists of an image I followed by 27 P images. The parameters below are linked to the video sequence and / or set by the service provider: γ = 0.055 σ ² - 13355

T = = 27 α = = 100238 β = = 2413

The method then seeks to minimize the distortion MSE _ch under the rate constraint E.

The MSE _Ch equation can be written in full form:

T τ-t _i T τ-t,

+ σ 'P ₂ ff - - dt dτ + σ ² P ₃ ff - î— dt dτ

₂ ^, 4 ^J. 1 ¹ + ⁺ γ ^γ t ^t ? iot 1ι + ⁺ γt

with the flow constraint: E = Citi + C ₂ (t ₂ - 1) + C ₃ (T - 1 ₂ ) - C ₂ T The process forms the Lagrangian F = MSE _ch + λ.E:

By deriving the equation of MSE _Ch , the process obtains:

The method combines the previous equations to obtain a new

equation of which ti is the solution:

(C ₂ -C ₃ ) (P ₁ -P ₂ ) Ln [l + γ (Tt ₁ )] - (C ₁ -C ₂ ) (P ₂ -P ₃ ) Ln 1 + γ Cl ^~ ^ 2 = 0

C ₂ - C ₃ t ₂ is linked to t _\ by the relation: t ₂ = T - C1 - C2

The values of probabilities P _e of errors for the different coders (for a given performance) and the signal-to-noise ratios of the channel are as follows (these are data fixed according to the video sequence and / or fixed by the supplier of service for example):

The digital application leads to

Thus, for a very noisy channel (Eb / N0 = 13db), the process codes the first nine images (t ₁ = 9) with the 1/3 coder, the following eight images (t ₂ -tι -1 = 18-9 -

1 = 8) with the encoder 2/5 and finally the last ten images (Tt ₂ + l = 27-18 + l = 10) with the encoder Vz.

In the case of a very good quality channel (Eb / N0 = 17db), the process codes the first image (tι = l) with the 1/3 coder, the 24 images (t ₂ -tι-l = 26- ll = 24) following with the 2/5 encoder and finally the last 2 images (Tt ₂ + l = 27-26 + l = 2) with the encoder

Vz.

A particular embodiment of a method of protection against errors according to the invention takes place in several stages illustrated by FIG. 3. The first three stages are identical to those described with reference to FIG. 1 and are not described again. During a fourth step, the method inserts a parameter 6 in a header of each compressed video image to be transmitted. This parameter defines the distribution of the number of redundancy bits associated with the image corresponding to the header. The insertion 6 into a header of a compressed video image is carried out by the channel coder which codes the image. FIG. 4a is a diagram of a video transmitter 7 according to the invention. The video transmitter 7 implements a method 1 for protection against errors according to a previous subject of the invention. Only the part of the transmitter 7 specific to the invention is shown, the transmitter 7 further comprising conventional elements such as a transmission antenna, transmission electronics, etc. The specified part consists of an 8 video encoder, a processing unit 9 and a suitable 10 channel encoder.

The video coder 8 is adapted to divide the video sequence into sub-sequences each consisting of a compressed video image I followed by a determined number of compressed video images P. According to a particular embodiment, the video coder 8 shares the sequence in twenty-five image sub-sequences comprising an I image and twenty-four P images.

The processing unit 9 controls each 10-channel encoder. It 9 is suitable for calculating the distribution of the number of redundancy bits transmitted and associated with a compressed video image based on the position of the compressed video image in a video subsequence. It 9 communicates this distribution to the 10-channel coder. This distribution is a function of the position of the images in a sub-sequence. The distribution is calculated so as to minimize an average distortion on the transmitted sequence. Each redundancy bit is associated with a video image

The 10-channel coder is suitable for adding redundancy bits to the data to be transmitted and for varying the number of redundancy bits to be transmitted according to the distribution communicated by the processing unit. The redundancy bits are obtained by known coding techniques such as turbo-codes, convolutional codes, Reed-Solomon codes. According to a particular embodiment, the addition of redundancy bits is constant whatever the type of the image: the 10-channel coder adds redundancy bits in constant number regardless of the type of the image and varies the number of redundancy bits transmitted by removing redundancy bits from the video stream before transmission using a so-called punching technique.

FIG. 4b is a diagram of a video receiver 11 adapted according to the invention for the implementation of an error correction method described below.

An error correction method according to the invention takes place in several stages illustrated in FIG. 5. The method 12 has as input data a stream of video images corresponding to a video sequence divided into sub-sequences consisting of a given number of video images, emitted by a transmitter according to the invention. The data 13 corresponding to the flow is tainted with errors. These errors were introduced during the transmission of data on an unreliable link. The method 12 consists in correcting the transmission errors by means of redundancy bits added by the transmitter and transmitted with the data 13 constituting the stream. The addition of redundancy bits follows a determined distribution such that the number of redundancy bits transmitted and associated with a compressed video image is a function of the position of this image in a subsequence. The distribution is calculated before transmission so as to minimize an average distortion on the transmitted sequence. The method 12 comprises a first step 14 which consists in determining by means of a channel decoder the distribution of the redundancy bits. This information is inserted into the stream during transmission. According to a particular embodiment, the channel decoder extracts from the header of each transmitted compressed video image, a parameter which defines the distribution of the number of redundancy bits associated with this image. Knowing the distribution of the redundancy bits, the method corrects in a second step 15 the data which is marred by transmission errors according to known methods.

FIG. 4b is a diagram of a video receiver 11 adapted according to the invention for correcting transmission errors according to an error correction method described above. Only the part of the receiver 11 specific to the invention is represented, ie a suitable 16-channel decoder. The receiver 11 also comprises conventional elements such as a reception antenna, reception electronics, a video decoder, etc. The receiver li has as input data a stream of video images corresponding to a video sequence divided into sub-sequences made up of a given number of video images, emitted by a transmitter according to the invention. The data corresponding to the flow is tainted with errors. These errors were introduced during the transmission over an unreliable link of the data corresponding to the flow.

The 16-channel decoder corrects transmission errors by means of redundancy bits added by the transmitter in the data constituting the stream. The 16-channel decoder determines the distribution of the redundancy using the information on the distribution of the redundancy introduced into the stream during transmission. This distribution of the number of redundancy bits transmitted and associated with a compressed video image is a function of the position of this image in a sub-sequence and is calculated on transmission so as to minimize an average distortion on the transmitted sequence.

According to a particular embodiment of the 16-channel decoder, the latter extracts from the header of each compressed video image to be transmitted a parameter which defines the distribution of the number of redundancy bits associated with this image.

Claims

1. Method (1) for protecting a stream of compressed video images against errors which occur during the transmission of this stream over an unreliable link consisting in transmitting redundancy bits with the compressed video images, implemented by a video transmitter (7) comprising a video coder (8), a processing unit (9) and at least one channel coder (10), all of the compressed video images being from a video sequence (2) d input, the method comprises the step which consists in: - sharing (3) by means of the video coder (8) the sequence (2) input video in output sub-sequences each consisting of a compressed video image I followed by a determined number of compressed P video images, each compressed video image being coded in the form of data, characterized in that the method (1) further comprises the steps which consist in: - calculating (4) by processing unit (9) a bit distribution function redundancy, this distribution which is a function of the position of the compressed video images in a video sub-sequence is calculated so as to minimize an average distortion on the transmitted sequence, the average distortion being expressed by the mean square error MSE or the ratio PSNR peak noise signal,

- for each compressed video image, to be added (5) by means of one of the coders (10) channel of the redundancy bits to the data coding this image so that the number of transmitted redundancy bits varies according to the distribution function .

2. Method (1) for protection against errors according to claim 1, in which the distribution of the redundancy bits is such that the ratio composed of the number of transmitted redundancy bits associated with a compressed video P image divided by the size in bits of this compressed video P image decreases throughout each sub-sequence limited to all of the compressed video P images.

3. Method (1) for protection against errors according to one of claims 1 and 2, further comprising the step which consists of:

- Inserting (6) by means of the channel coders (10) in a header of each compressed video image to be transmitted, a parameter defining the distribution of the number of redundancy bits associated with this image.

4. Method (1) for protection against errors according to one of claims 1 and 2, further comprising a preliminary step of establishing the transmission of the stream which consists in: inserting (6) in the data to be transmitted by means coders (10) channel, parameters defining the given distribution of the number of redundancy bits transmitted and associated with a compressed video image.

5. Method (1) for protection against errors according to one of claims 1 to 4, in which the calculation (4) of the distribution of the redundancy bits by the processing unit (9) consists of:

- to acquire by the processing unit (9) a parameter R whose value indicates a number of time boundaries within a sub-sequence as well as a set of types of channel coder (10), these boundaries determining the type of coder (10) channel among the different types of game to be used to add (5) redundancy bits in the data to be transmitted corresponding to each P image,

- to determine by the processing unit the values of the time boundaries t _r which make it possible to minimize the distortion due to transmission errors of the video images transmitted while respecting a bit rate constraint, the distortion being expressed by the following formula:

and the flow constraint being expressed by the following formula

∑C _r (t _r -t _r _ ₁ ) -C _m T = 0,

formulas in which: - R is the number of types of channel coder considered,

- C _r , redundancy, is the inverse of the efficiency of the channel encoder r, C _m T characterizes the transmission rate,

- P _e (r) is the probability of error for a given Signal to Noise ratio, a given channel and a type of channel coder, - σ ² characterizes the impact of an error on the frame, - f _sat a function which characterizes the effect of the error,

- γ is the attenuation of the error over time,

- t _r , r = 0..R are the limits of the regions, t _r -t _r-1 is the number of frames in the region r, - to = 0,

- t _R = T, length of the subsequence.

6. Method (1) for protection against errors according to one of claims 1 to 4 in which: - whatever the image P of each sub-sequence, the addition (5) of the redundancy bits is carried out by a single (10) channel encoder and,

- for all the P images of a sub-sequence, the variation in the number of redundancy bits is obtained by punching by the coder (10) channel the bits at the output of this coder (10) channel with a coding rate determined for each image P.

7. Method (1) for protection against errors according to claim 6, in which the calculation (4) of the distribution of the redundancy bits by the processing unit (9) consists of:

- Acquiring by the processing unit (9) a parameter R whose value indicates a number of time boundaries within a sub-sequence as well as a set of coding rates, these borders determining the coding efficiency at use to add (5) redundancy bits in the data to be transmitted corresponding to each image P,

and the flow constraint being expressed by the following formula:

formulas in which:

- R is the number of channel coding rates considered,

- C _r , redundancy, is the inverse of the efficiency of the channel r encoder,

- C _m T characterizes the transmission rate, - P _e (r) is the probability of error for a given Signal to Noise ratio, a given channel and a coding rate,

- σ characterizes the impact of an error on the frame,

- f _sat a function which characterizes the effect of the error,

- γ is the attenuation of the error over time, - t _r , r = 0..R are the limits of the regions, t _r -t _r-1 is the number of frames in the region r,

- to ^≈ O,

- t _R = T, length of the subsequence.

8. Method (1) for protection against errors according to one of claims 5 and 7, in which to determine the borders t _r which minimize the distortion, the method (1) consists in using a Lagrangian.

9. Video transmitter (7) for the transmission of a stream of compressed video images over an unreliable link, all of the compressed video images corresponding to an input video sequence (2), this transmitter (7) comprising :

a video encoder (8) suitable for sharing the input video sequence (2) into output sub-sequences each consisting of a compressed video image I followed by a determined number of compressed video images P, each image I , P compressed video being coded in the form of data, at least one coder (10) channel adapted to add redundancy bits in the data coding each compressed video image to be transmitted,

- a processing unit (9) which controls each channel encoder (10), characterized in that each channel encoder (10) is further adapted to vary the number of redundancy bits to be transmitted, according to a distribution of the number of redundancy bits transmitted and associated with a compressed video image, and the processing unit (9) is adapted to calculate this distribution which is a function of the position of the compressed video image in a video sub-sequence so as to minimize an average distortion on the transmitted sequence, the average distortion being expressed by the mean square error MSE or the peak signal to noise ratio PSNR.

10. Method (12) for correcting errors implemented by a video receiver (11), these errors occurring during the transmission of a stream of compressed video images over an unreliable link, all of the video images compressed images originating from a video sequence (2) divided into sub-sequences consisting of a given number of compressed video images, the method (12) consisting in correcting the transmission errors by means of redundancy bits present in the stream and the number of which varies as a function of a distribution determined during transmission, this method (12) is characterized in that it comprises the step which consists in: - determining (12) by means of a decoder ( 16) channels the distribution of the redundancy bits defined in a header of the image or at the start of the transmission, this distribution of the number of redundancy bits transmitted and associated with a compressed video image being a function of the position of this image in a sub-sequ in this.

11. Video receiver (11) for receiving a stream of compressed video images after transmission over an unreliable link, all of the compressed video images being from a video sequence (2) divided into sub-sequences made up of a given number of compressed video images, this receiver (11) comprises:

- a channel decoder (16) for correcting transmission errors by means of redundancy bits present in the stream and is characterized in that the channel decoder (16) is suitable for determining the distribution of the redundancy bits, this distribution of the number redundancy bits transmitted and associated with a compressed video image being a function of the position of this image in a sub-sequence and being defined in a header of the image or at the start of the transmission.

12. Mobile terminal comprising a video transmitter (7) according to claim 9.

13. Mobile terminal comprising a video receiver (11) according to claim 11.

14. A computer program product loadable directly into the internal memory of a digital video transmitter (7), comprising portions of software code for the execution of the steps of a method (1) according to one of claims 1 to 8, when the program is executed on the digital video transmitter (7).

15. Support usable in a digital video transmitter (7) and on which is recorded a computer program product according to claim 14.

16. A computer program product loadable directly into the internal memory of a digital video receiver (11), comprising portions of software code for the execution of the steps of the method (12) according to claim 10, when said program is executed on the digital video receiver (11).

17. Support usable in a digital video receiver (11) and on which a computer program product according to claim 16 is recorded.