FR2770726A1 - METHOD FOR TRANSMITTING DIGITAL VIDEO IMAGES - Google Patents

Abstract

The invention relates to a method of transmitting video signals in which one merely sends information which indicates that an area of a current image 1 is unchanged with respect to a reference image 2 if the area of the image current 1 has little variation from the reference image 2.

Description

 A method of transmitting digitized video images.

 The invention relates to a method for transmitting digitized video images. This method applies more particularly to the transmission of video images in a fixed plane.

 Digital transmission of images requires the use of a high-speed transmission channel.

In order to be able to use lower bit rate channels, it is known to use image compression coding. I1 is known for example the M-JPEG codes and
MPEG.

The M-JPEG code has the particularity of sending in a compressed way the differences between a current image and the immediately preceding image. The code M
JPEG is particularly interesting when important areas of the image are unchanged. On the other hand, the efficiency of the M-JPEG code is quite limited when there is a lot of movement. In addition, the compression ratio of the M-JPEG code is also sensitive to the variations of illumination as well as to the parasites.

 The MPEG coding has the particularity of sending three types of images: real images, predicted images according to the previous image, and bidirectionally predicted images. The MPEG code is particularly interesting for images that vary constantly. A disadvantage comes from the large number of calculated image that requires a fast image succession.

 If one considers a video surveillance system that uses telephone lines, one skilled in the art perceives that the transmission constraints require transmitting few images per second by using an image compression code. The person skilled in the art realizes that MPEG coding can not be used because, in case of image motion, the large difference between two real images (a few seconds) does not make it possible to obtain predicted images close to reality. The skilled person also realizes that the M-JPEG coding does not sufficiently compress the data when there is a movement.

 The object of the invention is to provide an image compression code particularly well adapted for this kind of application. The surveillance cameras filming in a fixed plane, the invention will exploit the fixed plane of the image to reduce the information to be transmitted. The code used in the invention will suppress the transmission of information of little use by simply sending information that indicates that an area of the current image is unchanged from a reference image if the area of the current image has little variation from the reference image.

 The invention relates to a method for transmitting video signals of at least one current image area in which the video signals are coded and for transmitting the coded video signals in which a variation is measured in the current image area. relative to a reference image, and it is decided to transmit information indicating that the current image area corresponds to a reference image area if the variation is less than a threshold.

 It is possible to decide to send information representative of the differences between the current image and the reference image if these differences are greater than a threshold. In a variant, a degraded information representative of the differences between the current image and the reference image is sent, if these differences are greater than a defined variation threshold. The degradation of the coding of the information can be done according to the flow available on a transmission channel.

Preferentially, an image compression coding is used, and the information corresponding to the reference image is translated to correspond to an optimized use of the compression coding used for the transmission.
the invention will be better understood and other features and advantages will appear on reading the description which follows, the description referring to the appended drawings among which:
FIG. 1 represents the block diagram of the invention,
FIG. 2 schematically represents the image information to be transmitted,
FIG. 3 represents a quantization diagram, and
Figure 4 shows an entropy coding table.

 The system of FIG. 1 receives a digitized current image 1. The current image 1 corresponds to the last image filmed by a fixed-plane camera. A first reference image 2 has been previously stored in a memory, the first reference image being stored at a time when the fixed plane of the camera is considered immobile. A differentiation circuit 3 receives the current image 1 and the first reference image 2 in order to subtract these two images 1 and 2 to obtain a difference image. A transformation circuit 4 receives this difference image and performs a discrete spatial transformation of the reference image. Preferably, the transformation performed is a discrete cosine transformation (DCT).

 In order to limit the size of the transformation circuit 4, the difference image is divided into several areas. Classically, we will use areas that correspond to identical blocks of 8 * 8 points.

avec C Ç, = 1/ Wpour u,v = O Cli' Cu C,=lautrement
et dans laquelle Svu est le coefficient à la position (u, v) dans le bloc transformé et le coefficient Syx le point à la position (x, y) du bloc à transformer.In order to simplify the description, a block division of 4 * 4 points will be described below. The transformation circuit 4 provides at its output a plurality of coefficient matrices, each coefficient matrix being associated with a block. The TCD equation for each block is as follows:

with C, = 1 / W for u, v = O Cli 'Cu C, = lautrement
and in which Svu is the coefficient at the position (u, v) in the transformed block and the coefficient Syx the point at the position (x, y) of the block to be transformed.

The coefficient S0O corresponds to the average value of the block to be transformed.

 An extraction circuit 5 receives the matrices of coefficients from the transformation circuit 4 and calculates a spatial activity As for each block.

The spatial activity As of a block is obtained by performing the sum of the squares of all the coefficients of the matrix associated with the block except for the coefficient Soo which corresponds to the average luminous energy of the block.

The spatial activity As quantizes the difference between the current image 1 and the reference image 2 for a block. If the spatial activity is zero, this means that the block of the current image 1 is identical to the block of the corresponding reference image 2. The greater the spatial activity As, the more the differences between the images are important for the block. The spatial activity As is calculated using the following formula:

 A discrimination circuit 6 receives from the extraction circuit 5 the spatial activities As of all the blocks. Discrimination circuit 6 will compare all As space activities to a mobility threshold to provide a mobility indicator bit for each block.

The mobility threshold is set at a few percent of the maximum variation of spatial activity. The mobility bit is in a first state if the block is considered immobile, ie if the spatial activity As is less than the mobility threshold. The mobility bit is in a second state if the block is considered mobile, ie if the spatial activity As is greater than or equal to the mobility threshold.

 In one variant, the discrimination circuit 6 compares the spatial activities As with a plurality of thresholds and provides an activity vector indicating the distribution of the spatial activities As with respect to the different thresholds of the plurality of thresholds. By way of example, a second mobility threshold set at 50 W of the maximum variation of the spatial activity As is used and the discrimination circuit 6 provides the activity vector (n1, n2, n3), n1 indicating the number of blocks considered immobile, n2 indicating the number of blocks considered mobile whose spatial activity As is lower than the second mobility threshold, n3 indicating the number of blocks considered mobile whose spatial activity As is greater than the second mobility threshold.

 A switching circuit 7 receives, on a first input, the coefficients of the matrices associated with all the blocks coming from the transformation circuit 4, on a second input, a null matrix [V0] which has as many coefficients as a matrix associated with a block, and on a third input the mobility bits associated with each block. The switching circuit provides on its output the image information to be transmitted.

 FIG. 2 is a simplified representation of the information provided by the switching circuit 7.

The image represented in FIG. 2 comprises sixteen blocks 70, each of the blocks comprising a matrix of sixteen coefficients. Those skilled in the art will understand that this is an explanatory example and that the number of blocks and the number of coefficients are much greater in practice (for example, an image of 256 * 256 points is divided into 1024 blocks of 64 coefficients). This transformed image has two types of matrices: real matrices [R] and substituted matrices [V0]
The real matrices [R] consist of the svu coefficients obtained by the TCD of the corresponding block. The substituted matrices [V0] are in the example of the null matrices where all the coefficients of the matrix are zero. The use of a substituted matrix [V0] corresponds to the transmission of information indicating that the block of the current image 1 corresponds to a block of the reference image 2 having the same position if the spatial activity As the block is below the mobility threshold.

 A quantization circuit 8 receives the image information from the switching circuit 7 and the control signals from a compression control circuit 9. The quantization circuit 8 will then transform the values of the different coefficients into values. quantified. Quantification is performed as shown in FIG.

 In FIG. 3, a distribution curve 80 is shown which corresponds to the presence density of the different values of coefficients supplied by the switching circuit 7 for all the blocks of the image.

Although this distribution curve 80 is represented by means of a solid line, it is in fact a discrete curve having 2n values (where n is the number of bits used to code a coefficient).

 The compression control circuit 9 provides a desired number of quantized coefficients to the quantization circuit 8 in order to have only a limited number of values to encode. The quantization circuit 8 will then define the quantized coefficient values as a function of the desired number and the values of the coefficients. For example, the quantization circuit 8 defines a number equal to the desired number of ranges of values distributed linearly between the extreme values of the values of the coefficients. Quantified coefficient values are then associated with each interval. The quantized coefficient value may be equal to either the median of the interval or the average of the coefficient values in the range.

The quantization circuit 8 will replace the values of the coefficients of the image of FIG. 2 by the values of the quantized coefficients corresponding to the intervals at which the values of the coefficients are situated.

The image obtained constitutes a quantified image.

 In FIG. 3, the quantization curve 90 represents the presence density of the quantized coefficient values. In a simplified example, eleven quantized coefficient values a1 to a10 and 0 are used.

 The quantization circuit 8 provides a coding circuit 10 with a quantized image and also the presence densities of the quantized coefficient values.

 Conventionally, the number of quantized coefficients is defined from the transfer capacity of the transmission channel. In a variant, the number of quantized values and the distribution of these quantized values are defined by the compression control circuit 9 from the activity vector. Such a variant makes it possible to increase the precision of the blocks considered mobile when they are few and also makes it possible to increase the precision on the values having a high density of coefficients. This amounts to sending degraded information, representative of the differences between the current image and the base image, if these differences are greater than a defined threshold of variation. The threshold of variation can be variable according to the bandwidth of the channel (because for example of the unavailability of certain telephone lines).

The number of quantized coefficient values can be obtained simply by using a correspondence table having as inputs the activity vector and possibly the channel rate. By way of example, a correspondence table of the following type can be used:

<tb><SEP>SEP> channel rate: <SEP> 64 <SEP> kbs <SEP> 128 <SEP> kbs
<tb> n <SEP> 2 <SEP> n2 <SEP> + <SEP> n3 <SEP> 14 <SEP> 20
<tb> n2 <SEP>><SEP> n <SEP> + <SEP> n3 <SEP> 11 <SEP> 15
<tb> other <SEP> cases <SEP> 10 <SEP> 12
<Tb>
Of course, those skilled in the art will be able to use a larger table to more finely select the desired number of quantized coefficients in order to obtain the best possible image quality. For this purpose, it will be possible to use a plurality of mobility thresholds in order to obtain an activity vector that gives a more precise reflection of the spatial activity distribution As.

 The coding circuit 10 will define an entropy coding table 100 shown in FIG. 4. The entropy coding table 100 is a correspondence table. In a first column of table 100, the eleven quantized values will be written in decreasing order of density of presence. In a second column, we will enter the coding values that will correspond to the quantified values. The quantized value with the highest presence density is bit-coded, in the example using a "0". The quantized value with the second highest presence density is coded on two bits, in the example "10". The quantized value with the ith highest probability of presence is coded on i bits, in the example using i - 1 "1" followed by a "0". The quantized value with the lowest probability of presence is encoded on as many bits as the value with the second lowest probability of presence, in the example "1111111111". The last bit is always zero except for the coding of the quantized coefficient that has the lowest density of presence.

 After defining the entropy coding table 100, the coding circuit 10 will replace each of the quantized values by the coding value associated therewith in order to obtain an encoded image. The coding circuit provides at its output the coding table 100 and the coded picture.

 We can notice that the highest density of presence corresponds to the coefficient having the value 0.

This coefficient written for example on 12 bits in the image of FIG. 2 is here coded on a single bit "0" to be transmitted. The information from the substituted matrices [V0] therefore undergo the highest compression ratio.

 A multiplexing circuit 11 will then time-multiplex the data from the coding circuit 10 and the timing information from the compression control circuit 9. The multiplexed data is then sent by a transmission channel, for example a telephone line. .

The multiplexed data is received in a receiver by a demultiplexing circuit 12 which provides all the coded data to a decoder circuit 13. The decoder circuit will perform the decompression of the data using the coding table 100 and the information of the decoder. synchronization to obtain an uncompressed image. The decoder circuit 13 will carry out an inverse cosine transformation in order to obtain an intermediate image 14 by applying the formula:

 An addition circuit 15 adds the intermediate image 14 to a second reference image 16 identical to the first reference image 2 to obtain a received image 17. This second reference image 16 must be previously loaded into the receiver in a non-identical manner. degraded.

 Those skilled in the art will notice that this system corresponds to an M-JPEG system in which the extraction circuits 5, discrimination 6 and switch 7 have been added. In addition, the reference image does not correspond to the last image received. For this, first and second update circuits 18 and 19 have been added.

 The first update circuit 18 serves to update the first reference picture 2 and to send update information. This first update circuit 18 is triggered by an operator. The update can be carried out on the complete image, in this case the entire image is sent on the transmission channel possibly in compressed form without loss, the compression control circuit 9 indicating in the synchronization information it is a reference image. The demultiplexing circuit 12 orients the reference image on the second update circuit 19, this is indicated by the synchronization information. When the second updating circuit 19 receives an image, it records it as a second reference image 16 for reception. The update requires a much higher flow than to transmit a current image which paralyzes the system.

 An update variant consists of updating the block-by-block reference image. A current image is stored by the first update circuit 18. Each time a current image is sent in compressed form, one or more blocks are updated in the first reference image 2 and sent as data in a multiplexed manner. . The compression control circuit 9 serves to control the sending of the blocks according to the volume of data corresponding to the current image. The demultiplexing circuit 12 will then orient the data on the second update circuit 19 or on the decoder circuit 13 as a function of the synchronization information. The second update circuit 19 will replace blocks in the second reference picture 16 as blocks arrive. The block-by-block update takes longer but allows you to continue using the system during the update. Another possibility is to compare the stored image with the first reference image 2 in order to change in the first reference image 2 only blocks that are not identical to those of the stored image.

 The substituted matrices [Vo] are null matrices in the example because these null matrices have the particularity of corresponding to maximum compression with the compression used in the M-JPEG encoders while indicating that the block corresponding to the matrix is unchanged. It goes without saying that the person skilled in the art can adapt the invention to image transmission systems other than M-JPEG. In this case, the areas of the image considered immobile should be replaced by information indicating that the block corresponds to the reference image, said information being advantageously encoded in order to occupy as few data as possible.

Claims (8)

 1 - Method for transmitting video signals of at least one current image area (1) in which the video signals are coded and the coded video signals are transmitted, characterized in that  a variation in the current image area (1) is measured with respect to a reference image (2),  - It is decided to transmit information ([V0]) indicating that the current image area (1) corresponds to a reference image area (2) if the variation is less than a threshold.  2 - Process according to claim 1, characterized in that  - It is decided to send an information representative of the differences ([R]) between the current image (1) and the reference image (2) if these differences are greater than a threshold.  3 - Process according to one of claims 1 to 2, characterized in that  the coding is an image compression coding,  the information corresponding to the reference image corresponds to an optimized use of the compression coding used for the transmission.  4 - Process according to claim 3, characterized in that  processing and transmitting information corresponding to image blocks (70) of identical size, each block corresponding to a zone.  5 - Process according to one of claims 1 to 4, characterized in that  - We decide to send a degraded information, representative of the differences between the current image and the reference image, if these differences are greater than a defined threshold of variation.  6 - Process according to claim 5, characterized in that  the information coding degradation is modified as a function of the available bit rate on a transmission channel.  7 - Method according to one of claims 1 to 6, characterized in that  the variation is measured by performing a two-dimensional discrete transform (DCT) of image signals representative of the current image (1) and the sum of the squares of the coefficients of this current is represented as a signal representative of this variation (As); discrete transform except for the square of the coefficient representative of the average brightness (Soo).  8 - Process according to one of claims 3 to 7, characterized in that the compression coding is an entropy coding.