GB2195216A - Video transmission system - Google Patents

Abstract

At a television transmitter, a sequence of television fields is generated and every second or third field is extracted for transmission unaltered. A motion analysis is performed at the transmitter on each field not extracted, relative to the previous or the nearest extracted field. Supplementary information is also derived to enable each field not extracted to be predicted from the previous or nearest extracted field. Each field not extracted is low pass filtered and transmitted together with the supplementary information and the extracted fields. The extracted and low pass filtered fields are produced by an arrangement of a single field delay 13 supplying a strobed multiplexer 16 and a low pass filter 12 supplying the multiplexer 16 directly and through a series pair of field delays 14 and 15. The supplementary information is obtained from the high frequency content of the fields which is obtained by a subtractor 17 bridging the filter 12. The high density content is stored for three consecutive fields in field stores 18, 19 and 20 supplied directly, with one field delay 21, and with two field delays 21 and 22 respectively. Processing machines 26 and 27 produce two streams SI(1) and SI(2) of supplementary information for transmission; eg relating to shifts between corresponding elements in adjacent fields. <IMAGE>

Description

SPECIFICATION Video transmission system The invention relates to video transmission systems.

In an article entitled 'Reduced bandwidth requirement for compatible high definition television transmission', at pages 297 to 305 of Proceedings of the 38th Annual Broadcast Engineering Conference, National Association of Broadcasters, Washington, D.C. (1984), a video transmission system in which the higher spatial frequency content of a video signal is transmitted at a reduced frame rate is described by W. E. Glenn, K. G. Glenn, J. Marcinka, R. Dhein, and I. C. Abrahams. In a chapter entitled 'Semi-compatible High Definition Television Using Field Differential Signals', at pages 491 to 494 of 'Links for the Future', edited by P. DeWilde and C. A. May, published in 1984 by IEEE/Elsever Science Publishers B.V. (North Holland), B. G. Haskell describes a video transmission system in which a field differential signal is transmitted at a lower bandwidth than an original field.Although these known techniques perform adequately both in static pictures and on rapidly changing scenes, they give rise to perceptible distortion on picture material involving scenes with simple motion such as translation. The present invention aims to avoid such distortion.

In accordance with the present invention, not all fields are fully transmitted, and instead, only some of the original video fields are fully transmitted, and those which are not fully transmitted are reconstructed from information which describes the relationship of these fields to the fully transmitted fields, and low spatial frequency information from the fields which are not fully transmitted. By fully transmitted is meant that the full spectral content is transmitted.

In a first example of the invention, some of the original video fields are fully transmitted, the low frequency components of the fields not fully transmitted are transmitted, and higher frequency components for the fields not fully transmitted are derived by using information describing their relation to the fields which are fully transmitted. The fully transmitted fields and the low frequency components of the other fields can be transmitted in analogue form. The relationship of the fields which are not fully transmitted to those which are fully transmitted is principally a measure of the extent to which the scene in a field which is not fully transmitted has moved relative to the position of the same scene in a fully transmitted field.

A reconstruction error can be defined as the difference between the original video field not fully transmitted its reconstruction based on the nearest fully transmitted field.

The invention is based on the following considerations.

If a picture is stationary and not every field is transmitted, the missing fields can be derived by repetition of the transmitted fields.

During regular motion, such as uniform motion over a block, the viewer's eye tracks the motion in the scene and therefore it is necessary to reconstruct each field at full spatial resolution in the receiver. This can be accomplished by informing the receiver how the missing field is related to the information in the closest transmitted field. When there is a change of scene, information in adjacent fields cannot usually be related. In that case, the receiver should reconstruct a low spatial frequency version of the field not transmitted, since this will be acceptable because a viewer can tolerate a temporary loss of spatial resolution immediately before a change of scene and immediately after a change of scene.

Examples of the present invention are advantageous for transmitting a high definition television signal with a field rate in excess of current European standard, which is 50Hz.

The probable field rate for such examples lies in the range of 80-100Hz. The invention is most advantageous in this context if the original signal is produced by the scene being scanned sequentially (progressively).

The transmitting of only some video fields at full spatial frequency is referred to herein as temporal subsampling, and the ratio of fully transmitted to not fully transmitted fields is referred to as the field subsampling ratio. An advantageous field subsampling ratio is 1/2 or 1/3. Where the field subsampling ratio is 1/3, half of the fields not fully transmitted would be derived from a fully transmitted field immediately following the respective fields not fully transmitted, and the other half of the fields not fully transmitted would be derived from fully transmitted fields respectively immediately proceeding the fields not fully transmitted.

Sequential scanning is preferred in spite of its apparently high bandwidth because it allows each field to represent all the video information in the scene over the full spatial frequency spectrum. It is therefore possible for the information in one field to be accurately derived from the information in an adjacent field if there is a simple motion between them. Sequential scanning furthermore allows the vertical spatial bandwidth of fields not fully transmitted to be reduced to the same extend as the horizontal bandwidth. This allows the total bandwidth of the fields not fully transmitted to be proportional to the square of the bandwidth reduction in each spatial direction. For example, a reduction of 1/2 in each spatial direction would allow each field not fully transmitted to be transmitted in 1/4 of its original bandwidth.This would give rise to an overall bandwidth reduction of (1+1/4 + 1/4):3 or 1:2.

It can be shown that because the low fre quency content of fields not fully transmitted is transmitted separately in the first example, motion parameters should relate to the high frequency picture information. The supplemen tary information transmitted should include in addition to positional information some mea sure of the extent to which a field not fully transmitted is related to the fully transmitted field to which it is related. The derived high frequency information then has a reduced am plitude if there is not a high degree of inter field correlation. The fields extracted for transmission unaltered are referred to now as reference fields, and other fields as non-refer ence fields. In the present example, it is desir able that when there is uniform motion which the eye tracks, the resulting picture should be as good as stationary pictures in MUSE.

Hence in a MUSE-like system, motion vector estimation for motion following filtering may be used. Furthermore, motion compensation techniques may be used for bit rate reduction.

Motion vectors and their measurement and their use in bandwidth reduction are discussed in Preprint No. 128-49 of a paper entitled "HDTV Bandwidth Reduction by Adaptive Subsampling and Motion Compensation DATV Techniques" by Graham A. Thomas presented at the 128th SMPTE Technical Conference in New York, U.S.A., on 24th to 29th October 1986. The preprint is published by Society of Motion Picture and Television Engineers, Inc., of 595 W. Hartsdale Avenuer, White Plains, New York, N.Y. 10607, U.S.A. A more detailed treatment of motion vectors and their aplication to picture processing is disclosed in patent application no. 87 02582, serial no. 2 188 510. Copies of the specification and drawings of patent application no. 87 02582 are filed herewith.

Blind interpolation of the even fields of a sequence from its odd fields, and vice versa, can be effected by using phase correlation. In each part of a picture, a selection of motion vectors are identified and each decoded pixel is associated with the vector which best describes the motion between the two adjacent input video fields. Consequently, more than one type of motion can be dealt with. This segmentation is computationally intensive and is carried out in decoding. It may therefore be advantageous to restrict the classes of motion to two and transmit a low bandwidth classification signal. Alternatively, only one motion per block may be used and a compensation signal used. Furthermore, both possibilities may be used together.

An alternative technique for motion compensation is an adaptive search and approximation technique which does not explicitly estimate the exact displacement between fields. The search is used to derive an estimate of the relative shift between blocks in the non-reference fields and reference fields: this is accurate only to the nearest whole integer values of horizontal and vertical shifts, in pixels. The approximation technique involves computing the optimum values of a small set of weighting coefficients, centred on the shift values derived from the search technique, and minimises the reconstruction error in the non-reference field.

In the case of the adaptive search, one method comprises computing the total squared difference between the non-reference and shifted reference field blocks for each pair of shift coordinate values and selecting the value of shift coordinates which gives the minimum total squared difference. The total squared difference between the non-reference and shifted reference field blocks is the sum of all the individual values of the square of the difference between a pixel value in the nonreference field block and the value of the corresponding pixel in the shifted reference field block.

In the approximation technique, for all the pixel positions in a block of pixels in a nonreference field a single set of weighting coefficients is calculsted. This set of coefficients is used to estimate each pixel value in the said block by applying the weighting coefficients to corresponding pixels in the shifted reference field block. The set of weighting coefficients is calculated by determining the values of weighting coefficients which give the minimum totalqsquared difference between the estimated and true pixel values in the said block in the non-reference field. One method of calculating the required weighting coefficients is a regression method which finds the least total squared differences.This method has the advantage that the processing required at the receiver is reduced to that of filtering using received coefficients, there is not an abrupt transition between stationary and moving modes, and the approximation has minimum error even when the motion is complex or when there is a scene change. However, signalling overheads are higher than in the case of motion estimation when only the X and Y shift values have to be transmitted.

As an example, a system will now be described which is substantially compatible with the existing 625 line, 50Hz, 2:1 interlace standard. The field rate at the transmitter input is 75Hz with 625 lines per field and sequential scanning is used. The field subsampling ratio in this example is 1/3. Although the total number of lines per picture is the same as in the existing standard, the usable vertical spacial bandwidth is about 50% higher than at present. The fields which are fully transmitted are transmitted at an average rate of 25Hz, which is compatible with the existing standard if the order in which the lines of the fully transmitted fields are transmittd is 1,3,4,...

2,4,65..., i.e. the odd lines are transmitted before the even lines. This re-ordering of the lines does not represent a large overhead at the transmitter because retiming is in any case needed to transmit a single 75Hz field over 1/25 of a second. Provided the additional information i.e. low bandwidth extra fields, and supplementary information, is transmitted in a manner which causes it to be ignored by a receiver with conventional circuitry, the viewer should be presented with a picture which is comparable in quality with that produced by high quality film. This result is expected because both fields at the receiver corresponding to a single field fully transmitted from the transmitter are formed by information originated in the same field scan at the transmitter.If the increased vertical resolution causes unacceptable interfield flicker on a conventional viewing standard, the example would still nevertheless be usable for monitoring and continuity purposes as in a professional broadcasting environment.

Embodiments of the present invention may take the form of a digitally assisted television (D.A.T.V.) transmission system which relies on motion compensation techniques, temporal subsampling, and reconstruction in the receiver using side information (the digital assistance) to derive an approximation to the missing fields, together with an error signal to allow the correct field information to be reconstituted from the approximation.

A 625 line 25Hz sequential system would have a bandwidth of 5.5MHz, but as the permissible vertical resolution is about 7.4MHz, a horizontal bandwidth of the same magnitude is allowed. This is the bandwidth of the scanned signal. Allowing a bandwidth for the compensation signal of for example 40% of the full bandwidth brings the total bandwidth to approximately 10.5MHz. The difference between the 575 lines used for an active picture and the 625 lines is allowed for digital supplementary information.

The aspect ratio and transmission of chrominance information can be accommodated as thought desirable.

The invention will now be described, by way of example, with reference to the accompanying drawing the sole figure of which is a block diagram of part of a television transmitter embodying the invention.

The drawing shows schematically apparatus 10 for processing a digitised video signal in accordance with a method embodying the invention. The digitised video signal, consisting of a sequence of sequentially scanned fields so that the field and frame rates are identical, is supplied to an input 11 to the processing apparatus 10. The full video frequency spectrum input signal is transformed into an output signal consisting of a sequence of fields in which every third field has its full video spectrum of frequencies and all other fields have only their low video frequencies by an arrangement of a low pass filter 12 and three field delays 13, 14 and 15 supplying a multiplexer 16 from which the said output signal is taken.In each group of three fields in the input video signal the first is low pass filtered by the filter 12 and supplied sequentially through the two field delays 14 and 15 to one input of the multiplexer 16, the second field is passed through only the field delay 13 to a second input of the multiplexer 16, and the third field is low pass filtered by the filter 12 and thence passed directly to a third input of the multiplexer 16. Consequently all three fields of the group are applied synchronously to the multiplexer 16. The multiplexer 16 blanks out the two groups in which the first and second fields of the group which is not blanked out would appear as the respective third field.The multiplexer 16 may furthermore be adapted to retime and rearrange the data in the field coupled through the single field delay 13, or such retiming and rearrangement may be effected at a later stage in the transmitter. For example, where the input video signal at input 11 has a field rate of 75Hz and each field consists of 625 sequentially scanned lines, the field coupled through the delay 13 is expanded in duration to occupy the time of three consecutive fields, i.e. 40 milliseconds, and its lines are rearranged so that its odd lines occupy the first 20 milliseconds and its even lines occupy the second 20 milliseconds.

The output from the multiplexer 16 thus provides every third field extracted for transmission with its full spectrum of video frequencies, and the other fields low pass filtered for tranmission.

To obtain supplementary information to enable the low pass filtered fields to be restored to substantially their original full video spectra, the difference between each input video field and its low pass filtered version is produced by the operation of a subtractor 17 having one input connected directly to the input 11 and its other input connected directly to the output of the low pass filter 12. The difference output from the subtractor 17, which is the high frequency content of each video field, is supplied to three stores 18, 19 and 20.

The store 18 receives the difference output directly, the store 19 receives the difference output through one field delay 21, and the store 20 receives the difference output from a further field delay 22 which is in series with the field delay 21. The field delays 13, 14, 15, 21 and 22 can be implemented by shift registers with suitable shift pulse inputs 23.

The three individual fields in each group of three consecutive fields in the input video signal are applied synchronously to the three stores 18, 19 and 20, the first being applied to the store 20, the second to the store 19 and the third to the store 18. The stores 18, 19 and 20 are however only enabled at the occurrence of one in three of the input video signal fields so that, as in the operation of the multiplexer 16, the two groups of three fields in which the first and second fields of the stored group would have been the third field are blanked out. A strobe signal at F/3 where F is the field rate is applied to the enabling inputs 24 and 25 of the multiplexer 16 and stores 18, 19 and 20 respectively.

The second stored field, held in the store 18, contains the high frequencies of the field fully transmitted through the multiplexer 16 and the first and third stored fields, held in the stores 20 and 18 respectively, contain the high frequencies of the low pass filtered fields transmitted through the multiplexer 16. The contents of the stores 19 and 20 are supplied to a first processing machine 26 which generates and supplies first supplementary information Sl(1) for transmission, and the contents of the stores 19 and 18 are supplied to a second processing machine 27 which generates and supplies second supplementary information Sl(2) for transmission.The first supplementary information Sol(1) comprises shift values to the nearest pixel describing the relationship of blocks in the field in store 20 to corresponding shifted blocks in the field in store 19, and small sets of weighting coefficients, each set applying to a respective shifted block in the field stored in store 19. It should be noted that each block is a selected rectangular or diamond shaped region of the video field. The size of the blocks is fixed at a value which gives the best result for the normal expected maximum movement.The second supplementary information S1(2) comprises shift values to the nearest pixel describing the relationship of blocks in the field in store 18 to corresponding shifted blocks in the field in store 19, and small sets of weighting coefficients, each set applying to a respective shifted block in the field stored in store 19.

The shift values and the weighting coefficients are calculated as described hereinbefore.

Alternatively, the processing machines 26 and 27 may rely entirely on the production of motion vector information to provide the relationship between on the one hand the high frequencies of the fields not fully transmitted and, on the other hand, the high frequencies of the fields fully transmitted.

The two streams of supplementary information Sol(1) and S1(2) and the output from the multiplexer 16 are timed so that the correct supplementary information can be transmitted with the expanded fully transmitted field and the adjacent two low pass filtered fields. Such timing arrnngements will be well understood by those skilled in the art and need not be further described. The supplementary information and the multiplexer output are then used to modulate one or more carrier frequencies in any suitable manner for propagation. However, the manner in which the output signals from the multiplexer 16 and the supplementary information are combined for transmission may take any other of the many possible forms which will be apparent to those skilled in the art.

Claims (11)

1. A method of transmitting video signals, in which a sequence of video field signals is generated, a selection of the video field signals is transmitted with the full range of frequencies in a complete defined range, and signals in a restricted range of frequencies in the lower part of the complete defined range are transmitted which together with the transmitted selection allows the said sequence to be substantially reconstructed.

2. A method of transmitting video signals, in which a sequence of video fields is generated, only some of these fields are fully transmitted, and each of those which are not fully transmitted is represented by supplementary information which describes a predetermined relationship of the respective field to a selected nearest one of the fully transmitted fields, and low spatial frequency information from the said respective field which is not fully transmitted.

3. A method according to claim 2, wherein the predetermined relationship comprises a measure of the extent to which a scene depicted by a video field which is not fully transmitted has moved relative to the position of the same scene in the respective selected nearest one of the fully transmitted fields.

4. A method according to any preceding claim, wherein the field rate has in the range 60Hz to 100 Hertz.

5. A method according to any preceding claim, wherein each video field is formed by sequential scanning.

6. A method according to any preceding claim, wherein alternate fields are fully transmitted.

7. A method according to any one of claims 1 to 6, wherein every third field is fully transmitted.

8. A method according to claim 2, wherein the field rate is 75 Hertz, there are 625 lines per field, and the scanning is sequential, every third field is fully transmitted, and the fields which are fully transmitted are transmitted at 25 Hertz with the order of the lines thereof being such that the odd lines are transmitted before the even lines.

9. A method of transmitting video signals, in which a sequence of television fields is generated, every second or every third field is extracted for transmission unaltered as reference fields, motion analysis is performed on each field not a reference field relative to the previous or a selected nearest reference field, supplementary information is derived to enable each field not a reference field to be substantially reconstructed from the previous or selected nearest reference field, and the fields which are not reference fields are low pass filtered and transmitted together with the supplementary information.

10. A method according to any one of claims 2 to 9, wherein the supplementary information comprises coarse estimates of displacements between high frequency contents of fields not fully transmitted and high frequency contents of respective selected nearest fully transmitted fields, and the values of weighting coefficients to be applied to the said high frequency contents of the respective selected nearest fully transmitted fields.

11. A method of transmitting videosignals, substantially as described hereinbefore with reference to the accompanying drawing.