GB2205711A - Motion vector selection in television images - Google Patents
Motion vector selection in television images Download PDFInfo
- Publication number
- GB2205711A GB2205711A GB08728450A GB8728450A GB2205711A GB 2205711 A GB2205711 A GB 2205711A GB 08728450 A GB08728450 A GB 08728450A GB 8728450 A GB8728450 A GB 8728450A GB 2205711 A GB2205711 A GB 2205711A
- Authority
- GB
- United Kingdom
- Prior art keywords
- motion
- field
- vector
- samples
- fields
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/01—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/01—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level
- H04N7/0135—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level involving interpolation processes
- H04N7/014—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level involving interpolation processes involving the use of motion vectors
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/186—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a colour or a chrominance component
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
- H04N19/523—Motion estimation or motion compensation with sub-pixel accuracy
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
- H04N19/527—Global motion vector estimation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
- H04N19/533—Motion estimation using multistep search, e.g. 2D-log search or one-at-a-time search [OTS]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
- H04N19/577—Motion compensation with bidirectional frame interpolation, i.e. using B-pictures
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N9/00—Details of colour television systems
- H04N9/79—Processing of colour television signals in connection with recording
- H04N9/87—Regeneration of colour television signals
- H04N9/877—Regeneration of colour television signals by assembling picture element blocks in an intermediate memory
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N9/00—Details of colour television systems
- H04N9/79—Processing of colour television signals in connection with recording
- H04N9/87—Regeneration of colour television signals
- H04N9/89—Time-base error compensation
- H04N9/896—Time-base error compensation using a digital memory with independent write-in and read-out clock generators
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Television Systems (AREA)
Abstract
Apparatus for processing motion vectors associated with a digital television image comprises a vector filter (36 Fig. 2) and a vector calculator (37) for deriving a plurality of motion vectors respectively representing the motion of a pixel in the image, a circuit 101, 104 for testing the accuracy of each motion vector by differencing pixels shifted in dependence on the motion vector from two successive fields of the image to a common position in a field interpolated between the two fields, and a vector selector 103 for selecting a motion vector from the plurality in dependence on the testing. <IMAGE>
Description
2 2 0 5 7 11 MOTION VECTOR SELECTION IN TELEVISION IMAGES This invention
relates to motion vector selection in television images. Such motion vector selection is particularly, but not exclusively, used in television standards converters and in slow motion processors.
International television programme exchange necessitates standards converters due to the different television standards used in different countries, for example, the 625-line 50-fields per second (625/50) PAL system used in the UK, and the 525-line 60-fields per second (525/60) NTSC system used in the USA.
Many different standards converters have been previously proposed. One of the best known is the ACE (Aavanced Conversion Equipment) developed by the British Broadcasting Corporation.
Basically ACE operates on an input digital television signal line-by line to derive interpolated samples required to form an output digital television signal. Interpolation is done not only spatially using four successive horizontal scan lines of the input television signal, but also temporally using four successive fields -of the input television signal. Thus, each line of the output television signal is derived by multiplying respective samples from sixteen lines of the input television signal by respective weighting coefficients.
Further details of ACE will be found in UK patent specification
GB-A-2 059 712 and in 'Four-field digital standards converter for the eighties' by R N Robinson and G J Cooper at Pages 11 to 13 of Television' (the journal of the Royal Television Society) for January/February 1982.
Although ACE gives good results, there is the problem that the equipment is very bulky. To overcome this problem, we have previously proposed a television standards converter comprising three field stores and four 4-line stores for receiving an input digital television signal of one standard and deriving therefrom arrays of sixteen lines, each array consisting of four successive lines from each of four successive fields of the input television signal. A weighting coefficient store stores sets of siAteer! weighting coefficients, respective sets corresponding to positions both spatia and temporal of respective lines of an output digital television 2 signal of a different standard, relative to the sixteen lines of the input television signal. Two interpolation filters then derive 'line by-line the output television signal by multiplying corresponding sample values Irom each of the sixteen lines of the input television signal by a respective weighting coefficient in a set of weighting coefficients and sum the resulting products to form an interpolated sample value, and four output field stores - receive and store the derived lines of the output television signal. To store the additional lines which are derived when the output television signal has more lines than the input television signal, a 45-line store is interposed between one of the interpolation filters and the output field stores. Further details will be found in our UK patent specification GB-A-2 140 644.
The performance of such standards converters which employ vertical/temporal interpolation techniques represents a compromise between generating blurred pictures while maintaining good motion portrayal and maintaining vertical resolution but at the expense of judder'. The former is a result of post filtering in order to prevent disturbing alias effects; the latter is a result of the intrusion of the adjacent two-dimensional repeat sample structures.
We have therefore proposed that motion vector estimation should be incorporated in television standards converters and in slow motion processors. The problem with the majority of existing motion vector, estimation methods is that their use is biased towards video conference type applications where generally the subject matter is either a single person's head and shoulders or a small group of people seated around a table. With television images of this type the motion is relatively simple in comparison with broadcast television images where for example at a horse race meeting the camera could be following the leaders in a race. In this situation the motion would be complex, for example, because the camera would be panning. Thus, the background may well be moving at speeds greater than eight-pixelp, per field, while in the foreground there would be at least:qne horse galloping. This means that the motion vector estimation pethod must.
try to track the horses legs, which rray well be moving in.-.differentl directions to that of the already moving background, and_the method therefore involves problems of motion vector selection.
3 According to the present invention there is provided a method of processing motion vectors associated with a digital teievision image, the method comprising the steps of:
deriving a plurality of motion vectors respectively representing the motion of a pixel in said image; testing the accuracy of each said motion vector by differencing pixels shifted in dependence on said motion vector from two successive fields of said. image to a common position in a field interpolated between said-two fields; and selecting a motion vector from said plurality i n dependence on said testing.
According to the present invention there is also provided apparatus for processing motion vectors associated with a digital television image, the apparatus comprising:
means for deriving a plurality of motion vectors respectively representing the motion of a pixel in said image; means for testing the accuracy of each said motion vector by differencing pixels shifted in dependence on said motion vector from two successive fields of said image to a common position in a field interpolated between said two fields; and means for selecting a motion vector from said plurality in dependence on said testing.
The invention will now be described by way of example with reference to the accompanying drawings, throughout which like parts are referred to by like references, and in which:
Figure 1 shows in very simplified block diagrammatic form a television standards converter; Figure 2 shows in block diagrammatic form a first television standards converter; Figure 3 shows in block diagrammatic form a second television standards converter; Figure 4 shows part of the standards converter of Figure 3 in more detailed block diagrammatic form; Figure 5 shows time charts for explaining the operation of Figure 4; Figure 6 shows part o- the standards converter of Figure 2 in more detailed block diagrammatic form; 4 Figure 7 shows time charts for explaining the operation of Figure 6; Figure 8 shows part of the standards converter of.1--igure 2 in more detailed olock diagrammatic form; Figure 9 shows time charts for explaining the operation of Figure 8; Figure 10 shows part of the standards converter of Figure 2 in more detailed block diagrammatic form; Figure 11 shows diagrammatically the shifting of field store pictures under control of motion vectors; - Figure 12 shows part of the standards converter of Figure 2 in more detailed block diagrammatic form; and Figure 13 shows diagrammatically the problem of selecting motion vectors associated with a plain area of an image.
In order more readily to understand the motion vector selection which forms the subject of the present invention, the form and operation of two standards converters and a slow motion processor which use such motion vector selection will first be described. The standards converters to be described maintain vertical resolution and remove the 'judderl by compensating for motion between fields. In effect the motion between consecutive fields is analyzed. These fields can then be 'aligned' pixel bv pixel such that they represent static pictures upon which conversion can then take place. As a result, vertical resolution can be maintained.
The standards converters to be described can be divided into two parts. The first part is analogous to a known standards converter performing ver ti cal/ temporal interpolation to convert between 525,160 and 625/50 television standards. Alone, this would generate an output in which vertical resolution would be maintained but with the added effect of judder. To remove this judder four fields of the input digital television signal which are used in the conversion process are aligned under the control of motion vectors generated from a motion analyzer which forms the second part of the standards converter.
This is shown in very simplified diagrammatic block form in Figure 1. The video portion of an input digital television signal of one standard, which may for exampe have been derived by sampling an analogue television signal at 13.5 MHz, is supplied to an interpolator 1 from which the video portion of the required output television signal of a different standard is derived. A motion analyzer 2 receives the luminance video and derives motion vectors which provide data representing the motion between successive fields of the input television signal to control the operation of the interpolator 1. The interpolator 1 operates in a generally similar manner to the corresponding portion of a known standards converter, for example as referred to above. It also, however, contains the means to align the four fields used in the interpolation, under the control of the motion vectors.
The repositioning of the four fields is performed in two stages.
The first stage involves varying the address of a variable delay element associated with each field to reposition the picture to the nearest line or sample. The second stage uses interpolation techniques both vertically and horizontally to reposition to within +1/16 line or +1/8 of a sample. Even with no movement, both the above techniques are used to enable conversion of line standards.
The vertical interpolator has four taps per field allowing effectively an 8-tap vertical filter to be applied to the static pictures. An 8-tap interpolator allows good vertical resolution to be maintained with minimal distortion. The effect of distortion in the horizontal interpolator is less of a problem, so a 2-tap horizontal filter is used, although a 4-tap horizontal filter, for example, may be used.
The temporal interpolator is used in normal operation to enable interpolation of perspective changes or when no sensible motion vector can be detected, in which case the interpolator 1 must revert to normal standards conversion operation where no picture re-positioning occurs.
When converting from a high field rate to a lower rate, the incoming fields are interpolated such that an interpolated field can occasionally be dropped without any movement deterioration. All the interpolation is done at the input field rate and passed to a time base corrector which then spreads the fields generated over the required time period for the output standard.
The above operation is necessary when converting from 525/60 to 625/50. It is also evident however that 625 lines must be generated 6 where only 525 lines exist in the input signal.
To overcome the line number conversion problem a second time base corrector is used at the input to produce a signal having 585 lines at %,he 60 Hz rate. A 585-line format can contain all the active picture information in the 625-line format. Following this first time base corrector there are occasional lines which have no video information.
The interpolator stores are frozen during this time, so that an additional interpolated line can be generated from the same lines used to generate the previous output line. This process allows 625 lines to be interpolated from the original 525.
The reason for selecting the 585/60 format will now be explained in more detail. A 625-line picture contains 288 active lines in each field, and 720 samples in each horizontal line at the sampling rate of
13.5 MHz. The circuits, to be described below, of the television standards converters of Figure 2 and 3 use techniques which allow the picture to be shifted horizontally by plus or minus twenty-four samples. This requires a minimum horizontal blanking of forty-eight samples. The total number of sample positions required in a field is therefore:
(720 + 48) x 288 = 221184.
There are clearly considerable advantages in using a 13.5 MHz clock throughout the system, in which case the number of clock cycles within a 60 Hz period (more exactly a 59.94 Hz period) is:
225225.
If 576 lines of data are required in one frame, the number of horizontal samples would be 782.03125. Although this number is sufficient to contain the required (720 + 48) samples, the fractional sample would mean that the structure was non-orthogonal on a line to line basis. This would cause significant design difficulties in the 3 5 rest of the standards converter, so the number of lines required was gradually increased, from 576, until a whole number of sampies, in fact 770, existed in each line.
7 The only format that achieves the orthogonal structure is the 585/60 format, which in addition gives a useful vertica-11 bianking of four lines in the first field, five lines in trie second field and fifty samples of horizontal blanking.
In the 625/50 to 625/50 slow motion mode referred to below there is no requirement to store the active video of the 625 format within a Hz period, so the interpolation and other processing is done in the normal 625/50 format.
When converting from a low field rate to a higher rate the input time base corrector is required to produce a video stream at the output rate. This is done by occasionally repeating an input field.
When the repeated field occurs, all the interpolator stores must be frozen so that the interpolation is applied to the same input fields used to create the previous output field.
If this technique were not used, two sets of interpolator and movement detector would be required to make up the missing field.
The above operation is performed when converting from 625/50 to 525/60. To allow 625 lines to exist during a 60-fields per second period again requires the 585/60 intermediate format to be adopted.
During this process some of the interpolated lines will not be required, as only 525 have to be produced from the origina! 625. A time base converter is therefore required on the output to produce the final 525/60 format.
The amount of interpolation required is determined by comparing input and output synchronization pulse phases.
As mentioned above, motion analysis is performed on the luminance of the input video. The method employed involves a number of stages to arrive at a single motion vector for each pixel. Movement can be detected in the range +24 pixels horizontally and +8 (field rate) vertically.
In a first stage, motion in the picture at points on the screen spaced sixteen samples horizontally and eight lines vertically is determined using a block matching technique. The original motion vectors in a field are calculated every sixteenth sample and every eighth line. Each one of these points is at the centre of a search block. Conceptually each block is scanned +24 samples horizontally, and +8 and -8 samples vertically over the next field each time
8 generating the summation of the differences between the two fields over the area of the search block. The minimum overall difference then indicates in which direction the object at that point has movec.
In practice, the above technique is applied in separate steps which greatly reduces the amount and complexity of hardware requirej:
Step 1 Test for minimum difference in just three positions, centre position, sixteen samples to the left, sixteen samples to the right.
Step 2. Starting from point indicated above.
Test for minimum difference in nine positions symmetrically distributed about the above starting point in steps of eight samples or lines.
Step 3. Starting from point indicated above.
Test for minimum difference in nine positions symmetrically distributed about the above starting point in steps of four samples or lines.
Step 4. Starting from point indicated above.
Test for minimum difference in nine positions symmetrically distributed about the above starting point in steps of two samples or lines.
Step 5. Starting from point indicated above.
Test for minimum difference in nine positions symmetrically distributed about the above starting point in steps of one sample or line.
Step 6.
After step 5, the motion of the object has been detected to the nearest pixel. A more accurate vector value can be achieved by adding a sixth step in which the difference produced at the final position indicated by step 5 is compared with the two differences above and below to adjust the vertical vector value and with the two differences to the left and right to adjust the horizontal vector value.
The above technique relies on achieving correlation between the reference search block and a similar block of video data on the following field (the search positions). In step 5 it is possible the true movement was a half pixel more or less than detected, but it is necessary for the best correlation to occur at this point, even although exact correlation cannot be achieved. To ensure this occurs, 9 the picture can be filtered both vertically and horizontally by a gaussian filter which has +6 dB attenuation at 1,,2 Nyquist frequency.
Similarly, for step 4, the picture can be filtered with a 6 dB attenuation at 1/4 Nyquist frequency, which allows a one pixel error in detection.
Step 3 uses a picture filtered with a 6 dB attenuation at 1/8 Nyquist frequency allowing a two pixel error.
Step 2 uses a picture frequency with a 6 dB attenuation at 1/16 Nyquist frequency allowing a four pixel error.
Finally, step 1 uses a picture filtered with 6 dB attenuation at 1/32 Nyquist frequency allowing an eight pixel error. In addition, because the pictures are so heavily filtered during steps 1, 2, 3 and 4, the samples can be reduced, for example halved in number, which still further greatly reduces the number of calculations and amount of hardware required.
The effective search block size is sixteen lines high and forty- eight samples long. A large search block is necessary to detect accurately the movement of large plain areas. The oentral part of plain areas are unimportant as the values of the pixels at these points do not change from one field to the next, but the edges of such objects are obviously important. If the detection of motion is limited to +24 samples horizontally and:L8 lines vertically then a block of the above size would be the minimum size to ensure accurate motion detection.
In the standards converters, depending upon the conversion modes, the luminance video entering the motion analyzer 2 is in various forms of 585-lines/60-fields per second. This might comprise repeated lines for 525 input or repeated fields for 625 input. In addition, the input contains both field polarities. The first process is to ensure a continuity of data and single field polarity for the motion estimation processing. This is done by interpolation on the input data by a vector interface to maintain continuity, and filtration horizontally to aid subsequent motion detection/correlation.
Separate outputs from this circuit are passed to motion estimation vector filters and motion detection field stores," vector selectors. The output of the vector interface is, as described above, spatially continuous, single field polarity data. The output to the field stores/vector selectors depends upon the input and output modes.
In some modes it is continuous, and in others it contains repeated lines/fields. Tne vector filters and vector calculators perform the steps outlined above.
The processing of the various steps is performed by vector calculators and a vector processor. The vector calculators perform steps 1 to 5 and the vector processor performs step 6. In addition, the vector processor performs the second stage in the motion estimation, as follows:
For each 8 x 16 block a choice is made of four from seven motion vectors, the seven motion vectors being the one for that particular block and the six for the six nearest blocks respectively.
In addition, the vector processor also determines the four most common motion vectors throughout the whole input field, these being called modal motion vectors. The primary use of the modal motion vectors is in the border areas close to the edge of a field where it is not possible actually to calculate any local motion vectors. Also, if any one or more of the local motion vectors are equal, then these are substituted for by the modal motion vectors.
In the next stage of motion detection, for each pixel, the four motion vectors are tested by producing the difference between the extrapolated positions on field 0 to field 1. During standards conversion a field needs to be interpolated between two fields; say between field 0 and field 1. So the motion vectors generated between these two fields are considered to be most representative of the motion. Four motion vectors are used from these two fLelds. To decide which is the correct motion vector a pixel from field 0 is compared with a pixel from field 1 using the motion vector to decide where the pixel to be generated had come from on field 0 and where it has gone to by field 1. Mathematically, if the position x, y, z must be generated, where; x = horizonal position, y = vertical position, z = temporal position between field 0 and field 1, the pixels used for comparison are as shown below. Field 0 is assumed to be at z=O and field 1 at z=1.
Pixel from field 0
X0 = x - (Vhz) y 0 = Y - (Vvz) Pixel from field 1 x x + (1-7)Vh y y + (1-Z)VV Vh = horizontal component of vector Vv vertical component of vector For each motion vector a modulus of the difference between the pixels indicated in field 0 and field 1 is found. The minimum difference is assumed, as a first estimate, to indicate the correct motion vector. If a number of motion vectors produce a very similar difference then these motion vectors are tested again using a comparison between fields -1 and 0.
Pixels from field -1
1 x- = x - (1'+,z)Vh Y- 1 = y - (1+,)Vv The minimum modulus of difference of tne remaining motion vectors produced by this second test is then considered to represent most accurately the motion vector.
If a number of motion vectors again have similar differences then an option exists to assume no movement. If only the horizontal component varied and the vertical comporient did not, then only the horizontal component would be set to zero and the vertical component would be maintained at the detected value. If only the vertical component varied, then the horizontal component would be maintained and only the vertical component set to zero., If the pixel difference chosen is too large then an option exists to set the whole motion vector to zero in both directions.
A final stage is applied once every pixel has been assigned a motion vector. Here the motion of each pixel is tracked from one field to the next and a recursive filter applied to the vector value.
12 This removes the effects of noise and small movement estimation errors ana also smooths the trajectory of the motion vectors.
There are two possible ways of tracking the motion of a pixel.
In the first, the motion vector for a pixel -in field t is used to point to a pixel in field (t+1). The motion vector determined for this pixel in field (t+1) is then recursively filtered to form the final motion vector for the pixel in field (t+1).
In the second, the motion vector for a given pixel in field t is used to point to a pixel in field (t-1). The motion vector from this pixel is then recursively filtered with the motion vector for the given pixel to form the final motion vector for this given pixel in field t.
In either case the final output is a motion vector for each pixel which is passed from the motion analyzer 2 to the interpolator I to be employed in aligning the four fields used in the standards conversion process.
The first standards converter for converting an input digital 625-line 50-fields per second television signal to an output digital
525-line 60-fields per second television signal is shown in detailed block form in Figure 2.
The incoming video at 50-fields per second and a sample rate of
13.5 MHz, that is CCIR 601 data, is supplied to a demultiplexer 31 which separates it into luminance components Y, synchronizing signals SYNC and chrominance components UV. The luminance components Y are supplied to a 4-field luminance time-base corrector (TBC) 11Y and.the chrominance components UV are supplied to a 4-field chrominance TBC
11C. The synchronizing signals SYNC are supplied to a control 32 together with an input field polarity signal from an external input, and an output field synchronizing reference signal from another external input. The TBCs 11Y and 11C occasionally repeat fields, so that the output is at 60-fields per second. The control to the TBCs
1Y and 11C which causes the repetition of a field is derived from the input field synchronization pulses, and the required output field synchronization pulses. The comparison of the synchronization pulses also provides a temporal offset figure which indicates the amount of temporal interpolation required at the outputs of the TBCs 11Y and 11C such that smooth motion at 60-fields per second would be observed.
13 When converting from 50-fields per second to 60-fields in this way a line conversior of 625 to 525 is neuessary. T t is therefore necessary to maintain the original 625 lines of information at a 60 fields per second rate so that they are all ava e to form the interpolated lines.
The standards converter uses an intermediate standard which is capable of containing all the active vertical informatiop of the 50 fields per second signal at the 60-fields per second rate. The intermediate standard also contains all the active line information arranged orthogonally line by line still using the original 13.5 MHz sample rate.
The intermediate standard used, and which is as explained above capable of meeting all these requirements, is a 585-line format at 60 fields per second. When sampled at 13.5 MHz each line of this format has exactly 770 samples. It is clear therefore that 585 lines is sufficient to contain the 576 active lines of the 625-line format at a 60-fields per second rate. As the active line width is only
720 samples there is still fifty samples of horizontal blanking.
The luminance data (D) from the luminance TBC 11Y is supplied by way of a processing compensating delay 17Y to a luminance temporal shift register 16Y comprising four field stores (FS) 12Y, 13Y, 14Y and
15Y. The luminance TBC 11Y also supplies a temporal freeze signal (F) by way of the delay 17Y to the shift register 16Y. The chrominance TBC 11C supplies the chrominance data (D) by way of a processing compensating delay 17C to a chrominance temporal shift register 16C which comprises four field stores 12C, 13C, 14C and 15C. The chrominance TBC 11C also supplies a temporal freeze signal by way of the delay 17C to the shift register 16C. Associated with the shift register 16Y is a luminance interpolator 1Y which receives inputs from each of the f ield stores 12Y, 13Y, 14Y and 15Y, and derives the 585 line format. The output of the luminance interpolator 1Y is supplied to a 2-field luminance TBC 18Y. Associated with the shift register
16C is a chrominance interpolator 1C which receives inputs from each of the field stores 12C, 13C, 14C and 15C, and also derives the 585 line format. The output of the chrominance interpolator 1C is supplied to a 2-field chrominance TBC 18C. When the outputs of the
TBCs 11Y and 11C are frozen, during a repeat field, the shift
14 registers 16Y and 16C are also frozen, so that four distinct consecutive fields of the input always exist in the shift registers
16Y and 16C. Thus the shift registers 16Y and 16C are used to provide the temporal taps for the interpolators 1Y and 1C. Each temporal tap produces four line taps at a position depending on the
motion vectors, so that a two-dimensional filter can be used to provide the necessary interpolation. The interpolated picture will contain 576 active lines, so that a correct picture will be obtained when every sixth line in one field is dropped. The 484 lines left produce the active picture portion of the 525-line forma. To enable lines to be dropped in this way, the outputs from the interpolators 1Y and 1C are supplied to the 2-field TBC 18. The TBCs 18Y and 18C write in all 576/2 lines, but on.ly read out the required 484/2 lines to provide the required output television signal. The outputs of the luminance TBC 18Y and of the chrominance TBC 18C are supplied to a multiplexer 34 which multiplexes the luminance components Y and the chrominance components UV to provide output CCIR 601 data in the form of a digital 525-line 60-fields per second television signal.
The control 32 supplies control signals (C) to the luminance TBC 11Y and to the chrominance TBC 11C. The control 32 also supplies control signals to the luminance TBC 18Y and the chrominance TBC 18C.
It also supplies interpolation control signals (IC) to the luminance interpolator 1L and the chrominance interpolator 1C.
The luminance data only, as supplied by the luminance TBC llY, is also supplied to the motion analyzer 2 shown in the upper part of Figure 2, so that motion vectors can be generated. In fact a frame delay is necessary between the TBCs 11Y and 11C and the shift registers 16Y and 16C to allow for the time taken to process the motion vectors. The freezing of the shift registers 16Y and 16C must therefore also be delayed by one frame, and these delays are provided by the delays 17Y and 17C.
The motion analyzer 2 comprises a vector interface 35 to which the luminance data from the luminance TBC 11Y is supplied, together with the interpolation control signal from the control 32. The vector interface 35 supplies data interpolated to 625 lines to a vector filter 36 and a vector calculator 37 which together perform the motion estimation described above. The output-of the vector calculator 37 is supplied to a modal motion vector processor 38 and also to a sub-pixel motion estimator. The motion vector processor 38 supplies four outputs and the sub-pixel motion estimator one output to a motion vector reducer 40 which supplies four outputs to a vector selector 41, with which the present invention is particularly concerned.
The vector interface 35 also supplies the data interpolated to even fields to a processing compensating delay 42 to which it also supplies the received interpolation control signal, and also a temporal freeze signal (F) generated at the vector interface 35. The data from the delay 42 is supplied to a temporal shift register 43 which comprises three field stores 44, 45 and 46 which supply respective data outputs to the vector selector 41. The delay 42 supplies the interpolation control signal to the vector selector 41 which supplies the selected motion vector to a recursive motion vector filter 47, the output of which is the motion vector data which is supplied to the luminance interpolator 1Y and to the chrominance interpolator 1C.
The way in which the motion analyzer 2 derives the motion vector data has been described in detail above, and will be further described below, but the operation of the elements 35 to 43 and 47 will now be briefly described.
The vector interface 35 receives the luminance data from the luminance TBC 11Y, and the interpolation control signals from the control 32. It supplies 625-line data, normally contained within the 585/60 format, to the vector filter 36. It also supplies data to the delay 42. These data must contain a picture which is the s ame line standard as the required output, again normally contained within the 585/60 format. Each field of the interpolated data is also made to appear even.
The vector filter 36 produces the filtered picture data required for steps 1 to 5 above of the motion detection.. The filtered picture data are supplied in sample reduced form to the vector calculator 37.
The vector calculator 37 operates on the filtered and sample- reduced data from the vector filter 36 using an algorithm described in the terms of the steps 1 to 5 above of the motion detection. The process is essentially a two-dimensional binary search for motion down to pixel/line resolution. For each field, 1200 motion vectors are
16 generated and are supplied to both the modal motion vector processor 38 and to the sub-pixel motion estimator 39. It also supplies surrounding weighted absolute difference (WAD) va.-4,.ues as calculdted by step 5 above to the sub-pixel motion estimator 39. For details of WAD calculations, see 'Advances in Picture Coding', Musmann et al, Proceedings of the IEEE, April 1985. The specific WAD value which is the minimum in step 5 above of the motion detection provides a figure of merit (FOM).
The vector processor 38 calculates the four most common motion vectors that are detected in each field, and supplies them to the vector reducer 40.
The sub-pixel motion estimator 39 receives the motion vectors from the vector calculator 37 together with the surrounding WAD values. From these it estimates sub-pixel movement to be appended to the motion vector values. With each motion vector its corresponding final WAD value is also supplied to the vector reducer 40.
The vector reducer 40 receives the motion vectors from the vector processor 38 and from the sub-pixel motion estimator 39. For each motion vector from the sub-pixel motion estimator 39, the six motion vectors closest to it are grouped together. For each motion vector there are then eleven choices. The reduction process selects four motion vectors from the eleven for supply to the vector selector 41.
The vector reducer 40 supplies the vector selector 41 with four representative motion vectors for each sixteen pixel by eight line block of the picture. By comparing pixels over three fields, the vector selector 41 selects the single best motion vector for each pixel in the picture, as described in more detail below. The motion vector selected is supplied to the motion vector filter 47.
The delay 42 delays the data by one frame less twenty-one lines to compensate for other delays in the system.
The temporal shift register 43 holds and supplies the three fields of data used by the vector selector 41.
The motion vector filter 47 tracks a motion vector from one field to another so applying some filtering to the motion vectors by 31-5 combining motion vectors in different fields, so reducing motion detection errors. The output of the motion vector filter 47 is supplied to the luminance and chrominance interpolators 1Y and 1C to 17 control the alignment of the field data.
Exactly the same hardware can be used as a slow motion processor with good motion portrayal for either a 625/50 or a 525/60 television signal. It is not however necessary to use the vertical interpolator to provide the line number conversion. In all cases the control 32 determines what action is required by recognizing the input/output standard from the input and output field synchronization. pulses. In slow motion the input field polarity is used.
Whereas in 50-fields per second to 60-fields per, second conversion a field was occasionally repeated, in slow motion the field is repeated the same number of times as the input field is repeated.
As repeated fields are not written into the shift registers 16Y and
16C, the shift registers 16Y and 16C again _contain distinct consecutive fields. Indeed if a video tape recorder reproduces without any interpolation of its own, the original interlace structure is maintained allowing full resolution pictures to be produced. The temporal offset required is calculated by comparing the actual field rate pulses, be they 50-fields per second or 60-fields per second, with the rate at which a new field is received. To determine the temporal offset in this way, the system needs a signal to be available which indicates the true field polarity of the field being repeatedly replayed. The vertical interpolator will always produce the field polarity required at the output.
Conceptually the TBCs 11Y and 11C are not really required for slow motion operation, but their presence does provide a frame synchronization facility and also simplifies the'sYstem configuration.
The second standards converter for converting an input digital 525-line 60-fields per second television signal to an output digital
625-line 50-fields per second television signal is shown in detailed block form in Figure 3.
In this case, interpolation requires that all the input data is available in a consecutive form. In this case it would not therefore be possible to convert to 50-fields per second before the interpolators 1Y and 1C. The input data however contains only 484 active lines and the interpolators 1Y and 1C must produce 576. The 2 -he fleld TBCs 18Y and 18C are therefore positioned at the front of 11 standards converter to provide the necessary time slots for 484-line 18 to 576-line conversion.
The original continuous line structure is written into the TBICS 18Y and 18C but is read out in the 585-line standard with approximately every sixth line being blank. The interpolators 1Y and 1C are then used to produce a continuous picture at the output line rate by freezing its line stores during the blank input line, and producing the required additional line at the output, so ensuring that a spatially correct picture is produced. The required temporal offset is detected and applied as in the first standards converter, although the interpolation is applied such that a field can occasionally be dropped leaving the motion smooth. The field is dropped such that 60 fields per second to 50-fields per second conversion is achieved. The dropping of a field is achieved by using the 4-field TBCs 11Y and 11C at the output.
is Thus the second standards converter differs from the first standards converter shown in Figure 2 in only minor respects. In particular, the luminance TBCs 11Y and 18Y are interchanged, and the chrominance TBCs 11C and 18C are also interchanged. Also, no temporal freeze signals are required.
In both cases the control 32 has various functions as follows; controlling the reading and writing of the TBCs 11Y, 11C, 18Y and 18C; generating a temporal offset number, and in the case of the first standards converter the temporal freeze signal, and generating a vertical offset number together with vertica'L interpolation control signals. These functions will now be described in more detail.
Firstly, the 2-field luminance and chrominance TBCs 18Y and 18c always switch between field stores at the end of every 60 Hz field.
However, the operation of the 4-field luminance and chrominance TBCs
11Y and 11C depend on the mode of operation, and their control is also associated with the generation of the temporal offset signal. In fact, the control of the luminance and chrominance TBCs 11Y and 11C is determined from the input and output field synchronizing signals.
The derivation of the temporal offset signal in the case of 525/60 to 625/50 operation will now be described with reference to Figures 4 and 5.
In Figure 4, the control 32 is shown as including a line counter 61, and first and second latches 62 and 63. A line clock signal is 19 supplied to a clock terminal of the line counter 61, while the input field synchronizing signal is supplied to a eset terminal of the line counter 61 and to a clock terminal of the second latch 62. The output field synchronization signal is supplied to a clock terminal of the first latch 62. The output of the line counter 61 is supplied to the input of the f irst latch 62, the output of which is supplied to the input of the second latch 63, the output of which is he temporal offset signal supplied to the luminance and chrominance shift registers 11Y, 11C, 18Y and 18C.
The Anput and output field synchronizing signals are shown in
Figures 5A and 5B respectively. Figure 5C shows the output of the line counter 61 which repetitively counts from 0 to 524. Figures 51) and 5E show the outputs of the first and second latches 62 and 63 respectively. By latching the counter. 61, the required proportion of the input field period is determined. The temporal shift value tn indicates the position between two input fields where the output field must be interpolated such that when the shaded field shown in Figure
5A is dropped, continuous motion still occurs. Thus, the field which uses the temporal offset shown shaded in Figure 5E is the one that is dropped. It will be seen by reference to Figures 5A and 5B, that the field which is dropped is the one which does not have a new temporal shift associated with it. The field (arrowed) which is to be dropped is indicated to the following circuitry by the temporal freeze signal.
The derivation of the temporal offset signal in the case of 625/50 to 525/60 operation will now be described with reference to Figures 6 and 7.
In Figure 6, the control 32 is shown as including a line counter 71 and a latch 72. A line clock signal is supplied to a clock terminal of the line counter 71, while the input field synchronizing signal is supplied to a reset terminal of the line counter 71. The output field synchronization signal is supplied to a clock terminal of the latch 72. The output of the line counter 71 is supplied to the input of the latch 72, the output of which is the temporal offset signal supplied to the luminance and chrominance shift registers 11Y, 11C, 1BY and 18c.
The input and output field synchronizing signals are shown in
Figures 7A and 7B respectively. Figure 7C shows the output of the line counter 71 which repetitively counts from 0 to 624. Figure 7D shows the output of the latch 72. By latching the counter 71, the required proportion of the input field perioc is determined. Thus, the temporal shift value tn again indicates the position between two input fields where the output field must be interpolated, such that if the shaded field is repeated, continuous motion still occurs. The field which is repeated is the one which has two temporal shift values associated with it. The field (arrowed) which is to be repeated is indicated to the following circuitry by the temporal freeze signal.
The deviation of the temporal offset signal in the case of slow motion whether at 525/60 to 525/60 or 625/50 to 625/50 is the same, and will now be described with reference to Figures 8 and 9.
In Figure 8, the contr?l 32 is shown as including a line counter 81, a field counter 82, first to fourth latches 83 to 86, an exclusive-OR gate 87 and a scaler 88. The input field synchronizing signal is supplied to a clock terminal of the first latch 83, to a clock enable terminal of the field counter 82, and to a second reset terminal of the line counter 81. The input field polarity signal is supplied to the first latch 83 and thence to the second latch 84 and also to one input of the gate 87. The second latch 84 supplies an output to the second input of the gate 87, the output of which is supplied to a first reset terminal of the line counter 81, to a reset terminal of the field counter 82 and to a clock terminal of the third latch 85, which forms a speed detector latch. A line clock signal is supplied to a clock terminal of the second latch 84, and tQ respective clock terminals of the line counter 81 and the field counter 84. The output of the line counter 81 is supplied to an input terminal of the scaler 88, and the output of the field counter 82 is supplied to an input of the third latch 85 and also to an offset input terminal of the scaler 88. The output field synchronizing signal is supplied to a clock terminal of the fourth latch 86. The output of the third latch is supplied to a scale factor terminal of the scaler 88, the output of which is supplied to the fourth latch 86, the output of which is the temporal offset signal.
The input field synchronizing signal and the input field polarity signal are shown in Figures 9A and 9B respectively. Figure 9C also indicates the input field synchronizing signals and Figure 91) the
21 output field synchronizing signals. Figures 9E and 9F indicate the operations of the field counter 82 and the line counter 81, which are respectively counting fields and lines from 0 to N. Figure 9G indicates the output of the fourth latch 86 which is the temporal offset signal. Figure 9H indicates the temporal freeze signal (which is active when low), and, as indicated by the arrows, the shaded field that uses the temporal offset shown is a repeat of the previous field that used the temporal offset t1.
To generate the temporal freeze signal, the control 32 is shown in Figure 10 as including a synchronous RS flip-flop 91, a latch 92, an inverter 93 and an AND-gate 94. The output field synchronizing signal is supplied to one input of the flip-flop 91, to the input of the inverter 93 and to a clock enable terminal of the latch 92. The input field synchronizing signal is supplied to the other input of the flip-flop 91, while a line clock signal is supplied to clock terminals of the flip-flop 91 and the latch 92. The output of the flip-flop 91 is supplied to one input of the gate 94, which receives at its other input the output of the inverter 93. The output of the gate 94 is supplied to the input of the latch 92, the output of which forms the temporal freeze signal. The operation of this circuit is such that if more than one output field synchronizing pulse follows an input field synchronizing pulse, a freeze occurs.
Referring back to Figure 2, the generation of the vertical offset number by the control 32 will now be described. The same address generator which reads data from the luminance TBC 11Y into the luminance interpolator 1Y and the motion analyzer 2, also addresses an erasable programmable read-only memory (EPROM) which provides the vertical offset number together with vertical freeze signals when required.
(In the Figure 3 arrangement which is used for 525/60 to 625/50, the read addresses of the luminance TBC 18Y-are used, but in all other modes the read addresses of the luminance TBC 11Y are used.) The vertical offset number is generated assuming that both the input and the output fields are even, and it then indicates the position between two input lines where the output line must be interpolated such that a non-distorted picture would be produced if:
a line were occasionally dropped in 625/50 to 525/60 conversion, 22 or a line were occasionally repeated in 525/60 to 625.'50 conversion.
When a line is repeated by the luminance TBC 11Y (18Y), a vertical freeze signal is generated.
If the input fields are not both even, then the interpoiators 1Y and 1C must make use of the input field polarity and output field polarity to ensure correct interpolation.
The contents of the EPROM are generated in a way similar to that described above in connection with Figure 10 for the temporal offs'et signal, using the known line position in both a 525 and a 625 picture.
The form and operation of the vector selector 41, with which the present invention is particularly concerned, will now be described in more detaii with reference to Figures 11 to 13.
As explained above, 1200 motion vectors per field are generated, each motion vector being associated with a block of pixels arranged as eight lines by sixteen horizontal samples.
Of course one motion vector is not indicative of the motion within a particular block regardless of the size of that block.
Consequently it is necesary to provide a choice of motion vectors for each block such that every pixel within that block will have a fair chance of its motion being accurately estimated.
In the present case four motion vectors are selected from seven local motion vectors. These four motion vectors are then passed to a second processor, that is the vector selector 411, which selects one from four.The motion vectors are represented in the cartesian coordinate system and as such are available as a vertical component and a horizontal component. Associated with each motion vector is an FOM (figure of merit) which is based on the WAD (weighted absolute difference) within a block. This represents a quantitive assessment of the degree of correlation within a block.
The 1200 motion vectors form a motion vector array of thirty vertically by forty horizontally which is generated for each field.
Vertically the motion vectors are each displaced by eight lines whilst horizontam.11y they are each displaced by sixteen pixels, each motion 3 vector belng associated with an eight by sixteen block ol pixels.
Sub-pixel motion vector estimation means that it is possible t.-- 23 take one step beyond the motion vector estimate already determined as described above, by calculating to sub-pixel resolution the original motion vector estimate. In the example to be described it is possible to calculate to a resolution of one eighth of a line vertically and one quarter of a pixel horizontally.
As stated above, the vector reducer 40 (Figure 2) receives the motion vectors from the vector processor 38 and from the sub-pixel motion estimator 39. For each motion vector from the sub-pixel motion estimator 39, the six motion vectors closest to it are grouped together. For each motion vector there are then eleven choices. The vector reducer 40 selects four motion vectors which represent the possible motion of a portion of a picture within a block of pixels arranged as eight lines by sixteen horizontal samples.
The vector selector 41 determines which one of the four motion vectors most accurately, if at all, represents the motion of each pixel within the block. In particular, a technique which uses more motion vector is than two fields of data to determine the correct employed. Only the luminance signal is considered in the present description, the processing for the chrominance signal being similar.
The video data, used to determine the four motion vectors for each of the 1200, eight by sixteen blocks, is first filtered both vertically and horizontally to cause slight picture smear, as this improves the accuracy of the following process to be described. The vertical filter also interpolates every other incoming fild so that all fields appear to be even.
The video data is then applied to two field stores so that three temporal taps are available, as indicated in Figure 11. These three outputs are then converted to the required output field polarity by three further simple vertical interpolators. If the temporal taps are called field 0, field 1 and field 2, as indicated, then the four motion vectors to be processed represent motion between field 0 and field 1.
The motion vectors which are to be generated for each pixel will eventually be used to provide standards conversion by shifting the picture in another set of field stores in the luminance interpolator
1Y (Figure 2), such that a spatially correct picture can be generated between field 0 and field 1. The same portion of an object is taken
24 from each field and combined to produce the required output field.
Therefore, the four incoming vectors must also be tested for their accuracy at this position between field 0 and field 1.
Let the required output interpolatea fieid position be t, wnere t=O represents the position of field 0 and the maximum value of t, t=1, represents the position of field 1. Let V represent the motion vector to be tested with horizontal and vertical components. Wi th reference to each pixel:
If the picture at field 0 is shifted by tV and the picture at field 1 is shifted by -(l-t)V, the required output pixel should exist, at the same time, in both the shifted field 0 and the shifted field 1.
By subtracting these two pixels, a first attempt at determining which motion vector most accurately describes the motion of the required pixel in question can be made.
All four motion vectors are used individually to generate a pixel difference between field 0 and field 1, and the minimum difference is found. If however any of the three remaining motion vectors exhibit a pixel difference which is within a preset window of the minimum, these motion vectors are also considered possibly to represent the motion of the single pixel under test.
The number of related motion vectors could now be either one or, in the worst case, still four. To detefmine the correct motion vector, the remaining motion vectors are this time used to shift field
1 and field 2, providing a temporally extrapolated test. As indicated in Figure 11, field 1 is shifted -(l-t)V and field 2 is shifted (2 t)V. The motion vector which produces the minimum pixel difference, produced by this second stage of comparison, is then chosen as the primary choice of motion vector for the pixel under test.
An example of the portion of the vector selector 41 required to effect this processing is shown in block form in Figure 12. The four output motion vectors, here designated vectors A to D, are supplied to both of first and second programmable read-only memories (PROMS) 101 and 102, and also to a priority circuit 103. The pixels of field 0 and field 1 are supplied to the PROM 101, and the pixels of field 1 and field 2 are supplied to the PROM 102. The PROM 101 effects the shifting and differencing in respect of each of' the four vectors A to D as between field 0 and field 1, and supplies respective output
I difference signals (diff.) al to dl to a comparator circuit 104. By comparison of the magnitudes of the differehee signals al to dl the comparator circuit 104 selectively suplies output enable signals A to D to the PROM 102 so as selectively to enable the input thereto of the vectors A to D.
The PROM 102 then, if necessary, that is if more than one of the vectors A to D is enabled, effects the shifting and differencing in respect of the enabled vectors A to D as between field 1 and field 2, and supplies respective output difference signals (diff.) a2 to d2 as appropriate to a comparator circuit 105. By comparison of the magnitudes of those of the difference signals a2 to d2 supplied, the comparator circuit 105 supplies selected vector indicator signals a to d to the priority circuit 103 and thereby effects selection of the primary choice of motion vector, from the vectors A to D, for the pixel under test.
It is possible that an output pixel position under test could have more than one possible motion associated with it, indicated by motion vectors, other than the primary motion vector, having a pixel difference which is within a preset window of the minimum. There are two possible reasons for this situation to arise. Firstly, there may well be a number of different motion vectors surrounding the pixel, such as would occur in an image of an explosion. Secondly, if there were a large plain area around some fine detail, a large number of motion vectors would apparently be correct, according to the pixel difference criteria outlined. This is illustrated diagrammatically in Figure 13, in which 0 represents a sample in field 0 and X represents a sample in field 1.
The plain areas surrounding areas of detail could well produce the minimum pixel difference and cause the fine detail to be reproduced with excessive noise.
To prevent this situation occurring, the selection of three or four motion vectors, with similar pixel differences and whose magnitudes or directions are significantly different, can be used to represent a possible error in the motion vector selection. This is done separately for horizontal and vertical components and the corresponding components set to zero if such an error is detected.
Where this results in the loss of motion vectors, it is usually not 26 important, because it occurs in an image, for example, of an explosion which would in any case be visually confusing, and it would not be possible to track any errors in standards conversion. If there are only two matching motion vectors, the final choice of output
motion vector is chosen using a priority basis.
The motion vectors are always arranged in the order A, B, C and D. Vector A will normally be the central local motion. vector and vectors B, C and D will probably be global vectors again arranged in order depending on the number of times the motion vectors are detected. For example: if vector C where chosen as the actual minimum difference but vector A or B were also very close to this minimum value, then vector A or B would be chosen as the final output motion vector.
An output motion vector could also be in error if the minimum pixel difference were relatively large. In such a case the output motion vector would be set to zero,. so that the pixel would be reproduced as in normal standards conversion. An alternative is to conceal such missing motion vectors by using the motion vector values surrounding it.
Thus in summary, the processing performed by the vector selector
41 involves the following steps:
Firstly, selection by direct pixel matching. Secondly, refinement by extrapolation of match across a third field. Finally, additionally vector checking to correct the multiple selection of motion vectors.
Attention is drawn to seven other UK patent applications relating to subject matter similar to that of the present application, which we filed on the same day as the present application, and the disclosures in which are incorporated herein by reference.
27
Claims (26)
1 A method of processing motion vectors associated with a digital television image, the method comprising the steps of:
deriving a plurality of motion vectors respectively representing the motion of a pixel in said image; testing the accuracy of each said motion vector by differencing pixels shifted in dependence on said motion vector from two successive fields of said image to a common position in a field interpolated between said two fields; and selecting a motion vector from said plurality in dependence on said testing.
2. A method according to claim 1 comprising the step of further testing the accuracy of at least one of said plurality of motion vectors by differencing pixels shifted in dependence on said at least one motion vector from the second of said two fields of said image and from the field of said image immediately succeeding said second field to said common position, and selecting a motion vector from said plurality in dependence on said further testing.
3. A method according to claim 1 or claim 2 wherein said motion vectors are derived by determining motion in said image at points spaced a predetermined number of samples horizonally and a predetermined number of samples vertically by a block matching technique.
4. A method according to claim 3 wherein said block matching technique comprises the steps of:
testing for minimum difference in three positions, the centre position of a block, a predetermined number of samples to the left, and the same predetermined number of samples to the right; starting from the point indicated above, testing for minimum difference in nine positions symmetrically distributed about the above starting point in steps of a smaller predetermined number of samples or lines; starting from the point indicated above, testing for minimum 28 difference in nine positions symmetrically distributed about the above starting point in steps of a still smaller predetermined number of samples or lines; -ing for minimum starting from the point indicated above, test difference in nine positions symmetrically distributed about the above starting point in steps of a still smaller predetermined number of samples or lines, and starting from the point indicated above, testing for minimum difference in nine positions symmetrically distributed about the above starting point in steps of one sample or line.
5. A method according to claim 4 comprising a further step, subsequent to the last step of claim 4, of comparing the difference produced at the final position indicated by said last step with the two differences above and below to adjust the vertical vector value, and with the two differences to the left and right to adjust the horizontal vector value.
6. Apparatus for processing motion vectors associated with a digital television image, the apparatus comprising:
means for deriving a plurality of motion vectors respectively representing the motion of a pixel in said image; means for testing the accuracy of each said motion vector by differencing pixels shifted in dependence on said motion vector from two successive fields of said image to a common position in a field interpolated between said two fields; and means for selecting a motion vector from said plurality in dependence on said testing.
7. Apparatus according to claim 6 comprising means for further testing the accuracy of at least one of said plurality of motion vectors by differencing pixels shifted in dependence on said at least one motion vector from the second of said two fields of said image and from the field of said image immediately succeeding said second field to said common position, and means for selecting a motion vector fron.
said plurality in dependence on said further testing.
29
8. Apparatus according to claim 6 or claim 7 wherein said means for deriving motion vectors comprises a vector filter and a vector calculator which determine motion in said image at points spaced a predetermined number of samples horizontally and a predetermined number of samples vertically by a block matching technique.
9. Apparatus according to claim 8 wherein said vector calculator is arranged to perform the following steps on a digitized signal representing said image and supplied to said vector calculator by said vector filter; testing for minimum difference in three positions, 1.
-he centre position of a block, a predetermined number of samples to the left, and the same predetermined number of samples to the right; starting from the point indicated above, testing for minimum difference in nine positions symmetrically distributed about the above starting point in steps of a smaller predetermined number of samples or lines; starting from the point indicated above, testing for minimum difference in nine positions symmetrically distributed about the above starting point in steps of a still smaller predetermined number of samples or lines; starting from the point indicated above, testing for minimum difference in nine positions symmetrically distributed about the above starting point in steps of a still smaller predetermined number of samples or lines, and starting from the point indicated above, testing for minimum difference in nine positions symmetrically distributed about the above starting point in steps of one sample or line.
10. Apparatus according to claim 9 wherein said vector calculator is arranged to perform a further step, subsequent to the last step of claim 9, of comparing the difference produced at the final position indicated by said last step with the two differences above and below to adjust the vertical vector value, and with the two differences to the left and right to adjust the horizontal vector value.
11. A television standards converter comprising:
means comprising apparatus according to any one of claims 6 to 10 for -elevision analyzing the motion between consecutive fields of an input t signal of one television standard; means then to align said fields so as effectively to represent static pictures; and means to effect conversion using said static pictures to derive the required output television signal of a different television standard.
12. A television standards converter according to claim 11 wherein said means to effect conversion comprises an interpolator for effecting vertical /temporal interpolation, and said input television signal is supplied to said interpolator by way of a time base corrector which produces therefrom a 585-line 60-fields per second television signal.
13. A television standards converter according to claim 12 wherein said input television signal is a 625-line 50-fields per second signal, said time base corrector is a 4-field time base corrector, and the output of said time base corrector is supplied to said interpolator by way of a 4-field shift regist er.
14. A television standards converter according to claim 12 wherein said input television signal is a 525-line 60-fields per second signal, said time base corrector is a 2-field time base corrector, and the output of said time base corrector is supplied to said interpolator by way of a 4-field shift register.
15. A television standards converter according to any one of claims 11 to 14 wherein said means to align said fields operates to vary the address -of a variable delay element to reposition each pixel of the picture to the nearest line or sample, and then reposition each pixel of the picture both vertically and horizontally to a fraction of a line and a fraction of a sample respectively.
31
16. A television standards converter according to claim 15 wherein said vertical repositioning to a fraction. of a iine is done by a vertical interpolator with four taps per field, and said horizontal repositioning to a fraction of a sample is done by a horizontal filter having two or four taps.
17. A 625-line 50-fields per second to 525-line 60-fields per second television standards converter comprising:
a 4-field time base corrector for receiving an input 625-line 50- fields per second digital television signal; a motion analyzer comprising apparatus according to any one of claims 6 to 10 connected to the output of said time base corrector for analyzing motion in said input television signal; a shift register also connected to the output of said time base corrector; an interpolator for deriving samples of a required output 525-line 60 fields per second digital television signal in dependence on samples derived from said shift register and picture motion data derived by said motion analyzer; and a 2-field time base corrector for assembling said derived samples to form said output television signal.
18. A television standards converter according to claim 17 wherein said 4-field time base corrector derives a 585-line 60-fields per second television signal from said input television signal for supply to said shift register.
19. A 525-line 60-fields per second to 625-line 50-fields per second television standards converter comprising:
a 2-field time base converter for receiving an input 525-line 60 fields per second digital television signal; a motion analyzer comprising apparatus according to any one of claims 6 to 10 connected to the output of said time base corrector for analyzing motion in said input television signal; a shift register also connected to the output of said time base corrector; an interpolator for deriving samples of a required output 625-line 50- 32 fields per second digital television signal in dependence on samples derived from said shift register and picture motion data derived by said motion analvzer; and a 4-field time base corrector for assembling said derived samples to form said output television signal.
20. A television standards converter according to claim 19 wherein said 2-field time base corrector derives a 585-line 60'-fields per second television signal from said input television signal for supply to said shift register.
21. A television standards converter according to claim 18 or claim wherein said motion analyzer derives motion vectors in dependence on the motion between corre-sponding pixels in consecutive fields of said input television signal, and supplies said motion vectors to said interpolator so as effectively to align said pixels in said consecutive fields to represent static pictures.
22. A slow motion processor comprising:
an input circuit for receiving an input digital television signal; a motion analyzer comprising apparatus according to any one of claims 6 to 10 for analyzing motion in said input digital television signal; a shift register for holding successive different fields of said input television signal; an interpolator for deriving samples of a required slow motion output digital television signal in dependence on the degree of slow motion, samples derived from said shift register, and picture motion data derived by said motion analyzer; and a 2-field time base corrector for assembling said derived samples to form said slow motion output television signal.
23. A slow motion processor according to claim 22 wherein said input circuit is a 4-field time base corrector.
24. A slow motion processor according to claim 23 wherein said 4 field time base corrector derives a 585-line 60-fields per second television signal from said input television signal for supply to said 33 shift register.
25. A method of processing motion vectors, the method being substantia'Lly as hereinbefore described with reference to Figures 11 and 12 of the accompanying drawings.
26. Apparatus for processing motion vectors, the apparatus being substantially as hereinbefore described with reference to Figures 11 and 12 of the accompanying drawings.
Published 1988 at The Patent Office, State House, 66,171 High Holborn, London WCIR 4TP. Puzther copies maybe obtained from The Patent Office, Sales Branch, St Mary Cray, Orpington, Kent BR5 3RD. Printed by Multiplex techniques ltd, St. Mary Cray, Kent. Con. 1/87.
Priority Applications (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP88304620A EP0294961B1 (en) | 1987-06-09 | 1988-05-20 | Motion vector selection in television images |
DE3851786T DE3851786T2 (en) | 1987-06-09 | 1988-05-20 | Selection of a motion vector in television pictures. |
US07/200,503 US4864394A (en) | 1987-06-09 | 1988-05-31 | Motion vector selection in television images |
JP63142358A JPS63313984A (en) | 1987-06-09 | 1988-06-09 | Television image motion vector processor |
KR1019880006901A KR0141701B1 (en) | 1987-06-09 | 1988-06-09 | Motion vector processing in digital television images |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
GB878713455A GB8713455D0 (en) | 1987-06-09 | 1987-06-09 | Television standards converters |
Publications (3)
Publication Number | Publication Date |
---|---|
GB8728450D0 GB8728450D0 (en) | 1988-01-13 |
GB2205711A true GB2205711A (en) | 1988-12-14 |
GB2205711B GB2205711B (en) | 1991-09-04 |
Family
ID=10618610
Family Applications (6)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
GB878713455A Pending GB8713455D0 (en) | 1987-06-09 | 1987-06-09 | Television standards converters |
GB8728449A Expired - Lifetime GB2205710B (en) | 1987-06-09 | 1987-12-04 | Motion vector estimation in television images |
GB8728450A Expired - Lifetime GB2205711B (en) | 1987-06-09 | 1987-12-04 | Motion vector selection in television images |
GB8728445A Expired - Lifetime GB2205706B (en) | 1987-06-09 | 1987-12-04 | Motion vector reduction in television images |
GB8728447A Expired - Lifetime GB2205708B (en) | 1987-06-09 | 1987-12-04 | Motion vector processing in digital television images |
GB8728451A Expired - Lifetime GB2205712B (en) | 1987-06-09 | 1987-12-04 | Motion vector processing in television images |
Family Applications Before (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
GB878713455A Pending GB8713455D0 (en) | 1987-06-09 | 1987-06-09 | Television standards converters |
GB8728449A Expired - Lifetime GB2205710B (en) | 1987-06-09 | 1987-12-04 | Motion vector estimation in television images |
Family Applications After (3)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
GB8728445A Expired - Lifetime GB2205706B (en) | 1987-06-09 | 1987-12-04 | Motion vector reduction in television images |
GB8728447A Expired - Lifetime GB2205708B (en) | 1987-06-09 | 1987-12-04 | Motion vector processing in digital television images |
GB8728451A Expired - Lifetime GB2205712B (en) | 1987-06-09 | 1987-12-04 | Motion vector processing in television images |
Country Status (2)
Country | Link |
---|---|
KR (5) | KR970010043B1 (en) |
GB (6) | GB8713455D0 (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2231226A (en) * | 1989-04-27 | 1990-11-07 | Sony Corp | Motion dependent video signal processing |
Families Citing this family (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2222500A (en) * | 1988-09-05 | 1990-03-07 | Philips Electronic Associated | Picture motion measurement |
GB2231748B (en) * | 1989-04-27 | 1993-08-18 | Sony Corp | Motion dependent video signal processing |
GB2248361B (en) * | 1990-09-28 | 1994-06-01 | Sony Broadcast & Communication | Motion dependent video signal processing |
JPH04236664A (en) * | 1991-01-18 | 1992-08-25 | Sony Corp | Arithmetic circuit |
JP3123130B2 (en) * | 1991-06-28 | 2001-01-09 | ソニー株式会社 | Motion vector detection device |
EP0557204A3 (en) * | 1992-02-21 | 1995-10-18 | Sony Corp | Discrete cosine transform apparatus and inverse discrete cosine transform apparatus |
DE4226128A1 (en) * | 1992-08-07 | 1994-02-10 | Philips Patentverwaltung | Procedure for determining motion vectors |
KR940004467A (en) * | 1992-08-26 | 1994-03-15 | 오오가 노리오 | Discrete cosine inverters and their inverses |
JPH06103301A (en) * | 1992-09-17 | 1994-04-15 | Sony Corp | (8x8) discrete cosine transforming circuit and (8x8) discrete cosine inverse transforming circuit |
JPH06149862A (en) * | 1992-11-13 | 1994-05-31 | Sony Corp | Matrix data multiplying method and device |
JP3596519B2 (en) * | 2001-12-13 | 2004-12-02 | ソニー株式会社 | Image signal processing apparatus and method |
KR20070069615A (en) | 2005-12-28 | 2007-07-03 | 삼성전자주식회사 | Motion estimator and motion estimating method |
CN105212611B (en) * | 2013-08-29 | 2017-03-22 | 郑牧之 | Bamboo mat with springs and manufacturing method of bamboo mat |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS61200789A (en) * | 1985-03-04 | 1986-09-05 | Kokusai Denshin Denwa Co Ltd <Kdd> | System for detecting dynamic vector of object on picture plane |
DE3682068D1 (en) * | 1986-03-19 | 1991-11-21 | British Broadcasting Corp | TELEVISION MOTION MEASUREMENT. |
-
1987
- 1987-06-09 GB GB878713455A patent/GB8713455D0/en active Pending
- 1987-12-04 GB GB8728449A patent/GB2205710B/en not_active Expired - Lifetime
- 1987-12-04 GB GB8728450A patent/GB2205711B/en not_active Expired - Lifetime
- 1987-12-04 GB GB8728445A patent/GB2205706B/en not_active Expired - Lifetime
- 1987-12-04 GB GB8728447A patent/GB2205708B/en not_active Expired - Lifetime
- 1987-12-04 GB GB8728451A patent/GB2205712B/en not_active Expired - Lifetime
-
1988
- 1988-06-09 KR KR1019880006900A patent/KR970010043B1/en not_active IP Right Cessation
- 1988-06-09 KR KR1019880006903A patent/KR0141703B1/en not_active IP Right Cessation
- 1988-06-09 KR KR1019880006904A patent/KR0141705B1/en not_active IP Right Cessation
- 1988-06-09 KR KR1019880006901A patent/KR0141701B1/en not_active IP Right Cessation
- 1988-06-09 KR KR1019880006905A patent/KR0141702B1/en not_active IP Right Cessation
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2231226A (en) * | 1989-04-27 | 1990-11-07 | Sony Corp | Motion dependent video signal processing |
GB2231226B (en) * | 1989-04-27 | 1993-09-22 | Sony Corp | Motion dependent video signal processing |
Also Published As
Publication number | Publication date |
---|---|
KR0141702B1 (en) | 1998-07-01 |
KR890001385A (en) | 1989-03-20 |
GB2205710A (en) | 1988-12-14 |
GB2205708A (en) | 1988-12-14 |
KR0141705B1 (en) | 1998-07-01 |
KR970010043B1 (en) | 1997-06-20 |
KR890001384A (en) | 1989-03-20 |
GB8728445D0 (en) | 1988-01-13 |
GB2205706A (en) | 1988-12-14 |
GB2205710B (en) | 1991-09-04 |
GB2205706B (en) | 1991-09-04 |
GB8728450D0 (en) | 1988-01-13 |
KR0141701B1 (en) | 1998-07-01 |
GB8728447D0 (en) | 1988-01-13 |
GB2205711B (en) | 1991-09-04 |
KR890001380A (en) | 1989-03-20 |
GB2205708B (en) | 1991-09-04 |
GB8713455D0 (en) | 1987-07-15 |
GB8728449D0 (en) | 1988-01-13 |
GB2205712A (en) | 1988-12-14 |
GB8728451D0 (en) | 1988-01-13 |
GB2205712B (en) | 1991-09-04 |
KR0141703B1 (en) | 1998-07-01 |
KR890001381A (en) | 1989-03-20 |
KR890001383A (en) | 1989-03-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US4864394A (en) | Motion vector selection in television images | |
EP0294956B1 (en) | Motion vector reduction in television images | |
US4864393A (en) | Motion vector estimation in television images | |
US4862267A (en) | Motion compensated interpolation of digital television images | |
EP0294957B1 (en) | Motion vector processing in digital television images | |
EP0294955B1 (en) | Motion vector estimation in television images | |
EP0294959B1 (en) | Television standards converters | |
EP0294960B1 (en) | Motion vector processing in television images | |
EP0852096B1 (en) | Motion compensated interpolation | |
JPH01877A (en) | television format converter | |
GB2205711A (en) | Motion vector selection in television images | |
GB2205713A (en) | Motion compensated interpolation of digital television images | |
KR970010044B1 (en) | Method and apparatus for motion vector estimation in television images |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PE20 | Patent expired after termination of 20 years |
Effective date: 20071203 |