WO2008075261A1

WO2008075261A1 - Automatic noise reduction improvement in video sequences

Info

Publication number: WO2008075261A1
Application number: PCT/IB2007/055056
Authority: WO
Inventors: Perry G. Mevissen
Original assignee: Koninklijke Philips Electronics N.V.
Priority date: 2006-12-18
Filing date: 2007-12-12
Publication date: 2008-06-26

Abstract

To a video signal comprising groups (GOP) of intra (I-frame) and intercoded frames (P-frame, B-frame) a noise reduction is applied. The strength of the noise reduction strength has a modulation in time with a period substantially equal to the period (Tl) between intra coded frames. The noise reduction strength has a maximum substantially co inciding with the position of intra coded frames.

Description

Automatic noise reduction improvement in video sequences

BACKGROUND OF THE INVENTION

This invention relates generally to video signal processing and more particularly to improved methods of and systems for noise reduction in a video signal.

The invention also relates to a device comprising a system for noise reduction in a video signal.

In video signal processing the signal is comprised of intracoded and intercoded frames, for instance I-frames, P-frames and B-frames. The I-frames are intra-coded. The P- and B-frames are referred to as intercoded frames. Intra-code frames can be reconstructed without any reference to other frames; intercoded frames are reconstructed using data of other frames (forward or backward prediction).

Image signals are commonly encoded digitally for transmission and then decoded for display. For example, the Motion Pictures Experts Group is an ISO group that sets standards (MPEG standards) for compressing and storing video, audio and animation in digital form. MPEG-I is a standard audio and video coding format for low density storage media such as CD-ROMs video CDs and so forth. MPEG-2 is a standard coding format for broadcast video. In MPEG, the compression method involves I-frames or intra frames. A frame refers to a complete TV picture. It can be made up of multiple fields, such as a field of odd numbered lines and a field of even-numbered lines. An I-frame is an intra-coded video frame that is independent of other video frames in MPEG standard. I-frames are repeated at a regular interval to refresh the coding sequence. Between any two I-frames are P-frames (Predictive frames) and B-frames (Bi-directional predictive frames). The P-and B-frames only contain information or changes between the I-frames. An I-frame and its P-and B- frames are called a GOP (Group of Pictures). The video sequence is composed of groups of pictures (GOP 's) that follow one after the other. The invention is not restricted to MPEG, and could be used in other standards using intracoded and intercoded frames, such as DivX and Xvid.

Many compression methods, such as MPEG compression methods, are lossy compression methods. When images are coded, decoded, and subjected to digital post processing, the display can develop undeliverable characteristics. For example, when a portion of an image has saturated colors, that portion of the image can experience significant enhancement of noise when conventional filters are applied to peak the image. This can be particularly true of blue sections of an image, to which the eye is most sensitive.

Thus video signals typically involve some type of noise component. The noise can be present in the actual video signal. Noise can also be introduced by processing circuitry, such as at the encoder, decoder, transmitter, analog-digital converter (ADC), digital-analog converter (DAC) and so forth.

Noise reduction methods applied at the decoding end are known to reduce noise in the displayed video image. Such noise reduction algorithms typically comprise spatial or local recursive noise reduction algorithms. Conventional algorithms apply an average noise reduction (filter) strength in the noise reduction algorithms.

Although the conventional noise reduction methods for video sequences do provide a considerably reduction in noise, there remains a desire to improve the noise reduction.

Accordingly, it is desirable to provide improved methods and systems for noise reduction of video sequences.

SUMMARY OF THE INVENTION

Generally speaking, in accordance with the invention, a method and apparatus for noise reduction of a sequence of video frames is provided.

The method in accordance with the invention is a method for noise reduction in a video signal comprising groups of intra and intercoded frames wherein to the frames of the video sequence a noise reduction is applied wherein the noise reduction has a noise reduction strength and the noise reduction strength has a modulation in time with a period substantially equal to the period between intra coded frames, wherein a maximum in the modulation substantially coincides with a intra coded frame.

The system in accordance with the invention is a system for noise reduction in a video signal comprising groups of intra and intercoded frames wherein the system comprises a noise reductor to apply to the frames of the video sequence a noise reduction wherein the noise reductor is arranged such that the noise reduction has a noise reduction strength and the noise reduction strength has a modulation in time with a period substantially equal to the period between intra coded frames, wherein a maximum in the modulation substantially coincides with a intra coded frame.

The invention is based on the following insight: When a scene is compressed with algorithms like for instance MPEG compression the signal to noise ratio starts modulating with the I-frames (I-frame pumping), as the P-, and B-frames are implicitly filtered stronger during reconstruction in the decoder.

If an image has been encoded and decoded than the noise may be worse on frames that were transmitted as intracoded frames, I-frames compared to intercoded frames such as P-/B-frames. Intracoded I-frames will contain more noise and quantization artifacts then the intercoded P- and B-frames because the latter two frame-types, being intercoded frame types, have been interpolated hence filtered stronger during reconstruction in the decoder.

If an image has been encoded and decoded than the noise may be worse on frames that were transmitted as intracoded frames, I-frame, compared to intracoded frames such as P-/B-frames. I-frames will contain more noise and quantization artifacts then the P- and B-frames because the latter two frame-types have been interpolated hence filtered stronger during reconstruction in the decoder. The noise is peaked at the I-frame. Herein below this effect will also be called I-frame pumping, i.e. the noise is pumped at the I-frames.

If an average (temporal or spatial) noise reduction factor is used in a noise reduction algorithm, averaging the noise reduction of a sequences of frames, thus comprising intracoded as well as intercoded frame, the required noise reduction strength is underestimated for intracoded frames and for the intracoded frames too much noise passes the noise filter formed by the noise reduction algorithm.

The invention is not restricted to a particular type of noise reduction, but it is in particular advantageous when use is made of recursive noise reduction and loop systems for recursive noise reduction.

When a signal with significant I-frame pumping is processed with a temporal recursive noise reduction algorithm then the noisier than average I-frame may break through the dynamic threshold of the temporal recursive noise reduction processor. The result is that the I-frame' s extra noise ends up in the recursive path of the loop system. This will result in just propagating the noise of the I-frame into the following B-frames. The effect of temporal recursive noise reduction is less effective or even counter effective if not implemented and programmed as a stable closed loop system.

"Noise reduction" algorithm include within the framework of the invention algorithms to prevent noise from occurring. Coring, a method to prevent sharpening/enhancing noise is an example of such noise preventing algorithms. Likewise "noise reduction' as a method step includes method steps aimed at preventing noise from arising, such methods including coring.

The invention accordingly comprises the several steps and the relation of one or more of such steps with respect to each of the others, and any method, system, device and computer program adapted to effect or resulting from such steps, all as exemplified in the following detailed description and drawings and the scope of the invention will be indicated in the claims.

BRIEF DESCRIPTION OF THE DRAWINGS

For a fuller understanding of the invention, reference is had to the following description, taken in connection with the accompanying drawings, in which:

Figs. Ia and Ib show a number of frames in MPEG display order and MPEG transfer order respectively.

Fig. 2 shows an unfiltered MPEG artifact meter output plotted along a temporal axis.

Fig. 3 illustrates a modulated noise reduction strength.

Fig. 4 shows unfiltered (or weakly filtered) Mu and filtered Mf MPEG artifact meter data.

Fig. 5 shows unfiltered (or weakly filtered) Mu and filtered MfMPEG meter data and the confidence factor C.

Fig. 6 schematically shows temporal recursive noise reduction.

Fig. 7 schematically shows spatial recursive noise reduction.

Fig. 8 schematically shows a device in accordance with the invention

The Figs, are not drawn to scale. Generally, identical components are denoted by the same reference numerals in the Figs.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Figs. Ia and Ib illustrate sequences of MPEG frames. Fig. Ia illustrates the MPEG display order. The I-frames are infra coded frames, i.e. they can be reconstructed without any reference to other frames. The P-frames are forward predicted from the last I- frame or P-frame, so it is impossible to reconstruct them without the data of another frame (I or P). The B frames are forward en backward predicted from the last/next I-frame or P-frame. P-frames and B-frames are referred to as inter coded frames. The frame for backward prediction follows the predicted frame. That would require, if the frames would be coded according to the MPEG display order, to suspend the decoding the B-frames till the next P- or B-frame appears. Fig. Ib illustrates an MPEG coded sequence. In the data stream the frames appear in such an order (I P B B B P B B B I P B B B P B B B I etc.) that the referred to frames precede the referring frames. The only task of the decoder is to reorder the reconstructed frames. MPEG DO stands for MPEG Display Order, MPEG TO stands for MPEG Transfer Order. Ti denotes the period of a GOP (group of pictures). The intraframes are in time separated by T₁. Ti is below also called the GOP interval.

It is remarked that everywhere in this document where the term MPEG is used also all other derivations of the MPEG compression are meant. (Examples: DivX, XVid, ASP etc.)

Video signals typically involve some type of noise component. The noise can be present in the actual video signal. Noise can also be introduced by processing circuitry, such as at the encoder, decoder, transmitter, analog-digital converter (ADC), digital-analog converter (DAC) and so forth.

The inventor has realized that when a scene is compressed with algorithms like MPEG compression the signal to noise ratio starts modulating with the intracoded I- frames (in this application the word I-frame pumping is used to describe this effect), because the intercoded P-, and B-frames are implicitly filtered stronger during reconstruction in the decoder.

If an image has been encoded and decoded than the noise may be worse on frames that were transmitted as intracoded frames (I-frames) compared to intercoded frames (P -/B-frames). Intracoded I-frames will contain more noise and quantization artifacts then the intercoded P- and B-frames because the latter two frame-types have been interpolated hence filtered stronger during reconstruction in the decoder.

MPEG artifact meters exist that with some filtering produce a value representative for the average amount of MPEG compression artifacts. However if the MPEG meter is evaluated temporally unfiltered, then in the temporal domain a clear peaked pattern can be seen with its peaks coinciding with the I-frames. This effect is shown in Fig. 2 where M_u is the unfiltered MPEG meter output is plotted along a temporal axis. Analysis of this 'beat' allows predicting when an I-frame will occur because MPEG streams are mostly coded with a fixed GOP interval T₁.

For an I-frame the noise reduction (both temporal recursive and spatial) should be set stronger then on the P- and B-frames. This is the core of the invention. Within the framework of the invention "noise reduction" includes method to prevent noise such as coring, a method to prevent sharpening/enhancing noise. Such methods are performed by or via algorithms. In such algorithms the coring strength is a parameter of the algorithm In coring methods in accordance with the invention a modulated coring strength is applied to I- frames stronger than to P and B-frames.

The position of the I-frames can be found by measuring the cadence in the artifacts. Advanced high end market digital set-top boxes include image processing algorithms that provide for the location of the I-frames. The location of the I-frames is then known and need not be found separately.

Fig. 3 illustrates that in the invention the intracoded I-frames are filtered stronger then the other frames. The image is valid for all types of noise reduction. NR stands for the noise reduction strength. The dotted line in Fig. 3 shows a steady reduction strength NR. The continuous line shows the modulated reduction strength with its peaks coinciding with the I-frames of the decoded MPEG stream. The average of the continuous line over one GOP-period can be equal to the level of the conventional steady reduction (dotted line) so the integral amount of reduction is still the same while the visual effect is better. The modulation D on NR is in the time domain, with a period T_1; equal to the I-frame period (GOP). Noise reduction is modulated such that an I-frame is filtered stronger then the P-, and B-frames. On the noise filtered output this will give more weight to the relatively cleaner P- and B-frames compared to the I-frames.

The modulation depth is the ratio between the higher noise reduction levels at the I-frame over the lower noise reduction levels at the P- and B-frames. The average amount of noise reduction is kept the same, such that a modulation does not change the noise reduction on average. To keep the average amount of noise reduction the same for all modulation depths, the average of noise reduction on the I-frame and the P- and B-frames must be equal such that

∑NR_a = NRΪ + ∑NR_m

GOP B+P

with NRu = un-modulated noise reduction and NR_m = modulated noise reduction, I stands for the intracoded frames and B+P for the intercoded frames.

In other words, the amount of extra noise reduction on the I-frames is in preferred embodiments compensated by applying less noise reduction on the P- and B-frames to keep the average over the GOP at the same level. In an example the relation can be formulated as follows:

NR_m' = NR_u + NR_u - (D -I)IA

And

NR_m ^{P B} = NR₁₁ - NR₁₁ (D - 1) /(A - (GOP - 1))

With D as modulation depth > 1 , A as the accuracy < D and NR_U as the unmodulated noise reduction and GOP is the number of frames in a GOP. In this example the P and B frames are treated equally.

The modulation depth D can be calculated from several statistics. An important parameter may be the confidence factor C. If not confident about the I-frame beat (C -> O) then D should be 1 or very close to 1, e.g. no modulation to very little modulation. The influence of C can be controlled too.

Another parameter could take into account the relevance of this whole algorithm at dark scenes. A dark scene can be evaluated in many smart ways with help of stochastic distributions and their deviations, relevance related to the size of dark surfaces, ignoring of sub-black levels etc. or in a simple way as the average luminance value of the image. Given here is that we know how dark the image is, expressed as Y. With this determined the modulation depth D can be slightly increased too when Y decreases.

At last there are the application flavors F, and F_max giving the application a means of control over the modulation.

In an embodiment the modulation depth D look as follows:

D = l + MIN(^ ₇-^β ,F, MAX ,

α is the amount of influence of C, β > 1 is the inverse amount of influence of Y , and

Y > 1 by limitation, with β rather chosen as a power function then as linear multiplier.

This provides a substantially linear relation to the gamma of the video signal. It is remarked that it is needed to have in the above exemplary formulas an NRu level of > 0 to start with. In the formulas for NR_m above, if NR_U = 0 then the whole modulation makes no sense. The resulting NR_m will be 0 if NR_U = 0.

In this example there needs to be a sufficient amount of NR_U to work with. Most digital signals have hardly to any analogue noise at all. So any conventional automatic analogue noise reduction algorithm would have set the amount of NR_U to a low level because of low analogue noise levels. The NR_U mentioned here preferably is a function of the conventional evaluated noise reduction level and a new evaluation of the noise reduction level based on MPEG measurement.

NR_u = f(NR_a,NR_d)

NRa being the noise reduction level based on analogue noise measurement and NRj being the un-modulated noise reduction level based on the filtered MPEG measurement M_f.

An embodiment of the invention is illustrated in Fig. 4 to 7.

Building elements of this embodiment are:

A conventional MPEG artifact meter or any meter capable of measuring artifacts or noise in a frame.

Temporal and/or Spatial noise reduction.

(optional) Sharpening functions with coring and anti-peaking mechanisms.

Embedded processor able to perform video frame synchronized calculations.

Given the following:

An MPEG artifact meter produces a temporally unfiltered (or weakly filtered) and normalized measurement M_u.

An MPEG artifact meter produces a temporally filtered and normalized measurement M_f.

On video signals that were once MPEG compressed the unfiltered (or weakly filtered) M_u and filtered MPEG meter data Mf looks like Fig. 4. Ti is the repetition period (GOP) of the I-frame. The occurrences of I-frames coincide with the peak in the unfiltered measurement M_u.

With a simple recurring comparison function it can be detected that the unfiltered M_u data is periodically (with the I-frame or GOP period) larger ten the average. So the 'beat' of the I-frame can be detected and the location of the next I-frame can be predicted. The modulation depth of the modulation has to be provided. This modulation depth can be provided in several manners. A simple one is a standard modulation depth related to the average value. The modulation depth is then not determined separately but follows from the average value. This can be a standard percentage modulation of the average value or a more complex, for instance non- linear function of the average value. The modulation depth may also be made dependent on the darkness of the shot. Preferably the modulation depth is, at least for a range of average image intensities a decreasing function for increasing average image intensity. The modulation depth D increases for relatively dark scenes where errors are more clearly visible. The dependence on image intensity may be such that above a certain intensity level no modulation or a very simple type of modulation is used, not dependent on image intensity, whereas below a threshold image intensity a dependence of the modulation depth on intensity is used, which becomes more complex as the intensity goes down. Some exemplary methods for determining D are given above.

Additional values for determining the modulation depth may be provided as follows.

For each time that the unfiltered MPEG artifact meter M_u is clearly bigger then the filtered MPEG artifact meter M_f exactly (within error margins) on a repetitive interval a confidence counter C can be increased (Fig. 5). For each time that the unfiltered MPEG artifact meter M_u is clearly bigger then the filtered MPEG artifact meter M_f but not on a repetitive interval, a confidence counter can be decreased. Such that a confidence C is calculated on every occurrence of an I-frame for instance by the following formula:

λ * M₁₁ (t) ≥ M_f (t)\AND[λ *M_u (t -T_j) ≥ M_f (t - T₁ )J}→ C(t - 1) + J

C(O = else → C(t -I) -I

Here λ is a factor to determine 'how clearly bigger' Mu should be then Mf on the I-frames.

Here J is a constant to be added when the event coincides with an I-frame peak in the MPEG meter.

Now that we have an MPEG Fig. Mf and a confidence factor C for the cadence we can start to control the temporal- and spatial noise reduction. The confidence factor C if low, reduces the modulation depth, if the confidence factor C is high, the modulation depth D increases. Fig. 6 shows the diagram of temporal recursive noise reduction. The local adaptive mixer compares the image to the previous image and calculates filter strength for each pixel. The result is stored in a frame memory so this result can be used for filtering the next image. The trick of the trade is to adjust the local adaptive mixer such that a proper balance is made in the recursion factor such that the filtering is strong enough to reduce noise and MPEG artifacts, but does not introduce too much (motion) artifacts. When a signal with significant I-frame pumping is processed with a temporal recursive noise reduction algorithm then the noisier I-frame may break through the dynamic threshold of the temporal recursive noise reduction processor. The result is that the I-frame's extra noise ends up in the recursive path. This will result in just propagating the noise of the I-frame into the following B-frames. The effect of temporal recursive noise reduction is then less effective or even counter effective if not implemented and programmed exactly as a stable closed loop system. In the embodiments of the invention the data stream is analyzed in e.g. an MPEG artifact meter (MPEG AM) which provides for a measure of Mf, the cadence, i.e. where the I-frames occur, and optionally the confidence factor C. The MPEG AM is an artifact meter that has an output for indicating the position of an intraframe (I-frame). The data coming from the artifact meter are used to control the local adaptive filter in the recursive noise reduction loop RNRL (or more generally any device and method in which the noise reduction is made to run in sync with the GOP frames with a maximum on the intraframes) which is schematically indicated in the dotted rectangle. This loop RNRL forms a noise reductor to reduce the noise in the GOP frames.

Fig. 7 shows the diagram of spatial noise reduction. Spatial noise reduction is like temporal noise reduction based on a local adaptive filter too. But in stead of comparing a pixel with a pixel from a previous (recursive) frame a pixel is compared to other pixels in its direct neighborhood. This is done in a spatial correlation local adaptive filter SCLAF The trick of the trade is to adjust the local adaptive filter such that the filtering is strong enough to reduce noise and MPEG, but does not wash out too many details in the image.

Both temporal and spatial noise reduction have means of global (= per total image) strength control. The global strength control sets a filter strength tendency to the local adaptive filters in both cases.

As mentioned before the embodiment provides for an MPEG Fig. Mf and a confidence factor C. Conventional methods only have a filtered MPEG Fig. to set temporally averaged filter strength to the noise reduction functions. It is remarked that motion can be used to determine a dynamic ratio between temporal noise reduction for still images, and spatial noise reduction for moving images.

Studies of MPEG meters in temporal domain show a clearly distinctive response of the meter output of moving and still images. After an I-frame peak the down slope of the meter values in time decreases clearly faster on moving images then on still images. If there is a lot of motion, the MPEG artifacts decay faster in time after an I-frame (due to more interpolations on sub-block shifted image data). So the down slope is steeper. This allows for a very elementary motion detector just indicating the amount of motion.

Advanced high end market digital set-top boxes include comparable image processing algorithms as a television. When this invention is applied in such a device then the cadence does not need to be 'looked for'. The location of the I-frames is given. So the confidence factor C is maximal and does not need to be determined.

The invention is also embodied in any device using the method or system in accordance with the invention, such devices including, without being restricted to television receivers, advanced digital content players and receivers/decoders. The invention can also be used in or for software/GPU video render applications and algorithms on any computer platform. All of these devices received data streams comprising intra and inter- frames. The problem of a cadence in the artifacts occurs. For all such devices the invention is useful.

Fig. 8 schematically show a device. The device has a receiver for receiving a signal (signal in) having intra and intercoded frames. The device comprises a system (which could be in hardware or in software or in a combination of hard and software) for applying a noise reduction (wherein "noise-reduction" may be noise-prevention) that is in cadence with the intraframes and has maximum at the intraframes. The noise reduced output signal is sent or internally transferred to a display part for displaying the images on a display D.

In short the invention can be described by:

To a video signal comprising groups (GOP) of intra (I-frame) and intercoded frames (P-frame, B-frame) a noise reduction is applied. The strength of the noise reduction strength has a modulation in time with a period substantially equal to the period (Ti) between intra coded frames. The noise reduction strength has a maximum substantially coinciding with the position of intra coded frames.

The invention is also embodied in any computer program comprising program code means for performing a method in accordance with the invention when said program is run on a computer as well as in any computer program product comprising program code means stored on a computer readable medium for performing a method in accordance with the invention. Computer program products can for instance include a graphic processor for a game-console.

It is remarked that, within the concept of the invention redactor and reduction loop calculator etc. should be broadly understood to comprise e.g. any piece of hard- ware (such as a redactor or reduction loop, etc.), any circuit or sub-circuit designed for performing or aiding in a reduction or in a loop etc. as described as well as any piece of soft-ware (computer program or sub program or set of computer programs, or program code(s)) designed or programmed to perform an action in accordance with the invention as well as any combination of pieces of hardware and software acting as such, without being restricted to the below given exemplary embodiments. In embodiments more than more than one element may be combined into one piece of hardware or software. One piece of software or hard ware may also perform more than one function.

It will be appreciated by persons skilled in the art that the present invention is not limited by what has been particularly shown and described hereinabove. The invention resides in each and every novel characteristic feature and each and every combination of characteristic features. Reference numerals in the claims do not limit their protective scope. Use of the verb "to comprise" and its conjugations does not exclude the presence of elements other than those stated in the claims. Use of the article "a" or "an" preceding an element does not exclude the presence of a plurality of such elements.

The present invention has been described in terms of specific embodiments, which are illustrative of the invention and not to be construed as limiting. The invention may be implemented in a method and a device, in hardware, firmware or software, or in a combination of them. Other embodiments are within the scope of the following claims.

Claims

CLAIMS:

1. A method for noise reduction in a video signal comprising groups (GOP) of intra (I-frame) and intercoded frames (P-frame, B-frame) wherein to the frames of the video sequence a noise reduction is applied wherein the noise reduction has a noise reduction strength (NR) and the noise reduction strength has a modulation in time with a period substantially equal to the period (Ti) between intra coded frames, wherein a maximum in the modulation substantially coincides with an intra coded frame.

2. Method as claimed in claim 1 wherein the modulation depth (D) is a function of the average intensity of the image.

3. Method as claimed in claim 2 wherein the modulation depth (D) increases, at least for a range of average image intensities, as the image intensity decreases.

4. Method as claimed in any of the preceding claims wherein a confidence factor (C) is calculated which confidence factor C influences the modulation depth (D).

5. Method as claimed in any of the preceding claims, wherein the noise is reduced by a temporal recursive noise reduction.

6. Method as claimed in any of the preceding claims wherein the noise is reduced by a recursive spatial noise reduction.

7. Method as claimed in any of the preceding claims wherein an artifact measurement (MPEG AM) is performed and from the artifact measurement the position of an intracoded frame is determined.

8. Computer program comprising program code means for performing a method as claimed in any one of claims 1 to 7 when said program is run on a computer.

9. Computer program product comprising program code means stored on a computer readable medium for performing a method as claimed in any one of claims 1 to 7 when said program is run on a computer.

10. System for noise reduction in a video signal comprising groups (GOP) of intra (I-frame) and intercoded frames (P-frame, B-frame) wherein the system comprises a noise reductor (SCLAF, RNRL) to apply to the frames of the video sequence a noise reduction wherein the noise reductor (SCLAF, RNRL) is arranged such that the noise reduction has a noise reduction strength (NR) and the noise reduction strength has a modulation (D) in time with a period substantially equal to the period between intra coded frames (I-frames), wherein a maximum in the modulation substantially coincides with a intra coded frame.

11. System as claimed in claim 10, wherein the system comprises a temporal recursive noise reduction loop (RNRL)

12. System as claimed in claim 10, wherein the system comprises a spatial recursive noise reduction loop (SCLAF).

13. System as claimed in any of the claims 10 to 12 wherein the system comprises an artifact meter (MPEG AM) and the artifact meter has an output for indicating the position of an intracoded frame (I-frame).

14. Device comprising a system as claimed in any of the claims 10 to 13.

15. Device as claimed in claim 14 wherein the device is a television receiver, receiver/decoder or content player.