MXPA97002867A - Digital vcr with hart truck reproduction current derivation - Google Patents

Abstract

The present invention relates to a digital video cassette recorder (210), consumer, can record a television signal in advance (09) having a signal format similar to MPEG. The predictive nature of the similar MPEG signal format requires that additional I frame data be generated and recorded together with a normal playback speed data stream (10) to facilitate non-normal speed, or trick-play broadcast. Additional frame I data streams (121, 131, 141) are generated specifically for each playback speed and written into recorded tapes to facilitate playback at predetermined speeds. Various methods of the invention are described for the derivation of streams of trick-play data of full resolution and reduced resolution. The trick play reproduction data stream of the invention is described for recording in real time by the consumer apparatus and generating normal and non-real time trick play data stream for use with pre-recorded digital media

Description

DIGITAL VCR WITH HDTV TRUCKED REPRODUCTION CURRENT DERIVATION This invention relates to the field of digital video recording and in particular, to the derivation, recording and reproduction of similar advanced MPEG television signals at non-normal speeds. BACKGROUND OF THE INVENTION A standardization committee has proposed a digital video cassette recorder employing a helical scan format. The proposed standard specifies digital recording of normal definition television signals, for example, NTSC or PAL and high definition television signals that have a compatible MPEG structure such as a proposed "Grand Alliance" or GA signal. The DN recorder uses a compressed component video signal format that employs intra-field / frame DCT, with adaptive quantization and variable length coding. The digital VCR or DVCR of DN can digitally record NTSC or PAL television signals and has sufficient capacity to record data to record an advanced television signal. A specification of the GA signal is included in a project specification document entitled "Grand Alliance" HDTV System Specification, published in "1994 Proceeding" of the "48th Annual Broadcast Engineering Conference Proceedings", March 20-24 1994. GA signal employs a compatible MPEG encoding method that uses a coded image between the frame, named frame I, an advanced predicted frame, named frame P and a bidirectionally provided frame, named a frame B. These three frame types are presented in groups known as GDI or Image Groups. The number of frames in a GDI is defined by the user but may include, for example, 15 frames. Each GDI contains a frame I, which can be adjacent to two frames, which are followed by a frame P. In a VCR, "Trucking" or analogue RT aspect for the consumer, such as a forward or backward image. backward, fast or slow movement, are easily achieved, since each recorded band usually contains a television field. Therefore, the reproduction at speeds different from the normal ones, can give like result the reproduction of head or heads, the crossing of multiple bands and recovery of recognizable segments of images. The segments of images can have boundaries and provide a recognizable and useful image. An advanced television or similar MPEG signal may comprise groups of GDI images. The GDI, for example, can comprise 15 frames and each frame can be registered by occupying multiple bands on the tape. For example, if 10 bands are distributed in each frame, then a GDI of 15 frames will comprise 150 bands. During the playback speed operation, data is retrieved from frame I, which allows the decoding and reconstruction of the predicted frames P and B. However, when operating a DVCR at an abnormal reproduction rate, the response heads transduce sections or segments from the multiple bands. Unfortunately, these DVCR bands no longer represent discrete records of consecutive image fields. However, since the predicted frames P and B require preceding data facilitate decoding, the possibility of reconstructing any useful frames of the reproduced pieces of data is greatly diminished. In addition to that the MPEG data stream is particularly inexorable for missing or unintelligible data. Therefore, to provide "Trick Play" or non-normal speed response characteristics requires that specific data be recorded, which, when played back in an RT mode, is capable of reconstructing the image without the use of frame information. adjacent or preceding. The specific data or "Trick Play" data must be semantically correct to allow decoding of MPEG. In addition, a selection of "Trick Play" speeds, it may require different derivation of RT data and may require places of bands recorded at RT specific speed. To be able to reconstruct without data of previous frames, it is required that the specific data of the "Trick Play" are derived from the frames I. The specific data of the "Trick Play" must be corrected syntactically and semantically to allow the decoding, for example, by a compatible GA or MPEG decoder. In addition, the "Trick Play" or RT data should be inserted into the MPEG-like data stream to be recorded along with the normal MPEG-like signal. This compartment of the data capacity of engraving channels, can impose impediments in terms of RT character data regime that can be provided within the available tracking capability. The data character regime of RT can be used or shared in a varied manner between the resolution in space and / or time in the derived or reconstructed RT image. The reproduced "Trick Play" image quality can be determined by the complexity of the RT data derivation. For example, a DVCR consumer must derive RT data during recording, essentially in real time and only with additional data processing expense added to the cost of DVCR. Therefore, the real-time consumer DVCR "Trick Play" image quality may seem inferior to the RT image data derived by the non-real time image processing using sophisticated digital image processing. With non-real time RT image processing, for example, an edited program can be processed, possibly on a scene-by-scene basis, possibly at non-real-time playback speeds, to allow the use of processing techniques of sophisticated digital images. Such non-real-time processing can inherently provide images of "Trick Play" of superior quality that can be obtained with real-time processing.

COMPENDIUM OF THE INVENTION A method for generating a representative MPEG compatible digital image signal, which when recorded facilitates reproduction at no more than one speed. The method comprises the steps of: receiving a data stream comprising a signal representative of MPEG compatible digital image; decoding the data stream to extract intra-coded data; storing specific coefficients extracted from the intracoded data to form an encoded frame of reduced character regime; periodically selecting the intracoded frame of reduced character regime to form a specific character stream for a trick-play speed; selecting between the character stream specific to the trick-play speed and the data stream to produce a stream of formatted register characters; and record the stream of formatted register characters. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a simplified block diagram of an inventive system for the real-time generation of a "trick-play" data stream having low resolution. Figure 2 shows a simplified block diagram of an additional inventive system for the real-time generation of a full-resolution "trick-play" data stream.

Figure 3 shows a simplified block diagram illustrating an inventive method for generating low resolution "trick-play" data streams for inclusion in pre-recorded digital records. Figure 4 shows a simplified block diagram illustrating an inventive method further to generate "trick-play" data streams for inclusion in the pre-recorded digital registers. Figure 5 illustrates the derivation of CD coefficients from predicted macroblocks. Figure 6 shows a simplified partial block diagram illustrating an additional inventive method for non-real time generation of pre-recorded records Figure 7 shows a simplified partial block diagram illustrating another inventive method for non-real time generation of pre-recorded records DETAILED DESCRIPTION In a digital video cassette recorder for consumers, considerations main in the generation The real time of a trick-play stream is the complexity and cost of processing required and the need to keep this cost at a reasonable level. For this reason, the processing used in the generation of a stream of trick-play data in real time is You can limit by extracting pieces from the existing character stream and implementing minor modifications to the character stream parameters. Streams of "trick-play" data can be produced in real time by extracting pieces of intra-information independent of the original data stream. This intra-information can come from intra-frames, intra-modules, and / or intra-macroblocks. The source selected for data derivation of frame I, depends on the form of non-renewal used in the original stream and for illustrative purposes, it is assumed that any method of renewal is used either intra-frames or intra-modules.

In a first inventive method of real time generation, a "Trick Play" data stream is derived from resolution in the low space. The trick play stream of resolution in the low space can, for example, have resolution in accordance with the CCIR 601 standard, (720 x 480 pixels), regardless of the original HDTV current resolution. Since the regime of available effective characters for trick-play streams is limited to nominally 2M characters / second, using resolution in the low space, this results in fewer characters used per frame, and therefore, can be achieved relatively high temporal resolution. However, this resolution in the low space can only be practical if an advanced decoder and television screen are capable of such resolution.

In a second inventive method, a trick-play stream having the same resolution, or pixel count, as the original HDTV material is generated. However, since the scheme of useful trick play characters is limited by the engraving channel capacity of nominally 2 M. characters / second, there is a compromise between the resolution in space and time. Therefore, the effective provision of a "Trick Play" mode of resolution in the entire space requires that the temporal resolution be reduced to remain in proportion to the capacity of the RT data channel. The first inventive method for real-time generation of "Trick Play" resolution data in the low space is illustrated in Figure 1. In this illustrative block diagram, trick-play speeds of 5x, 18x and 35x are generated. For each RT speed, low-resolution intra-coded frames are constructed from a transport stream similar to MPEG received. By detecting the information of MPEG headers of transport stream below the level of the module, one can extract, process and use intramodules, to create a single frame I in the memory 110. The extraction and processing stage 100 performs three tasks; extracting macroblocks for the construction of an RT framework I, recodes the CD transformation coefficients when DPCM coding is needed and discards unwanted CA transformation coefficients, when necessary. Having constructed and stored an I-frame of resolution RT in the memory 10, it is used in the generation of specific velocity data streams for each trick-play speed. A modulated radio frequency carrier, which responds to a compatible MPEG signal, is received by the receiver 05. The modulated vehicle can come from an antenna or a cable, not shown. The receiver 05 demodulates and processes the received carrier to produce an MPEG compatible advanced television transport stream 09. The advanced television transport stream 09 is demultiplexed in block 20 to obtain only the Packed Elementary Stream or CEP stream corresponding to the advanced television video information. The PEG stream is decoded in block 30 to extract the payload of MPEG encoded video stream. Having extracted the MPEG encoded stream, the required intra-coded information can be detected and extracted. The sequence detection block 40, examines the character stream for the presentation of a starting code characterized by twenty-five 0's followed by 1, followed by an MPEG video header indicating the 8-character address. The image detection is carried out in the block 50 and in the block module 60 the module layers 60 are detected. Since the decoded "trick-play" frame I is to be constructed, it is only extracted between modules. The ntramodules only contain intra-coded macroblocks and are characterized by an intra-module indicator in the module header. Therefore, when an intra-module indicator is set to 1, the entire module is passed to step 100 of "data extraction and processing". The undetection process of block 70 assumes that both intra-module or intra-module renewal techniques are employed and also that the intra-module indicator in the module header is set when appropriate. If the intramodule indicator is not fixed or if intra-macroblock renewal is used, then an additional level of detection below the macroblock level is required. The data extraction and processing step 100 is selected from the intra-coded macroblocks extracted in block 70, only intrainformation which is used to construct various trick-play data streams. In addition, block 100 performs any processing that may be necessary to ensure syntactic and semantic corrections for MPEG compatibility of the resulting reconstructed RT frame I. Since the reconstructed RT frame I is of resolution in the lower space than the original MPEG stream, only a subgroup of intra macroblocks detected is required. To determine which macroblocks or MB are to be maintained and which are to be discarded, you can use either a mathematical unction and a predefined lookup table. The resolution frame in the resulting lower space results from the work with various parts selected from the macroblocks. A stage of the controller 90 is coupled to the processing step 10 and provides either the calculation required by the mathematical function or provides the lookup table to determine the selection of macroblocks The relationship between the position of MB in the new low resolution frame I , (mb (?, j),? = 0, 1, 2, n-1, j = 0, 1, 2, m-1, where m and n are the new width and height of the frame I in MB respectively eiyj se refer to the row and column of MB) and the original full resolution frame ((MB (I, J), l = 0, 1, 2, N-1, J = 0, 1, 2, M-1, in where M and N are the original width and height of the frame and I and J are the row and column of MB), the relationship is given by i (row of low resolution) = [I (nl) / (N-1)] j (low resolution column) = [J (m-1) / (M-1)] where the product of the square brackets [x] denotes the value of the nearest whole number ax The low resolution RT frame I uses a subset of the macroblocks of the original frame being discarded two remaining unselected MBs Figure 5 illustrates an illustrated 420 signal comprising three intra-coded macroblocks MB1, MB2 and MB3, each comprising blocks 0, 1, 2, 3, 4, and 5 The macroblock 2 is crossed to illustrate the lack of use in the construction of the I frame of RT of reduced resolution The coefficients of CD of each block of illumination and chrominance are described in Figure 5 with dark strips The coefficients of CD are foreseen within each macroblock, with the CD coefficient of the first block of a MB being provided with the last CD coefficient of the immediately preceding MB of the module. The arrows in Figure 5 illustrate the prediction sequence. Therefore, if the preceding MB is not selected, for example, MB 2 of Figure 5, certain CD coefficients of the newly attached macroblock must be recalculated, as described by the "NES" arrows of Figure 5, and they are recoded using DPCM. This recoding process is carried out as the macroblocks are written to the frame memory I, 110. If the HDTV video stream originated from an interleaved scan source, an optional processing step may be included to remove the "flicker" interspersed exhibited by frozen interleaved fields containing movement. If the temporal resolution of the reconstructed trick-play stream is such that the same frame (two fields) is displayed for more than one frame period, then that interspersed "flicker" can be very noticeable. In the coded macroblocks of capo, this "flickering" artifact can be eliminated by copying the two upper blocks of the macroblock, blocks 0 and 1, to the two lower blocks, blocks 2 and 3. This copying within the macroblock, effectively forms both fields, thus removing the same any movement from field to field of the frame. This recoding process is performed as the macroblocks are written to the frame memory I 110.

An additional function performed by the processing step 100, is the removal of AC coefficients from each macroblock that can not be accommodated in the newly constructed RT frame I due to the low rate of characters available for the trick play streams. To achieve this, each block is decoded of variable length to the point where the block will be filled with zeros, indicating the last coefficient of that block. The number of characters for each block is stored and accumulated in a compensator. Characters are counted, and when an account exceeds a predetermined number, the remaining A coefficients are disabled or deleted. The number of characters per MB of RT depends on the global regime allowed for each trick-play stream and the resolution or temporary number of frame updates per second. The block diagram of Figure 1 illustrates the formation of trick play data streams having the same set of placed characters. If the regime differs significantly between RT speeds, for example, to provide deferred resolution at each speed, the number of AC coefficients retained in the frame memory I 110 will also differ for each speed. Therefore, the frame memory I 110 can not be shared and separate memories of the frame I can be required for each RT rate or character regime.

The low resolution RT frame I of the invention, assembled in the memory 110 of the frame I, is coupled to three stages of trick play generation, 5 times, block 145, 18 times, block 160 and 35 times, block In Illustrative Figure 1, each trick-play stream can distribute the same character regime and temporal resolution, which could represent a preferred configuration. However, each reconstructed RT-frame I is not used for each RT rate. For example, if the renewal regime of frame I in the original stream is once every fifteen frames (M = 15) and the temporal resolution used for each trick-play stream is selected to be three, that is, the number of frame times between frame updates, then for 5 times speed, (speed 5x) (3 frames repeats) / (15 frame renewals) = 1 0 therefore each frame I of RT will be used similarly for speeds of 18x and 35x, (18) (3) / (15) = 36 (35) (3) / (15) = 7.0 Therefore at the speed of 18 x, approximately every third or fourth frame I is used, and at speed of 35x is used every seventh frame I If the period of intrarenovation is assumed in an advanced television stream of 0 5 seconds (M = 15 for 30 fps source) then, a maintenance time of three frames for 5x speed is the highest possible RT temporal resolution. For simplicity and consistency, a maintenance time of three frames can be used for the remaining RT speeds. A higher temporal resolution of two frames or a single frame maintenance time can be used for higher RT speeds since lower temporal resolution at higher speeds can give a false sense of decreasing the actual trick play speed. Assuming that the effective trick-play character set is constant, the provision of a higher temporal resolution could consequently require a resolution quality in the lower space. The reconstructed RT frame I is read into the memory 110 and packed, according to the RT rate, by blocks 145, 160 and 170, which add the appropriate MPEG image headers and a CEP layer. The advanced television transport stream 09 is regulated by a compensator 15, which generates the signal 10, a transport current for normal reproduction speed processing. The normal reproduction transport stream 10 is coupled to the MUX multiplexer 150. The MUX multiplexer 150 is controlled by responding to the servo signals of the recorder 210 to generate a stream of output characters having a sequence, which, when recorded, produces a predetermined band format. The registered band format is selected to provide the regimen of registered RT characters and to facilitate the specific physical distribution of specific speed RT frame I packages within specific recorded bands. The recorded band format therefore facilitates playback at normal speed and predetermined trick play speeds. The RT frame I packets, 5x signal 121, 18x signal 131, and 141c signal 35c, are coupled to the multiplexer 150 which inserts the frame packages I for each RT speed into the normal reproduction transport stream. Therefore, a valid transport stream, similar to MPEG, is formatted to record the processing by the recorder 210 and register on the tape 220. To minimize the character regime of RT, instead of frames I of RT Repeats, frame repetitions or maintenance times, can be implemented by writing empty frames P between frames I in the video stream. An empty P frame results in the prediction of the prior frame encoder, that is, the RT frame I. Alternatively, frame repeats can be implemented by setting the DSM-trick-mode-indicator in the CEP layer and calculating the MRT / MTD Decoding Time Mark and Presentation Time values so that each I frame of RT is present the necessary number of times in a separate frame. Each frame repeat method produces the same result. However, the second method does not require extra processing of the RT current in reading and, therefore, does not add extra cost to the unit. However, the second method requires that the optional DSM-trick-mode-indicator be supported in advanced television decoders. With this second method, extra processing is implemented in the advanced television decoder. Any frame repetition method may be implemented during the generation of specific velocity current in blocks 145, 160 and 170. The trick play current generation techniques of the invention, described above, were employed to produce trick-play velocities of 5x, 18x, and 35x with a resolution in the space of 720 x 480 pixels and an effective trick-play data rate of 2.0 Mbps. Different trick-play speeds were evaluated and can be summarized by the following points: Each trick-play was generated by representing low-resolution compatible MPEG transport streams (720 x 480 pixels). Each RT stream contains only intra-coded frames thus allowing the same trick-play stream to be used for both RT Fast Forward and Fast Backward modes. To retain an aspect ratio of 16: 9, the image size in real space is sampled at 720 x 384 pixels, the rest of the area being above and below the black part of the RT image.

The temporal resolution is such that a constant maintenance time of three frames is used resulting in an effective regimen of 10 frames per second. Each frame I of the trick play streams comprises a selection of macroblocks sampled from the original stream. The character regime of 2.0 M. characters / sec. and the maintenance time of three frames allows the majority of the AC coefficients to remain in the macroblocks selected for normal test material. The resolution in the subjective, global space is fair, depending on the amount of movement and image complexity in the original material. An image frame of 10 fps, provides good temporal resolution. The trick play data stream can be decoded to produce recognizable trick play video images and, therefore, is acceptable for tape search use. The low-resolution real-time trick-play mode of the invention, previously discussed, produces images in space recognizable at a relatively high temporal resolution. However, as already mentioned, this mode can be used if an advanced television receiver / decoder unit can be operated at lower resolution, for example, such as that produced by CCIR recommendation 601. However, if operation is not provided at a lower resolution, then the trick-play data must be derived having nominally the same resolution in the space, that is, the same pixel count as the original source. Figure 2 illustrates an inventive illustrative system for generating real-time, full-resolution trick-play streams. Three trick-play speeds of 5 times, 18 times and 35 times are illustrated. The difference between the full resolution scheme of Figure 2 and the low resolution scheme illustrated in Figure 1 is in the data extraction and processing block 105, and the current generating blocks 15, 165, and 175. The decoding of transport current and the intradetection described in blocks 20, 30, 40, 50, 60 and 70, operate and function as described for the low resolution system of RT, the purpose of block 105 of the extraction stage and data processing, is to extract only the intrainformation required to form trick-play streams and to perform any processing that is required to guarantee the syntactic and semantic corrections of the resulting RT frame I. The functionality of block 105 differs from that of block 100 in that the regenerated frame I must have the same resolution, or pixel count, as the original data stream. Therefore, all intramacroblocks are used to reconstruct the new RT framework I. Since the MBs are not deleted, no CD transformation coefficient is required. The main function of the processing block 105 is the removal of the AC coefficients of each macroblock, which, as a consequence of a scheme of trick play characters, can be distributed in the new RT frame I. The low RT channel character scheme, nominally 2 M characters / sec, forces a transaction between the number of the used AC coefficients, ie, resolution in space, and the temporal resolution, or update rate of frame of the trick-play stream and the temporal resolution, or update rate of the trick-play stream. This transaction of the space against the storm is also present in the derivation of the low resolution current. However, in a full resolution frame, that is, the same pixel count, the DC coefficients alone probably represent more characters than all the coefficients, both CA and CD assembled in a low resolution RT frame. Therefore, any limited inclusion of even few AC coefficients in each full-resolution macroblock, will produce a significant reduction in the temporal resolution, that is, the frame update time will be extended, with no more frame repetitions. Therefore, to facilitate constant temporal resolution in streams of full resolution trick representation, a system can only use the CD coefficients of each macroblock, all the AC coefficients being discarded. Furthermore, the discarding of the AC coefficients reduces the processing complexity since only the variable length decoding of the DPCM value of the CD coefficient is required. Figure 2 illustrates an illustrative system in which each trick play speed has the same character regime, and therefore, the same frame memory I can be shared among the three RT speeds. As previously discussed, if original HDTV video images were generated by interleaved scanning, then an optional processing step may be included to remove interspersed "blinking" exhibited by frozen fields containing motion. One such method has already been described. However, since this illustrative high-resolution RT system only transforms CD coefficients, a simpler and more efficient method can be provided by setting the frame_pred_marker_dct flag in the extension_coding_counter image to "1". This indicator shows that all the MB were coded by the frame, therefore, a block previously coded by field, which could produce flicker ', is decoded as a block coded in the frame. The result is that each field is placed in the upper or lower portion of a block and any 'blinking' is removed. This flicker elimination method also reduces the number of characters used in the macroblock-mode section since the dct_type flag can no longer be present if the frame_pred_marco_dct_ is set to '1' The reconstructed RT frame I is assembled in memory 115. and is coupled to three stages of trick play generation, 5 times speed described in block 155, 18 times speed in block 165 and 35 times speed in block 175. The illustrative system of the Figure 2 it is assumed that in each trick-play stream it has the same effective character regime and therefore the same approximate temporal resolution, as discussed previously, each frame I of RT reconstructed for each velocity is not used. However, the use of frame I of RT may be further limited for the following reason. Although each frame I of RT has the same number of coefficients, for example, only CD, each frame I of RT can not have the same number of characters since the CD coefficients are encoded with variable length. Therefore, a constant temporal resolution or frame that retains time, can not be fixed for each trick-play stream. Instead of the frame retaining time varies slightly after time with the number of characters required to code or form each I frame of RT. For each trick-play speed, the respective "stream generation" steps, 155, 165 and 175, wait until enough characters have accumulated in the compensator 105 to encode a frame I of RT. Then if the RT frame I accumulated in the compensator at the time there is a new RT frame I, ie one that has not been encoded in the specific trick-play speed, the RT frame I is encoded and the number of used characters will be subtracted from those available. If each frame I was of the same size and each trick-play speed was distributed to the same effective character scheme, this scheme could be equivalent to that described for the low resolution system and the frame renewal period could be constant for all speeds . The reconstructed RT racks I are read from the memory 115 and packaged by the current generators 155, 165 and 175 to form a compatible MPEG of transport streams in exactly the same way as detailed for the low resolution system. The technique of generating streams of resolution truncated representation in the entire space of the invention, written before, was evaluated at an effective trick-play representation rate of 2.0 Mbps, for trick-play speeds of 5x, 18x and 35x. The performance should be summarized as follows: An MPEG compatible transport stream of only an independent RT frame I can record each trick-play speed. The temporal resolution varies with the scene complexity and is lower, having frame maintenance times longer than the low resolution trick play system in the previously described space. The average and variation in maintenance times experienced for the normal source material are shown in the following table.

Note: Because an identical effective trick play pattern is used for all speeds, the temporal resolution will always be similar (if not identical) for each speed. Each frame I of RT uses only CD coefficients. The overall quality of resolution in space is only fair since only CD coefficients are used. The quality of temporal resolution can vary between poor and fair, depending on the level of complexity within the material encoded by RT. However, the resulting trick-play images will be. You can recognize and accept for the use of tape search. The main differences between current derivation of real-time trick-play data and pre-recorded trick play result from cost concerns and lack of complexity imposed on a consumer recorder / player. The consumer unit must derive and record the trick-play data stream while recording normal playback data, ie, the trick-play data stream is derived in real time. With pre-recorded material, the trick-play data streams can be derived directly from a source of original images instead of a compressed MPEG encoded stream. The data streams of RT of specific velocity can be derived independently from each other and independently of the actual engraving event. Therefore, pre-recorded trick-play data can be derived in non-real time, possibly in non-normal or slower frame repetition rates. Since the concerns of the consumer's real-time method are no longer applied, the quality of trick-play broadcast achieved by the pre-recorded material can be significantly higher. A first method of the invention of the derivation of pre-recorded RT data, provides a resolution in the space of, for example, CCIR Rec. 601 which has a resolution of 720 x 480 pixels, without taking into account the HDTV current resolution. original. A second method of the invention involves a trick-play stream of the same resolution, i.e. pixel count, as the original HDTV material. Figure 3 illustrates an illustrative block diagram showing a method of the invention for generating pre-recorded trick play data streams. Without taking into account the format of the original HDTV video material 09, the temporary processing block 30, performs temporary sub-sampling which produces a progressive signal 31 of 30 Hz. The operation of this stage may differ depending on whether the material of the The original source is progressive with a frame rate of 29.97 / 30 Hz. With source material progressively scanned, the frame rate can be reduced by dropping the sequence every second frame. Falling from alternate frames a progressive sequence results in having half the temporal resolution of the original source material. With the source material interspersed, the frame rate remains the same but only one field of each frame is used. This process results in a progressive sequence of half the vertical resolution and the same frame rate. The frames scanned progressively, the signal 31 is coupled to the block 40, which generates a lower resolution signal having, for example, the resolution delivered by CCIR Rec. 601. Each frame scanned progressively is resampled at 720 x 480 pixels to retain the aspect ratio of 16: 9, and the upper and lower edges are filled with black to produce a 'letterbox' format of 720 x 480 pixels. The HDTV signal is now represented by the signal 41, having a resolution in the lower space of 720 x 480 pixels, progressively scanned with a frame rate of 30 Hz. The signal 41 is coupled to the blocks 50, 60, 70 which they implement the temporal subsampling that depends on the speed. Each trick-play stream is constructed to have the same temporal resolution or frame that holds the time of 2 frames, that is, each frame will be repeated once. Therefore, at trick-play speed of N times, the frame rate is reduced from 30 Hz to 30 / 2N Hz Therefore, the resulting registered frame rates are as follows, 5x becomes 30/10 Hz, 18x it becomes 30/36 Hz and 35x becomes 30/70 Hz Since each frame is presented twice and the display rate is 30 Hz, the effective speed of the scene content remains correct at each speed The blocks of temporary sub-sampling 50 , 60, 70, general output character streams 51, 61 and 71, respectively, which are coupled to the respective MPEG encoders 120, 130, and 140 to the character streams compatible with the MPEG format. of MPEG is the same for each speed, and since in real-time processing of pre-recorded environment it is not necessary to use the same MPEG encoding hardware to encode the normal representation current and each repressive current. Tricked Situation This community of use is indicated by the dotted line enclosing the blocks of the MPEG encoder 100, 120, 130, and 140 The temporarily sub-sampled character streams 51, 61 and 71 are encoded by MPEG as frames I Each frame I is repeats once using the DSM_truco_representac? ón_? nd? erdor, located in the CEP layer as previously described The resulting compatible MPEG currents representing the current 101 NP of normal playback speed, and the current speeds 121 of the 5x trick play, 18x stream 131 and 35x stream 141, are coupled for registration format by the multiplexer 150 The multiplexer 150 effectively selects between the different MPEG streams to generate a format signal of the synchronization block 200, suitable for record the processing by the record reproduction system 210 and writing the tape 220. As described above, the use of the predetermined RT speeds allows the specific rate RT data to be placed, or recorded, at specific locations in the sync block within the recorded bands. This multiplexer 150 formats the signal 200 of the synchronization block to locate data of the specific speed RT frame I at specific locations of the synchronization block within the recorded bands. These specific locations facilitate reproduction at various specific speeds of RT. Figure 6 is a partial block diagram illustrating a further arrangement of the invention of the non-real time "trick play" apparatus of Figure 3. The RT signals 51, 61 and 71 processed at specific speeds, are coupled to the memories 520, 530 and 540 which store the digital image signals processed 5 times, 18 times and 35 times, respectively. The original HDTV signal 09, is also stored in the memory 500. The production of the prerecorded media or tape is facilitated by the sequential selection between the different digital signal sources stored to form an output signal that is encoded by MPEG via the encoder 100 and recorded in the medium. A multiplexer 150 is controlled to select among the different sources of digital signals to form an output signal for MPEG encoding. The MPEG encoded signal 200 has the different signal components arranged so that a register can be reproduced at normal and trick-play rates. Therefore, the arrangement of the invention of Figure 6 facilitates the derivation of non-real time and independent of the sources of digital signals of normal reproduction and trick play to code them as MPEG compatible character streams. Figure 7 is a partial block diagram illustrating another inventive setup of the trick-play apparatus of non-real time of Figure 3. In Figure 7, the digital signals 09, 51, 61 and 71 processed from normal reproduction and trick play, are coupled to be encoded as compatible character streams by the encoder 100. With non-real time signal processing and pre-recorded material preparation, the signals 09, 51, 61 and 71, can be derived separately and individually coupled for MPEG coding by a single encoder 100. The individually encoded MPEG character streams 101, 121 131 and 141 are stored in memories 550, 560, 570 and 580 representing streams of normal playback characters and 5x, 18x and 35x, respectively. The memories 550, 560, 570 and 580 produce output signals 501, 521, 531 and 541 which are coupled to the multiplexer 150 which responds in a controlled manner to the recorder 210 to generate a registered recorded character stream of MPEG compatible in a way that provides playback at normal playback speed and at predetermined "trick play" speeds. The illustrative system of low resolution RT in space, illustrated in Figure 3, and described above, produces quality of truncated reproduction of significantly higher quality than that obtained from trickle current reproduction currents derived real time. The results produced can be summarized as follows. During recording, a low resolution MPEG compatible stream (720 x 480 pixels) of a single, independent I frame is written to the tape for each trick-play speed. The size of the image in real space is 720 x 384 pixels, to retain aspect ratio of 16: 9, presented in a "letterbox" format. The temporal resolution is effectively 15 frames / second for each trick-play speed and produces quality from good to excellent that remains constant for each speed. The resolution in space produced by a data regime of 2.0 Mbps and resolution of 720 x 480 pixels is good to very good, depending on the complexity of the original material. In general, the image quality of trick play displayed with this scheme is very high.

The low-resolution pre-recorded trick play system, shown in Figure 3 and described above, produces good quality space images at a relatively high temporal resolution. However, said low resolution method can be used by providing that the advanced television decoder / receiver unit can support the lower resolution display format. Figure 4 is an illustrative block diagram of a full resolution, pre-recorded trick play current generating system of the invention, which provides trick-play speeds of 5x, 18x and 35x. As previously discussed, the derivation of pre-recorded trick play data stream can be generated from the original, uncompressed, original material. Figure 4 illustrates the generation of normal playback and trick-play character streams, however, these can be generated independently of one another directly from the original HDTV material. Since this system provides full resolution, no sub-sampling is required in the space and therefore less processing is required than shown in Figure 3. Since the original or compressed, original material can be used, the frames that are intra-coded will be they can choose with accuracy to adapt the speed of the trick representation, instead of selecting the frames I of an encoded stream. In addition, a constant temporary renewal regime can be maintained, which is more pleasant for the user. The original HDTV video signal 09 is shown coupled to the MPEG encoder 100 which generates an MPEG 101 stream for normal playback speed operation. The signal 09 is also coupled for temporary sub-sampling in blocks 55, 65 and 75 respectively. For a trick-play speed of times N, to encode, only each font frame source Nth can be used. However, depending on a desired transaction between the resolution in space and the temporal, the actual frames used for coding may be closer to each 5Nth or 8Nth frame in order to provide a resolution in the acceptable space. Therefore, frame maintenance times, or temporal resolution, are similar to those of the full real-time resolution system described above. Having selected a frame that maintains or updates time, for example each frame 5Nth for each trick-play speed of N times, the signal 09 of the HDTV stream is temporarily sub-sampled for each RT rate. The 5 times RT current is derived in block 5 which temporarily subsamples by a factor of 1 / 5N or 1/25, ie, 1 frame in 25, is selected to generate output signal 56. Similarly, the RT current of 18 times, is derived in block 65, which temporarily subsamples by a factor of 1 / 5N, or 1/90 and generates the output signal 66. The RT current of 35 times, is derived in block 75, which sub-samples temporarily by a factor of 1 / 5N or 1/175 and generates the output signal 76. The three subsampled RT character stream signals, 56, 66, and 76, are coupled to encode MPRG in the encoder blocks 120, 130 and 140 respectively. Since the MPEG-compatible encoding is the same for each speed and because real-time processing is not necessary in a pre-recorded environment, the same MPEG encoding hardware can be used to encode the normal playback stream and each trick play stream This community of use is indicated by the dotted line enclosing the blocks of the MPEG encoder 100, 120, 130 and 140. The temporarily sub-sampled character streams 56, 66 and 76 are encoded by MPEG as I. frames. Frame update is constant through each trick-play stream, it is also the number of characters distributed for each frame I. Frame maintenance times or repeats of frame I can be implemented using the DSM_project_project_indicator as previously described. The resulting MPEG transport streams representing normal rate of reproduction NP 101, and current trick play speeds 121, 5x, stream 131 18x, and stream 141, 35x, are coupled to record by formatting by multiplexer 150. Multiplexer 15 effectively selects between different MPEG streams to generate a format signal of synchronization block 200, suitable for processing records by the record reproduction system 210 and writing to tape 220. As previously described, the predetermined RT speeds allow the specific rate RT data to be placed, or record in specific places within recorded bands. This multiplexer 150 formats the synchronization block signal 200 to distribute the data of the specific rate RT frame I at the specific synchronization block locations that facilitate playback at the different specific RT rates. The inventive arrangements of Figures 6 and 7 can also be applied to the non-real time "trick play" generation array of Figure 4. As described, the arrays of Figures 6 and 7 can facilitate the independent derivation of the digital signals of normal reproduction and trick-play for the subsequent MPEG format and coding for the production of pre-recorded tapes or controlled video of the user requesting the service. The concern of retaining full resolution in space and time, results in a trick-play quality that is very similar to that achieved by the full-resolution real-time method. However, this pre-recorded method has an advantage that the maintenance time of the frame is constant. The described trick play stream generation technique provides trick-play speeds of 5x, 18x, and 35x, which have full resolution in space, and an effective trick play mode of 2.0 Mbps. Performance can be summarized from as follows: 'During recording, an MPEG stream of a single I frame, independent, is written to the tape for each trick-play speed. The resolution in space is equal to that of the original material. The temporary resolution is fixed having a maintenance time of 5 frames. Each frame I uses all the coefficients of CD and some of CA. The global quality in space is fair. Recovered trick-play images can be recognized and are acceptable for tape search purposes. The following table summarizes the quality of trick play achieved by the different methods of the invention described.

GENERATION OF CURRENT CURRENT GENERATION REPRODUCTION PLAYBACK TRUCK OF TIME TRIED OF TIME NO REAL REAL QUALITY MODES IN SPACE: QUALITY IN SPACE: REPRODUCTION from poor to fair, only from poor to just, only TRICKED from using CD coefficients. they use CD coefficients and COMPLETE RESOLUTION TEMPORARY QUALITY: some of CA. from poor to acceptable, TEMPORARY QUALITY: retention times from poor to acceptable, frames of 5-8, variable retention times frames of 5, constant QUALITY MODES IN SPACE: QUALITY IN SPACE: REPRODUCTION from poor to good, depends on poor to very good, TRUCKED of the material, work depends on the material. RESOLUTION LOW used parts of Mbs. TEMPORARY QUALITY: TEMPORARY QUALITY: very good, good time, retention time, retention of frames 2, frames 3, constant. constant.

In view of the concerns previously addressed, the quality of the highest-quality trick play can be achieved, in real-time and pre-recorded material, by using trick-play data of lower resolution . However, the advanced television receiver / decoder must support the use of a low resolution mode. If full-resolution trick play modes are used, the quality provided can be increased by manipulating several parameters. For example, raising the effective character rate for each trick-play speed will allow an increase in resolution. However, a minimum character rate of approximately 2.0 Mbps is required. If the number of "Trick Play" rates provided is small, for example two in each direction, then the effective character rate for each remaining speed can be increased. . The effective temporal resolution, or number of frame repeats, results from the transaction between the temporal and spatial resolution. Therefore, each parameter can be optimized depending on the desired application.

Claims (10)

1. CLAIMS 1. A method to generate a representative signal of MPEG compatible digital image, which when recorded, facilitates the reproduction at more than one speed, said method comprising the steps of. a) receiving a stream of data comprising a signal (09) representative of MPEG compatible digital image; b) decoding (20, 30, 40, 50, 60, 70) said data stream (09) to extract intra-code data (71); c) storing specific coefficients extracted from said intra-coded data (71) to form an intra-codified frame of reduced character regime (111); d) periodically selecting said intra-encoded frame (111) of reduced character regime to form a specific character stream (121, 131, 141) for a trick-play speed; e) selecting from said specified character stream for said trick play rate (121, 131, 141) and said data stream (10) to produce a register formatted character stream (200); and f) recording (210) said register formatted character stream (200). The method of claim 1, wherein said intra-coded data (71) comprises intra-coded macroblocks. The method of claim 1, further comprising a step of; select discrete cosine transform coefficients of CD and discard discrete cosine transform coefficients of CA from said intra-coded macroblocks. The method of claim 1, wherein said intracoded frame of reduced character regime (111) comprises discrete cosine transform coefficients of CD. The method of claim 1, wherein a periodicity of said periodic selection of said intra-encoded frame (111) of the reduced character regime is related to said trick-play speed. The method of claim 1, further comprising a step from; waiting to select said intracoded frame (111) of reduced character regime until a predetermined number of characters has accumulated to form said intracoded frame of reduced character regime (111). The method of claim 1, further comprising a step of; controlling (FMT CTRL) said sequential selection of said specific character stream for said trick-play speed (121, 131, 141) and said data stream (10) to facilitate the reproduction of said compatible MPEG character stream (200) in said trick-play speed. The method of claim 1, wherein said sequential selection responds in a controlled manner to a format control signal (FMT CTRL) including a control signal 9211) of a recorder (210) by recording said compatible character stream of MPEG (200). 9. The method of claim 1, comprising a further step of: setting a frame_predict_dct flag in the section of? Cod_machine? Extended? Array? To '1' of said compatible MPEG character stream to prevent flicker in the interleaved image material. 10. The method of claim 2, comprising an additional step of; copy blocks 0 and 1 to blocks 2 and 3 within macroblocks that are coded in the field.