JP4863326B2 - How to process the data stream - Google Patents

Description

  The present invention relates to the field of digital signal processing and, more particularly, to a method for assembling transport data packets used, for example, to transmit MPEG-type data encoded in a high-definition television system.

  U.S. Pat. No. 5,168,356 issued to Acampora et al. Describes a system for processing high-definition television (HDTV) signals according to MPEG variable length coding. Yes. MPEG is a standard encoding format determined by the International Organization for Standardization (ISO). The standard is “International Organization for Standardization” ISO / IEC DIS11172, moving image for digital storage media and related audio encoding method (Coding for Moving Pictures and Associated DigiDigit. 23, 1991, which is referred to herein to describe a general encoding format, and in a system outside of Acampola, codewords (codewords) contain data. Priority is given to represent high priority information and low priority information in the stream. The stream is sent to a transport processor, which packs the codeword data into the form of transport cells, each containing a header section and a packed data payload section, A high priority data stream and a low priority output data stream are provided.

  The main function of the transport processor is to pack the variable-length code word data supplied (issued) by the priority processor in the previous stage into a packed data word format. The accumulated packed word is called a data packet, and a transport header is added in front of it.

US Pat. No. 5,168,356

  The transport packet format facilitates resynchronization and signal recovery at the receiver. For example, if the header data is added even after the signal has been destroyed or broken due to interference in the transmission channel, the receiver can re-entry the data stream if the transmitted data is lost or corrupted. Points can be determined based on the header data. Data synchronization in the MPEG compliant decoder is also performed by an image group GOP (Group of Pictures) having a start point at the boundary of the packet. As will be seen later, a GOP is a series of one or more images or frames that allow random access to a sequence of encoded video bitstreams. ing. Resynchronization can also be done by responding to the picture start codeword of an intra (internal or intraframe) encoded I frame and by placing the picture start codeword at a packet boundary, for example in an MPEG standard compliant system Easy to do. If an apparatus according to the principle of the present invention is used, transport packets can be easily formed and processed.

  In accordance with the present invention, in a system that transmits variable length code words (eg, MPEG code words) within a data packet, an incomplete data packet having fewer than a prescribed number has no function. Packs "zero words" to create fixed length packets and delimits between packets in which a defined code word appears.

  In the preferred embodiment disclosed, a special code word, packet alignment flag (PAF), is inserted into the MPEG data stream to signal the start of a group of images (GOP). The PAF is placed immediately before the 'image start' code word for the intra-frame encoded “I” frame that starts the GOP. The PAF informs that an 'image start' code word is imminent and has been over one clock cycle, during which one “housekeeping” function may occur at the beginning of the next packet. Performed before the image start 'sign word appears. These housekeeping functions include, for example, resetting the accumulator, checking the status of the header, and setting the 'last word indicator' for the data packet being built when the PAF appears. Is to occur. Since the GOP is supposed to start at the packet boundary, when a PAF appears, the data packet being built is terminated. Such a termination may result in a shortened packet containing stuffed words, less than the specified number. This shortened data packet is filled with blank (zeroed bit) words to form a complete data packet with a specified number of words, and between the packets in which the image start code word appears Define the boundaries. Thirty 32-bit fixed-length words form one data packet, and this data packet is preceded by one 32-bit header.

  According to the present invention, transport packets can be easily formed and processed.

FIG. 2 is a partial block diagram of a video signal encoder comprising a data wordta word controller, a data packer and a data / header synthesizer according to the present invention. It is a block diagram which shows the detail of the word controller of FIG. 1, and a data packer. FIG. 2 is a circuit diagram showing details of the word controller and data packer of FIG. 1. It is a block diagram which shows the detail of the word controller of FIG. 1, and a data packer. 2B is a truth table for the operation of the word state controller shown in FIG. 2A. The details of the pack data assembly circuit are shown. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is explanatory drawing which shows the example of generation | occurrence | production of the last word. It is a figure which shows the detail of the data and header synthesizer which are shown by FIG. FIG. 18 is a state transition diagram regarding the operation of the state controller shown in FIG. 17. 1 is a block diagram of a high resolution television encoding system including an apparatus according to the present invention. FIG. 3 shows an image representation of a sequence of image fields / frames of an encoded video signal. FIG. 3 shows an image representation of a sequence of image fields / frames of an encoded video signal. FIG. 20 shows an image representation of data block generation formed by the encoding / compression device in the system of FIG. FIG. 20 illustrates a generalized image representation of a data format formed by an encoding / compression device in the system of FIG.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram of a transport processor data packer 12 and packed data word controller 10. As mentioned above, the primary function of the transport processor is to pack variable length codeword data into fixed length (eg 32 bit) data words. The 30 data words accumulated constitute a data packet, and finally a transport header is added before the data packet. As will be described later with respect to FIG. 19, such a transport processor is used in a system for processing compressed video signals in MPEG format. Other features of MPEG format and processing will be described later in connection with FIGS.

  The controller 10 monitors the cumulative value of the length data word in relation to the packet alignment flag PAF, and the 32-bit data word assembled from the stream of variable length codewords. The integrity (completion) of the 960-bit data packet is confirmed. The length data Length is a parallel 6-bit word that matches the length of the variable-length code word, and defines the length of the variable-length code word. The binary value of the length word Length indicates the number of bits matching the length of the variable length code word that actually represents the MPEG code word to be transported. Each variable-length codeword appears on a 32-bit wide bus and has a variable number of valid bits (1-32) representing an MPEG format code.

  The PAF is generated by the input processor 14 and is one codeword before the picture start codeword PS (Picture Start codeword) of the MPEG type “I” (intra-coded) frame at the start point of the group of pictures (Group of Pictures). Appears. The PAF is generated by detecting the presence of an I frame picture start codeword PS by a digital comparator. The unit of the input processor 14 also includes a signal delay circuit for processing the picture start codewords PS and PAF so that the PAF is at the clock period of the codeword immediately before the I frame picture start codeword PS. It is supposed to occur. In addition, the delay circuit ensures that the output signals supplied to the units of the packed word controller 10 and the data packer 12 have an appropriate time synchronization relationship.

  The word address Word Address is sent to the data packer 12, and the data packer 12 receives the variable length codeword VLC to be packed to ensure proper concatenation of the input variable length codewords. A word control signal Word Control is also sent to the packer 12 to indicate the presence of a short word, mark (display) the last word in the packet, and the proper alignment position with the associated transport header added So that a sequence of 30 packed data words is obtained. The controller 10 tracks and monitors the completeness (completion) of the packet by accumulating the binary value of the length word Length. Each length value represents the number of significant bits in the associated codeword. One packet is complete (completed) when 960 bits are accumulated. The starting point setting or initialization of the count value (count) is performed by the appearance of the PAF, and the PAF resets the internal accumulator in the controller 10.

  Data packer 12 receives variable length code words (VLC) via a 32-bit parallel data bus. The valid bits are packed into 32-bit words according to the supervisory control signal supplied from the controller 10. Also, concatenation is performed to accept the final MPEG bit serial transmission order. The packed data supplied from the unit of the data packer 12 is transferred to the input FIFO data buffer (data FIFO) 16 of the data and header synthesizer 15 at a variable word rate. Further, the synthesizer 15 receives the data write enable signal Data Write Enable from the data packer 12, and the data packer 12 enables the FIFO data buffer 16 in the synthesizer 15 and stores the valid data in the FIFO data buffer 16. To be written. The data packet is packet complete when such 30 word transfers are complete, unless the PAF forces a short packet. The last word indicator supplied by the packer 12 marks the 30th word in the normal packet in this example, or the last word in the packet that is shortened by the appearance of the PAF.

  As long as there are pack data words that can be processed, the pack data words are transferred to the data / header combiner 15. Similarly, as long as there is a transport header that can be processed, the transport header is sent from the header generator 18 to the input FIFO header buffer (header FIFO) 17 of the synthesizer 15. Information used by the header generator 18 to form the header is obtained from the input processor 14 and the word controller 10. The header write enable signal Header Write Enable indicates that there is a processable header, and enables the FIFO 17 so that the header is written into the FIFO 17. The combiner 15 adds a header before each packed data payload and sends the resulting transport packet or block to the output rate buffer as shown in FIG. Further, the synthesizer 15 supplies an output data ready Data Ready signal indicating that the packed data word or the transport header is in a transmission ready (sendable) state. The header indicator signal Header Indicator indicates the clock period when the header is sent out. This signal acts as a transport packet boundary marker so that subsequent operations, such as Forward Error Correction (FEC), are properly supplied to the transport cell.

  Each header contains information about the data in the data packet associated with that header. The header information provides data assembly and synchronization at the receiver. The header information includes information such as a service type (eg, audio, video, data), a frame type, a frame number, and a slice number, for example. This type of header and its processing is described in relation to an HDTV digital signal processing system using MPEG type signal coding in US Pat. No. 5,168,356 to Acampola et al.

  The data packet may contain less than 30 packed data words, ie 1 to 29 words, in this example. The PAF supplied by the input processor 14 appears immediately before the picture start codeword PS of the intra-coded I frame when the GOP starts, as will be described later in connection with FIGS. A new packet is always started by a picture start codeword PS for an intra-coded frame, and the PAF preceding the previous one indicates the end of the data packet and the start of the new packet. This packet alignment or alignment with the picture start codeword PS is useful for fast acquisition of the data stream at the receiver. If PAF occurs during formation of a fixed length word, a shortened data packet will be formed. The remaining bits in the packed word being assembled are filled with a plurality of "zero value bits" (1-31) in the data packer 12. Further, the remaining words in the data packet are similarly filled with a plurality of “zero value words” (1 to 29) in the synthesizer 15, and the size of the transport packet is maintained at a predetermined length. Such a “zero word fill” (satisfaction word) request is indicated by the occurrence of the last word indicator Last Word Indicator before 30 data words have been transferred to the combiner 15.

  It is important to properly identify the last word Last Word in the data packet. The final word Last Word confirms that the packet assembled using the associated transport header is properly recorded (accommodated). The last word Last Word indicates the existence of a packet filled in the boundary (that is, an intra-coded frame) of the MPEG screen group GOP. GOP boundaries are important for resynchronization, for example in a television receiver / decombiner after a channel change. Determining the final word Last Word is by no means trivial. For example, when the packet is complete (completed), and when complete, the status of the packet, such as whether or not data segmentation or data segmentation is being performed on the next packet The final word Last Word is determined according to the specific information. The final word has a state when the final word Last Word is a word formed during the current clock period and a state when the final word Last Word is a word to be formed during the next clock period. Exists.

  Some specific examples relating to the formation of the final word Last Word are given below. If there is no PAF when the packet is complete, the last word (the 30th word in this example) is the last word and is marked as the last word by the last word indicator Last Word Indicator . This is an example of a “true” last word Last Word. PAF occurs when a packet does not have bits to segment into the next packet, ie when a word ends exactly at a packet boundary. Since the final word of the complete packet is actually the final word, it is marked as the final word Last Word. This is another example of a “true” last word Last Word. PAF also occurs when a packet is complete with the remaining few bits segmented into the first word of the next packet. In this case, two consecutive last words Last Word are formed and marked as such. First, the last word of the complete packet is marked as Last Last (“True Last Word Last Word). The first word of the next packet is then forced to shorten the packet by PAF, so In the case of a shortened packet, such as the latter, the last word Last Word appears before 30 words have been transmitted, and then “zero. The word "is then filled. Next is another example of the last word Last Word. PAF also occurs when the packet is in the state of an incomplete packet being assembled. The remaining few bits go to the next word. If the internal word is word-complete when segmented, the partial word is the last word Last Word, especially when the packet is in an incomplete packet being assembled and the internal word is complete with no bits segmented into the next word, a difficult situation arises. If an internal word is sent to the data / header synthesizer before later data (ie, the appearance of a PAF) indicates that this word is the last word, it is determined that this word is the last word. Already too late to mark properly, in this case, a zero word, called a “pseudo” final word, is generated and displayed as the final word. Is composed of 0 bits, which is different from, for example, part of a segmented (incomplete) final word filled with 0 bits. Or another example is described with reference to FIGS. 5 through 16.

  Among the important features of the system disclosed in the following description is the generation of a zero length PAF. A PAF with this length of 0 indicates that the beginning of the GOP will occur soon after, generates and marks the last word Last Word in the data packet, and if necessary, the pseudo-last word Pseudo Last Word can be generated and specific signals can be easily generated in relation to various states of occurrence of the last word Last Word.

  FIG. 2A shows details of the controller 10 of FIG. The controller 10 includes an accumulator 20 in a feedback circuit having a modulo 960 circuit 22. The feedback path includes a buffer register 23 that holds the newly accumulated value at the end of each length Length input cycle. The input PAF word and length word Length are sent to the modulo unit 22 and the accumulator 20 via the input register 24, respectively. The value of the length word Length is successively accumulated successively by the units of the accumulator 20, and the packet length is set to 960 bits by feedback synthesis of the accumulator 20 and the modulo 960 unit 22. The accumulator output supplied from the buffer register 23 represents the bit position in the packet and is sent to the packet state controller 25.

  The packet state controller 25 receives the PAF from the input buffer register 24 and supplies (generates) an output signal necessary for generating a write command in the word state controller 26. When the accumulator bit count (bit count value) reaches 960 or more, the packet complete output signal Packet Complete is supplied to the word state controller 26. When the accumulator bit count does not indicate a word boundary (ie, when the bit count is not equal to an integral multiple of 32), the controller 25 generates an output residual signal Remnant. When the accumulator bit count is zero, a true zero output signal True Zero is provided. This true zero output signal is an important signal for making an accurate final word decision only when a PAF is present. The logic circuit that generates these signals is shown in FIG. 2B as described below.

  When a zero-length empty codeword is received, that is, a zero-length word Length indicating the presence of a no-operation (NO-OP) codeword, the accumulator 20 is idle and holds the final bit count. To do. There are exceptions to this rule, and the PAF may always force the accumulator value to zero regardless of the bit count. Another exception is when the packet forms a complete packet exactly at the packet boundary (ie, the accumulator count is 960). Then, in the next clock cycle, the accumulator count is corrected through the modulo 960 unit 22 to the binary value of the next length word Length. When the accumulator count reaches 960 or higher, the packet is complete.

  In FIG. 2B, a 10-bit accumulator output representing the accumulated length is assigned to I0-I9. A packet is complete when the accumulated length of the packet codeword is greater than 960. When the four upper (MSB) bits I6-I9 of the accumulator are in a logic one state, that state is indicated and the four bits are provided to the AND gate AND30. When 10 bits of all accumulators are in the logic 0 state, true zero True Zero is displayed and the 10 bits are supplied to the OR gate OR31. When the five lower (LSB) bits I0 to I4 of the accumulator are in the logic 0 (zero) state, a “no residue” state is indicated and the five bits are provided to the OR gate OR32. The word address Word Address of the data packer is generated in response to the six lower bits I0-I5 of the accumulator. The AND gate array 34 forces the word address Word Address to a logic 0 state when the packet alignment flag (PAF) appears.

  FIG. 2C shows details of the data packer 12 of FIG. A variable length code word is sent to the data shifter 35. As the data shifter 35, a barrel shifter such as model 74AS8838 manufactured by Texas Instruments can be used. To determine the proper position of the valid bits of the variable length codeword to be concatenated, a low order bit (LSB) portion of the length accumulator output is generated from the packet state controller 25 (see FIGS. 2A and 2B). The word address Word Address is sent to the data shifter 35. When a 32-bit word is formed by concatenating variable length code words, the packed word is transferred to the holding register 36. A flag indicating whether or not there is pack data that can be processed is given by a word ready signal Word Ready supplied from the register 36, and the word is transferred to the data assembly circuit 37. The data assembly circuit 37 uses the control signals WEN1, WEN2 and WZERO supplied from the packed word controller 10 (FIG. 1) to enable data write enable Data Write Enable and the last word flag (indicator) Last Word. The packed data is supplied to the FIFO data buffer 16 in the data and header synthesizer 15 together with the Flag.

  After transmitting the last word in the packet so that the header of the next packet is inserted after the last word of the current packet, the next transport header is subsequently inserted in the combined data stream Is done. The header controller uses the accumulator output Accumulator Output (FIG. 2A) to indicate the position of certain codewords in the packet so that each position is described in the entry point field in the header. A logic array associated with the word state controller 26 and the data assembly circuit 37 generates a final word indicator Last Word Indicator and a flag that enables the data word to be written to the FIFO buffer. Table 1 below generates the final word according to PAF (Packet Alignment Flag), PC (Packet Complete Flag), TZ (True Zero) and REM (Remant Word Segmentation Indicator) which are inputs to the logical array. Indicates the action state. When the output signal of the controller 26 is supplied via the buffer register 28, an output signal is generated from the data assembly circuit 37. These signals write to generate a write enable signal W EN1 that indicates the last word that appears in the next clock cycle, a write enable signal W EN2 that indicates the last word that appears in the current clock cycle, and a pseudo last word Psuedo Last Word. Contains the zero signal W ZERO. This pseudo last word Psuedo Last Word is generated when the occurrence of PAF coincides with the packet formation time located at the internal codeword data boundary of the incomplete packet.

  Note 1: The last word flag Last Word Flag on the current word is indicated by the write enable Write Enable 2.

  Note 2: The last word flag on the word at the next clock, Last Word Flag, is indicated by Write Enable Write Enable 1.

  Note 3: Pseudo-word generation is indicated by write zero Write Zero and is flagged by write enable Write Enable 1.

  A truth table for generating WEN1, WEN2 and WZERO for various examples of operating conditions (cases 1-6 of Table 1) is shown in FIG. This will be described later in connection with FIGS. The algorithm for Table 1 is described in Appendix A. The output signal of the controller 26 is supplied to the output buffer register 28 before being supplied as an output word control signal to the data assembly circuit shown in FIG.

  The data assembly circuit shown in FIG. 4 constitutes the output circuit of the data packer 12 shown in FIG. That is, the data assembling circuit includes logical product gates AND42 and 44, logical sum gate OR46, and D-type flip-flops (DF / F) 43 and 45 in a circuit configuration as shown. The packed data word 32-bit width Packed Data Word is sent to the data FIFO via the AND gate AND42, and the word ready signal Word Ready supplied from the previous packer circuit is sent via the AND gate AND44. The data write enable signal Data Write Enable for the data FIFO 16 in FIG. Data write control signals WEN1, WEN2 and WZERO (FIG. 2A) supplied from packed word state controller 26 are supplied to flip-flops 43 and 45 and logic gate 46 as shown. W EN2 indicates the last word flag Last Word Flag associated with the current word, and W EN1 indicates the last word flag Last Word Flag associated with the word in the next clock cycle. The W ZERO control shows the generation of a pseudo last word Pseudo Last Word that is flagged by W EN1 as the last word (case 4 in Table 1). In this case 4, all ZERO words called pseudo last words are inserted into the packed data word stream Packed Data and written into the data FIFO 16. A word ready input signal Word Ready to the gate AND44 of the assembly circuit is provided by the holding register 36 (FIG. 2C) to indicate that there are 32 bits of packed data that can be processed.

  Next, an example of generation of the last word illustrated in FIGS. 5 to 16 will be described. Some of these examples show the action of zero-length NO-OP words that exist before and after the PAF.

  5 and 6 show different examples corresponding to Case 5 in Table 1. FIG. In FIG. 5, the packet is segmented into the next packet (ie, the accumulator bit value is greater than 960) to form a complete packet. In FIG. 6, the packet forms a complete packet exactly at the packet boundary (ie, the accumulator bit value is equal to 960) and produces residual data without segmentation into the next packet. Without forming a complete packet. In either case, the last word flag Last Word Flag is generated at the same time as the occurrence of the packet complete packet complete. This final word flag Last Word Flag is independently generated when no PAF is present, taking into account each indication of true zero True Zero and residual Remnant. On the other hand, if a PAF is present, the final word flag LAST Word Flag is generated taking into account each indication of true zero True Zero and residual Remnant.

  7 and 8 illustrate case 2 in Table 1. In FIG. 7, PAF occurs without segmentation immediately after the packet is complete, followed by a 32-bit picture start codeword PS. FIG. 8 is similar to FIG. 7 except that three inserted zero-length no-operation (NO-OP) codewords precede the picture start codeword PS. 7 and 8, the PAF is generated in time coincident with the packet complete signal Packet Complete, and the packet ends without causing the remaining data to be segmented in the next packet. FIG. 7 shows the more general case where a 32-bit long picture start codeword PS immediately follows. FIG. 8 shows that NO-OP words are allowed to be inserted.

  FIG. 9 is a diagram related to case 6a in Table 1. In this figure, the PAF is generated in a position that does not coincide with the packet complete signal Packet Complete. PAF occurs after the NO-OP word after the packet is complete without segmentation, followed by the picture start code word PS. In this case, the last word indication (flag) Last Word Indication occurs in association with the packet complete signal Packet Complete, but the accumulator state is idle with a value of 0 (zero) and the true zero indication True Zero is Since there will be a final word indication (flag) Last Word Indication will not occur in association with PAF.

  10 and 11 show an example of case 1 in Table 1. In FIG. 10, PAF occurs in a state where segmentation occurs immediately after a packet is complete, followed by a picture start codeword PS. FIG. 11 is the same as FIG. 10 except that the inserted NO-OP word is present prior to the picture start code word PS. In this case, since there is residual data to be segmented, two final word indicators (flags) Last Word Indicator are required. One of the last word indicators (flags) Last Word Indicator occurs during the packet full packet complete, and the other of the last word indicators (flags) Last Word Indication is one of the PAFs due to segmentation. Occurs after the clock period.

  12, 13 and 14 show an example of case 3 in Table 1. FIG. In these examples, PAF occurs at some position during packet formation, but does not occur on a word boundary (i.e. segmentation into the next word occurs), and packet complete indication Packet Complete Does not match. Then, as a result of the partial start of the word (due to segmentation), the final word signal Last Word typically occurs in the next clock period following the PAF. In FIG. 12, PAF occurs after several NO-OP words after a packet is complete with segmentation, followed by a picture start codeword PS. In FIG. 13, PAF occurs with segmentation occurring at the same time as the word is complete, followed by a picture start codeword PS. In FIG. 14, PAF occurs after a word is complete and after segmentation has occurred with several codewords.

  FIGS. 15 and 16 illustrate case 4 in Table 1 and explain the necessity of generating a pseudo last word Pseudo Last Word, which is a special type of final word. In this case, PAF occurs without a word being segmented, that is, immediately after a word completes on a word boundary that is a multiple of 32 (Figure 15) or just after that (Figure 16). To do. In this case, it is assumed that the complete word has already been supplied (supplied after the PAF) before the information that it was the last word is obtained. In that case, an all-zero pseudo-last word Pseudo Last Word is formed and supplied. The reason this is allowed is that MPEG allows any number of zeros (0) to precede and precede the start codeword, and guarantees that the picture start codeword PS will come next due to the occurrence of PAF. It is because it has been. Further, in these cases, the packet shortage is compensated by a zero value bit (empty) / word by the data / header combiner. In this case, the synthesizer will supply one fewer word, since one zero word is issued and it is pseudo-marked as the last word. In FIG. 15, PAF occurs as soon as a word is complete (with no segmentation), followed by a picture start codeword PS. In FIG. 16, the inserted NO-OP word follows after the word is complete (with no segmentation). Thereafter, a PAF occurs, followed by a picture start codeword PS.

  FIG. 17 shows a detailed configuration of the data / header synthesizer 15 (FIG. 1). The header component is written to the header FIFO 70 in response to the header write enable signal Header Write Enable whenever the header is generated by the header generator 18. Similarly, the packed data word is written to the data FIFO 72 in response to the data write enable signal Data Write Enable when the word is generated by the data packer 12. The last word indicator generated by the data pack process occurs with the last word in the packet, regardless of whether the last word is the 30th word. The header output and data output of each unit of the header FIFO 70 and the data FIFO 72 are multiplexed and supplied on a common bus by a multiplexer (multiplexer) 76, and further supplied to an output register 78. As shown in FIG. 19, the output register 78 supplies the data ready signal Data Ready, packet data packet data, header header, and header indicator header indicator to rate buffers 713 and 714. In response to the zero issue signal Issue Zero supplied from the FIFO state controller 74, the multiplexer 76 can issue a zero word in accordance with an instruction (command).

  Both units of the two FIFOs 70 and 72, the multiplexer 76 and the output register 78 are directed by a controller 74 which is a state machine. After power-on (or power-on) or resumption of a similarly acting operation, controller 74 waits for a processable header to appear. The processable header is sent to the output bus of the multiplexer 76 along with the data ready indicator Data Ready Indicator and the header indicator Header Indicator. The controller 74 then extracts the data and controls the state of the data FIFO 72 as long as there is data that can be processed until the last word indicator Last Word Indicator appears. Each transmitted data is accompanied by a data ready indicator Data Ready Indicator, and the data ready indicator Data Ready Indicator is sent to the output register 78. When the last word indicator Last Word Indicator appears, if it is found that 30 data words have already been supplied, the controller 74 re-checks the header FIFO 70 to see if there is more information to process. inspect. When less than 30 data words are provided, the controller 74 commands the multiplexer 76 using the zero issue instruction Issue Zero to issue a zero word to make up for the packet shortage. . All such zero words are supplied with a data ready indicator Data Ready Indicator. If there is no header and data to transmit, the controller 74 commands the multiplexer 76 to issue a zero word without a data ready indicator Data Ready Indicator for periods when there is no data to process. A flowchart (state transition diagram) showing the operation of the combiner 15 driven by the state machine as described above is shown in FIG. The data ready indicator Data Ready Indicator and header indicator Header Indicator are sent through the output register 78 to the rate buffers 713 and 714 of FIG. These indicators tell the rate buffer that data and header information is present on the bus, maintain header / data registration, and forward error correction (FEC) encoding and data Interleaving is performed. In this system (FIG. 19), FEC processing and interleaving are performed by sending the header first, ie sending one header to the rate buffer, before the data packet represented by the header is sent. Need to start. The empty flag signal Empty Flag sent from the header FIFO 70 and the data FIFO 72 indicates that the header and the data word are not present (empty), respectively, which causes the state controller 74, which is a state machine, to be in an idle state. This state is shown in FIG. 18 as “no header” and “no word” states in states 0 and 1. When the associated read enable signal Read Enable is sent to the header FIFO 70 or the data FIFO 72, respectively, the header / data selection signal Header / Data Select supplied from the controller 74 commands the multiplexer 76, and the unit of the header FIFO 70 is set. Either the header output Header Output or the data output Data Output of the unit of the data FIFO 72 is switched and connected to the signal bus on the input side of the output register 78. A zero-word is added to the incomplete shortened packet to generate a predetermined 30-word data packet, and the data packet is supplied to the output buffer 78. The capacity of the output buffer 78 is considerably larger than the capacity of the header buffer 70 and the data buffer 72 in the preceding stage. These buffers receive and process data efficiently without being interrupted. By adopting such an operation configuration that does not receive interrupts, the timing and synchronization functions are greatly simplified. Its function can be simplified, for example, by eliminating the difficult synchronization process in clock stop / start.

  A complete packet of a predetermined length can be properly used by adding the required number of empty words as described above. By using such complete packets, data can be searched and synchronized in any data state, such as in a variable length codeword system. The start codeword, in particular the I frame start codeword, is an inherent resynchronization point in a data stream compatible with MPEG (interoperability). The start codeword appears at the packet boundary. Such packet boundary processing can be performed by packet perfecting the truncated data packet and determining the packet boundary using an empty word of zero-value bits in the system described herein. The MPEG standard allows any number of zero words to precede the start codeword, so that the receiver / decoder ignores empty words with zero value bits. In this example, the output buffer 78 has a large capacity and a high temporal tolerance (resistance against consecutive identical values), and therefore constitutes a means suitable for executing an empty word pack operation. Note that there is little time available to pack empty words between the appearance of the packet alignment flag PAF and the appearance of the picture start codeword PS at the packet boundary (eg, one clock cycle). It is.

  FIG. 19 illustrates a high-definition television (HDTV) encoding system using an apparatus according to the present invention in the transport processor section. FIG. 19 shows a system for processing a single video input signal. However, it can be seen that the luminance (luminance) component and the chrominance component are processed separately, and the compressed chrominance component is generated using the luminance motion vector. The compressed luminance and chrominance components are interleaved to form macroblocks before parsing the codeword priority. Additional information regarding the system of FIG. 19 is described in US Pat. No. 5,168,356 to Acampola et al. The sequence (column) of image fields / frames shown in FIG. 20A is supplied to a circuit 705 for rearranging the fields / frames according to FIG. 20B. The rearranged sequence is supplied to the compressor 710. The compressor 710 generates a compressed sequence of frames encoded according to an MPEG format. The format has a hierarchical structure, which is illustrated in simplified form in FIG. The MPEG hierarchical structure format is composed of a plurality of hierarchies each having different header information. Each header includes a start codeword, data related to each layer, and an item (provision) for header extension as a normal header format.

  An MPEG format signal generated by this system will be described. (A) A continuous image field / frame of a video signal is encoded according to an I, P, B encoding sequence, and (b) is encoded at a screen level. Data is encoded in the form of MPEG format slices or blocks. In that case, the number of slices per field / frame takes various values, and the number of macroblocks per slice also takes various values. The I-coded frame is compressed in the frame so that it can be processed using only the I-frame compressed data when performing image reproduction. P-coded frames are coded according to the forward motion compensated prediction method, and P-frame coded data is generated from the current frame and the I or P frame that occurs before the current frame. The B encoded frame is encoded according to a bidirectional motion compensated prediction method. B-coded frame data is generated from the current frame and the I and P frames that occur before and after the current frame.

  The encoded output signal of the current system is segmented into fields / frames or groups of images (GOP) illustrated in box L2 row (FIG. 22). Each GOP (L2) includes a header, followed by a segment of image data. The GOP header includes data related to horizontal and vertical screen sizes, aspect ratios, field / frame rates, bit rates, and the like.

  Image data (L3) corresponding to each image field / frame includes a picture head, and slice data (L4) follows the picture header. The picture head contains the field / frame number and the picture code type. Each slice (L4) includes a slice header, and the slice header is followed by a plurality of data blocks MBi. The slice header includes a group number and a quantization parameter.

  Each block MBi (L5) represents a macroblock and includes a header, followed by a motion vector (MV) and a coding coefficient. The MBi header includes a macroblock address, a macroblock type, and a quantization parameter. The coding coefficient is exemplified in the layer L6. Each macroblock includes a total of six blocks including four luminance blocks, one U chrominance block, and one V chrominance block (see FIG. 21). A block represents a matrix of pixels, for example an 8 × 8 matrix, on which a discrete cosine transform (DCT) is performed. The four luminance blocks are, for example, a matrix of 2 × 2 adjacent luminance blocks representing a 16 × 16 pixel matrix. The chrominance (U and V) blocks represent the same area as the entire area of the four luminance blocks. That is, before compression, the chrominance signal is subsampled by two factors (1/2) in the horizontal and vertical directions with respect to luminance. One slice of data corresponds to data representing a rectangular portion in an image corresponding to an area represented as an adjacent macroblock group. One frame includes 360 slices, that is, 60 slices in the vertical direction × 6 slices in the horizontal direction.

  The block coefficient is supplied by the DCT for one block at a time. DC coefficients are generated first, followed by each DCT AC coefficient in order of its relative importance. A block end code EOB (end-of-block) is added to the end of each block of continuously generated data.

  In FIG. 19, the data supplied from compressor 710 is prioritized before being supplied to transport processor 712 which segments the data into a high priority (HP) component and a standard priority (SP) component. Processed by the processor 711. These components are coupled to each forward error correction coding unit 715 and 716 via rate buffers 713 and 714. The rate buffer temporarily stores the packed data and header, and then the FEC error correction coding circuit extracts the packed data and header. A rate controller 718 cooperates with buffers 713 and 714 to adjust the average data rate of the data supplied from compressor 710. The signal is then coupled to a transmission modem 717 whose HP and SP data quadrature amplitude modulates each carrier in a standard 6 MHz NTSC television channel.

DESCRIPTION OF SYMBOLS 10 Controller 12 Data packer 14 Input processor 15 Header synthesizer 16 Input FIFO data buffer (data FIFO)
18 Header Generator 17 Input FIFO Header Buffer (Header FIFO)

Claims (1)

A method for processing a data stream of variable length codewords representing an image, comprising:
Forming a packet from the data stream of the variable length codeword , wherein the packet may have a fewer number of bits than a predetermined number ;
Transporting the formed packet to an output unit when the formed packet is transported to the output unit to inform the beginning of a group of images (GOP) associated with the packet; The step, wherein an alignment flag is inserted into the packet;
Adding a non-functional zero value word to the packet as necessary to form a fixed length packet having the predetermined number of bits to form an output data stream, the output data The stream includes an image field or image frame of a video signal encoded according to I encoding, P encoding, and B encoding, where the I frame or I field is intra frame encoded and the P frame or P field Is forward predictive encoded, the B frame or B field is bi-directional predictive encoded, and the output data stream includes video data having macroblocks including luma blocks and chrominance blocks; Each of the luminance block and chrominance block is discrete Represents a matrix of pixels sine transform (DCT) is performed, a group of adjacent macroblocks form a slice of the video data, and said step,
Said method.

Images