MPEG2 Part1 박찬솔 Contents Audio overview Video overview Video encoding Video bitstream 2/47
Audio overview MPEG 2 supports up to five full-bandwidth channels compatible with MPEG 1 audio coding. extends the coding of MPEG 1 audio to half sampling rates (16 khz, 22.05 khz, and 24 khz) for improved quality for bit rates at or below 64 kbps per channel. MPEG 2.5 is an unofficial,,y yet common, extension to the audio capabilities of MPEG 2. It adds sampling rates of 8 khz, 11.025 khz, and 12 khz. 3/47 Video overview MPEG 2, profiles specify the syntax (i.e., algorithms) and levels specify various parameter (resolution, frame rate, bit rate, etc.). Main Profile @ Main Level is targeted for SDTV applications, while main Profile @ High Level is targeted for HDTV application 4/47
Video overview Levels Low level MPEG 1 Constrained Parameters Bitstream (CPB) supports up to 352 288 at up to 30frames per second. Maximum bit rate is 4Mbps Main level MPEG 2 Constrained Parameters Bitstream (CPB) supports up to 720 576 at up to 30frames per second and is intended for SDTV applications. Maximum bit rate is 15 20 Mbps. High 1440 level This Level supports up to 1440 1088 at up to 60 frames per second and is intended for HDTV applications. Maximum bit rate is 60 80Mbps 5/47 Video overview High level Profiles High Level supports up to 1920 1088 at up to 60 frames per second and is intended for HDTV applications. Maximum bit rate is 80 100 Mbps. Simple profile Main profile without the B frames, intended for software applications and perhaps digital cable TV. Main profile Supported by most MPEG 2 decoder chips, it should satisfy 90% of the SDTV applications. 6/47
Video overview Multiview profile It is possible to encode video from two cameras shooting the same scene with a small angle between them 4:2:2 profile Previously known as studio profile This profile uses 4:2:2 YCbCr instead of 4:2:0, and with main level, e increases the maximum bit rate up to 50 Mbps (300 Mbps with high level). It was added to support pro-video SDTV and HDTV requirements High profile Supported by MPEG 2 decoder chips targeted for HDTV applications. 7/47 Video overview 8/47
Video overview 9/47 Video overview Scalability SNR scalability This mode is targeted for applications that desire multiple quality levels All layers have the same spatial resolution. The enhancement layer increases the video quality by providing refinement data for the DCT coefficients of the base layer Spatial scalability each layer has a different spatial resolution The base layer provides the basic spatial resolution and temporal rate. The enhancement layer uses the spatially interpolated base layer to increase the spatial resolution. 10/47
Video overview Temporal scalability This mode allows migration from low temporal rate to higher temporal rate systems. The base layer provides the basic temporal rate. The enhancement layer uses temporal prediction relative to the base layer. The base and enhancement layers can be combined to produce a full temporal rate output Data partitioning This mode is targeted for cell loss resilience in ATM networks. It breaks the 64 quantized transform coefficients into two bitstreams. The higher priority bitstream contains critical lower-frequency DCT coefficients and side information such as headers and motion vector 11/47 Video overview Transport and program streams The MPEG 2 Systems Standard specifies two methods for multiplexing the audio, video, and other data into a format suitable for transmision and storage The Program Stream is designed for applications where e errors are unlikely. It contains audio, video, and data bitstreams The Transport Stream, using fixed-size packets of 188 bytes, is designed for application where data loss is likely. Also containing audio, video, and data bitstreams all merged into a single bitstream, multiple programs can be carried. Both the Transport Stream and Program Stream are based on a common packet structure, facilitating common decoder implementations ti and conversions. 12/47
Video encoding YCbCr color space MPEG 2 uses the YCbCr color space, surpporting 4:2:0, 4:2:2, and 4:4:4 sampling. The 4:2:2 and 4:4:4 sampling options increase the chroma resolution over 4:2:0, resulting in better picture quality 13/47 Video encoding Coded picture types There are three types of coded pictures. I(intra) pictures are fields or frames coded as a stand-alone still image. They allow random access points within the video stream. As such I pictures should occur about two times a secound P(predicted) pictures are fields or frames coded relative to the nearest previous I or P picture resulting in forward prediction processing B (bidirectional) pictures are fields or frames that t use the closest past and future I or P picture as a reference resulting in bidirection prediction D (DC) pictures are not supported in MPEG 2, except for decoding to support backwards compatibility with MPEG 1. 14/47
Video encoding 15/47 Video encoding Motion Compensation 16 X 8 motion compensation option Two motion vectors (four for B pictures) per macroblock are used, one for the upper 16 8 region of a macroblock and one for the lower 16 8 region of a macroblock. Dual-prime motion compensation option This is only used with P pictures that have no B pictures between the predicted and reference fields of frames. All of the necessary predictions are derived from these. 16/47
Video encoding Macroblocks 17/47 Video encoding Macroblocks in P pictures are coded using the closets previous I or P picture as a reference, resulting in two possible coding Intra coding no motion compensation Forward prediction closest previous I or P picture is the reference Macroblocks in B pictures are coded using the closest previous and / or future I or P picture as a reference, resulting in four possible coding intra coding no motion compensation forward prediction closest previous I or P picture is the reference backward prediction closest future I or P picture is the reference bi-directional prediction two pictures used as the reference: the closest previous I or P picture and the closest future I or P picture 18/47
Video encoding I pictures Macroblocks If the [macroblock quant] column has a 1, the quantizer scale is transmitted. If the [coded pattern] column has a 1, the 6-bit coded block pattern is transmitted as a variable-length code. 19/47 Video encoding 20/47
DCT Video encoding Each 8 8 block (of input samples or prediction error terms) is processed by an 8 X 8 DCT (discrete cosine transform), resulting in an 8 8 block of horizontal and vertical frequency coefficients Input sample values are 0 255, resulting in a range of 0 2,040 040 for the DC coefficient and a range of about 2,048 to 2,047 for the AC coefficients spatial and SNR scalability, non-intra blocks (blocks within a nonintra macro block) are also possible. 21/47 Video encoding 22/47
Video encoding Quantizing The 8 8 block of frequency coefficients are uniformly quantized, limiting the number of allowed values. The quantizer step scale is derived from the quantization matrix and quantizer scale and may be different ent for different ent coefficients and may change between macroblock. If the [macroblock quant] column has a 1, the quantizer scale is transmitted 23/47 Video encoding Zig-zag scan Starting with the DC component, generates a linear stream of quantized frequency coefficients arranged in order of increasing frequency. 24/47
Video encoding P picture Macroblock There are 26 types of macroblocks in P pictures Skipped macroblocks are present when the macroblock_address_increment parameter in the bitstream is greater than 1. 25/47 Video encoding There are 26 types of macroblocks in P pictures Skipped macroblocks are present when the macroblock_address_increment parameter in the bitstream is greater than 1. For P field pictures, the decoder predicts from the field of the same parity as the field being predicted, motion vector predictors are set to 0, and the motion vector is set to 0. For P frame pictures, the decoder sets the motion vector predictors to 0, and the motion vector is set to 0. If the [macroblock quant] column has a 1, the quantizer scale is transmitted. If the [motion forward] column has a 1, the 6-bit coded block pattern is transmitted as a variable-length length code. 26/47
DCT Video encoding AC coefficients are transformed in the same manner as they are for I pictures. Intra block DC coefficients are transformed differently; the predicted values are set to 1,024, unless the previous block was intracoded Input sample values are 255 to +255, resulting in a range of about 2,000 to +2,000 for the AC coefficients. Quantizing Intra blocks are quantized in the same manner as they are for I pictures. 27/47 Video encoding B picture Macroblock There are 34 types of macroblocks in B picture For B field pictures, the decoder predicts from the field of the same parity as the field being predicted. The direction of prediction (forward, backward, or bidirectional) is the same as the previous macroblock, motion vectors are taken from the appropriate motion vector predictors. For B frame pictures, the direction of prediction (forward, backward, or bidirectional) is the same as the previous macroblock motion vector predictors are unaffectied, and the motion vectors are taken from the appropriate motion vector predictors. 28/47
Video encoding If the [macroblock quant] column has a 1, the quantizer scale is transmitted If the [motion forward] column has a 1, horizontal and vertical forward motion vectors are successively transmitted. If the [motion backward] column has a 1, horizontal and vertical backward motion vectors are successively transmitted. If both forward and backward motion types are present, the vectors are transmitted in this order: horizontal forward vertical forward horizontal backward vertical backward If the [coded pattern] column has a 1 the 6-bit coded block pattern is transmitted as a variable-length code. 29/47 Video bitstream 30/47
Video bitstream Video sequence Sequence _end_code This 32-bit field has a value of 000001B7H and terminates a video sequence. Video sequence A sequence header should occur about every one-half second. Sequence_end_code This 32-bit string has a value of 000001B3H and indicates the beginning of a sequence header. Horizontal_size_value This is the 12 least significant bits of the width (in samples) of the viewable portion of the Y component. 31/47 Video bitstream Vertival_size_value This is the 12 least significant bits of the height (in scan lines) of the viewable portion of the Y component. Aspect_ration_information This 4-bit codeword indicates either the sample aspect ratio (SAR) or display aspect ratio (DAR) If sequence_display_extension is not present, the SAR is determined as follows: SAR = DAR (horizontal_size / vertical_size) If sequence_display_extension extension is present, the SAR is determined as follows: SAR = DAR (display_horizontal_size / display_vertical_size) size) 32/47
Video bitstream Frame_rate_code The actual frame rate is determined as follows: frame_rate = frame_rate_value (frame_rate_extension_n + 1) + (frame_rate_extension_d + 1) 33/47 Video bitstream Bit_rate_value The 18 least significant bits of a 30-bit binary number. The 12 most significant bits are in the bit_rate_extension. This specifies the bitstream bit rate, measured in units of 400 bps Vbv_buffer_size_value The 10 least significant bits of a 18-bit binary number B = 16 1024 vbv_buffer_size 34/47
Video bitstream Load_intra_quantizer_matrix This bit is set to a 1 if an intra_quantizer_matrix follows Load_non_intra_quantizer_matrix This bit is set to a 1 if a non_intra_quantizer_matrix follows. 35/47 Video bitstream User data User_data_start_code start This optional 32-bit string of 000001B2H indicates the beginning of User_data User_data These n 8 bits are present only if user_data_start_code is present. Sequence extension A sequence extension may only occur after a sequence header Extension_start_code This 32-bit string of 000001B5H indicates the beginning of extension data beyond MPEG 1. 36/47
Video bitstream Extension_start_code_ID This 4-bit field has a value of 0001 and indicates the beginning of a sequence extension. For MPEG 2 video bitstreams, a sequence extension must follow each sequence header Profile_and_level_indication 37/47 Video bitstream Chroma_format Sequence display extension Et Extension_start_code_ID t t This 4-bit field has a value of 0010 and indicates the beginning g of a sequence display extension. Video_format 38/47
Video bitstream Matric_cofficients This optional 8-bit codeword describes the coefficients used in deriving YCbCr from R G B 39/47 Video bitstream Matric_cofficients This optional 8-bit codeword describes the coefficients used in deriving YCbCr from R G B Sequence scalable extension Extension_start_code_ID This 4-bit field has a value of 0101 and indicates the beginning of a sequence scalable extension If sequence_scalable_extension is not present in the bitstream, no scalability is used. 40/47
Video bitstream Scalable_mode This 2-bit codeword indicates the scalability type of the video sequence 41/47 Video bitstream Macroblock layer 42/47
Video bitstream 43/47 Video bitstream 44/47
Video bitstream 45/47 Video bitstream 46/47
Video bitstream 47/47