TW201143452A - Image encoding device, image decoding device, image encoding method and image decoding method - Google Patents

Abstract

Time direct vector generating part 13 specifies a macro block located spatially at a same position with a macro block of an encoding subject from macro blocks constituting encoded pictures timely close to the macro block of the encoding subject, and generates time direct vectors of a time direct mode from moving vectors of four rectangular blocks located in the center of said macro block.

Description

201143452 * VI. Description of the Invention: [Technical Field] The present invention relates to a video encoding and decoding technology, a compressed image data transmission technology, a read encoding device, a video decoding device, a video encoding method, and an image decoding method. . [Prior Art] "For example, in MP_oving Picture Fine (3) 忭 Gr 〇up) or ITU TU. 26 χ", the international standard image coding method is to use a brightness signal of 16 x 16 pixels (pixel), and corresponding to the luminance Signal signal 16x16 pixel color difference signal 8χ8 pixels aggregated block data 〇31〇4 (18 7 3, hereinafter referred to as "tile set (paste (^〇131〇 (:1〇)) as a unit, and A method of compressing according to a motion compensation technique or an orthogonal transform/transform coefficient quantization technique. In the motion compensation processing of the shirt image encoding device and the video decoding device, the image is referred to in front or rear (piC1: ure), and In the block set unit, the motion vector (vector) is detected and the predicted image is generated. At this time, the reference image is referred to as a P image by referring to only one image, and the two images are simultaneously referred to. The inter-picture predictive coder is called a B picture. In the AVC/H.264 CIS0/IEC 14496-101 ITU-T H.264, which is an international standard method, when B picture is coded, it can be called Direct mode encoding mode For example, Non-Patent Document 1). That is, it is possible to select a coded material that does not have an action vector in a tile set to be encoded, and to use a motion of a tile set of other images that have been encoded as a vector, And an encoding mode of the motion vector of the tile set of the encoding object generated spatially for the predetermined calculation of the motion vector of the surrounding coded tile set. In the direct mode, there are two modes of the time direct mode and the spatial direct mode. The time direct mode refers to the motion vector of the other images that have been encoded. * The scaling of the motion vector is performed in response to the time difference between the encoded image and the image to be encoded, thereby generating a map of the encoded object. An action vector of a block set. The spatial direct mode refers to an action vector of at least one or more coded tile sets located around a tile set of the encoding object, and generates a tile set of the encoding object from the motion vectors. Motion vector. In this direct mode, by using the "direct_spatial_mv_pred_flag" set to the flag of the si ice header, Alternatively, one of the time direct mode or the spatial direct mode may be selected in units of slices. Here, Fig. 12 is a schematic diagram showing a method of generating an action vector in the temporal direct mode. In Fig. 12, "P" indicates P image, "B" indicates B image. In addition, the number 0-3 indicates the display order of the image, indicating that it is a display image of time ΤΟ, ΤΙ, T2, T3. The image encoding processing is based on P0, The order of P3, Bl, and B2 is performed. For example, assume a case where the tile set MB1 among the images B2 is encoded in the temporal direct mode. At this time, in 4 322764 201143452 of the encoded image located behind the time axis of the image B2, the motion vector using the image P3 closest to the image B2 is located in the same spatial position as the tile set MB1. The action vector MV of the tile set MB2. The motion vector MV is the reference image P0, and the motion vectors MVL0 and MVL1 used when the tile set MB1 is compiled are obtained by the following equation (1). MVLO = MVL\ II T2-T0 Γ3-ΓΟ Γ2-Γ3 T3-T0

xMV χλ/V (1) However, in AVC/H.264, when the tile set ΜΒ2 of the image Ρ3 is divided into two or more rectangular tiles, as shown in Fig. 13, the inclusion is included. The motion vector of the rectangular block of the upper left pixel of the tile set ΜΒ2 is used as the motion vector MV. (Prior Art Document) (Non-Patent Document) [Non-Patent Document 1] MPEG-4 AVCCIS0/IEC 14496-10) / ITU-T H·264 Specifications [Disclosure] (Problems to be Solved by the Invention) Due to the conventional technology Since the video encoding apparatus is configured as described above, the motion vector MV of the rectangular block including the upper left pixel of the tile set MB2 is constantly used. Therefore, when the cluster set MB2 of the image P3 is divided into two or more rectangular tiles, the motion vector MV used does not exhibit spatially the same pixel position as the cluster set MB1 of the encoding object. The action 'has the time precision of the direct vector, the coding efficiency

322764 S 201143452 The problem of deterioration. The present invention has been developed for the purpose of producing a close-to-be-wrapped subject, and the object of the present invention is to obtain a temporal direct vector of the operation of the video 2 block set for improving coding efficiency, and the present "tr-transcoding method". The generation time vector 7 is decoded by the same image decoding device and image decoding solution that is more efficient than the editing device. (The means for solving the problem) The image encoding device of the present invention is time-based. Sexually located in the composition, 11 in the means of construction, specific to a tiled block near the tile, and the action vector of the plurality of segmentation regions directly at the time of the tile image: The time direct vector of the time direct mode is generated. (Effect of the Invention) According to the present invention, the means for generating the object to be encoded temporally is configured to be from the block constituting the coded image which is specific to the image to be encoded. ' A tile located at the same position between the centers of the above-mentioned tiles, and the time of the direct mode is straight; := the motion vector of the domain generates the tile action of the code object Time, so it can produce close to the editing result, that is, it has the time to increase the bat $ = connected mode direct vector, and its effect [embodiment] The effect of the editing efficiency of the object block. Hereinafter, refer to the figure, (^1 An embodiment of the present invention will be described with reference to W («). 322764 6 201143452 " In the first embodiment, a description will be given of a rectangular region in which an image frame constituting an input image is equally divided into 16x16 pixels (Fig. A video encoding device that performs encoding in a frame and a video decoding device corresponding to the video encoding device in units of blocks. In addition, in the first embodiment, the video encoding device and the video decoding device are The syllabic method used in the AVC/H.264 specification will be described as an example. However, in the first embodiment, the video editing device and the video decoding device are used when the direct mode is used. The direct mode is used for the time, and the spatial direct mode can be used as needed. Fig. 1 is a view showing a configuration of a video encoding device according to the first embodiment of the present invention. In Fig. 1, the motion compensation predicting unit i Selecting a reference image of the ^ ^ frame from the motion compensation prediction reference image that is misplaced in the frame of the frame memory 8 to form a tile set of the input image (or a pair of the tile set) (sub) tile set) performs motion compensation prediction processing for each unit, generates an action vector of the tile set (a tile set of the encoding target), and generates an image, and executes the selected reference image for each tile. The processing of the output such as the number, the motion vector, and the predicted image. β , However, for the convenience of explanation, it is assumed that the predicted image is generated by using the tile set as a single motion vector. That is, the motion compensation prediction unit In the 1st system, the motion vector storage operation of the sub-block set that constitutes or divides the tile set is attached to the 1 memory u (see FIG. 2), and in the time direct mode, the pair 322764 7 201143452 constitutes the input image. a set of tiles of the edited image near each tile set, from the set of tiles constituting the same position in space at the time of the tile set, and, specifically, the tile set For example 4 ® blocks (segmented areas) A motion vector is generated in the motion vector of the memory layer 11 placed at the center of the tile set (the motion vector vector is stored, and the processing for generating the predicted image is performed by directly controlling the time of the direct mode using the time) The motion compensation pre-fighting subtractor 2 is used to calculate the difference: the difference between the image and the input image = the predicted difference signal generated by the system 1 is output to the coding mode: the pre-representation of the differential image is indicated The coding mode determination unit 3 evaluates the prediction efficiency from the subtraction signal, and selects at least one of the prediction efficiency = out of the prediction difference prediction difference signal that is rotated from the subtractor 2 2 in the operation compensation prediction unit i. The motion vector used for the generation of the pre-predicted differential signal, the tile set class, and the predicted image of the divided signal, such as the coding mode used in the tile set, is the tile set type (for example, any of the direct modes) Information) and the parameter indicates the inter-frame mode or code mode information ^ output to the variable length editing part 9: The number is the highest predicted prediction difference signal output to the compression of the second prediction 4 embodiment by a compression system for the determination of the prediction difference signal from the ^ coding mode DCT (Discrete Cosine (c〇sine). 3 rounds out the DCT coefficient, and quantizes the DCT coefficient, and then calculates the calculated DCT coefficient >1 reduction data (quantization coefficient) round a, set 5 and variable length coding The processing of the section 9. Bureau.卩 Decoding unit 322764 8 201143452 Further, subtraction H 2, coding mode determination unit 3, and compression quantization means. * The local decoding unit 5 determines that the DCT coefficients are obtained by subdividing the compressed data output from the compression unit 4 and performing inverse discrete cosine transform processing on the DCT coefficients. The processing of the output error signals of the three output differential signals. The adder 6 is a process of expressing the predicted image of the predicted video generated by the signal calculated by the local decoding unit 5 and the motion compensation predicting unit 1. "Partial solution (4) Bureau I code video loop chopper (i〇Qp filter) 7 system will compensate the encoding distortion contained in the adder 6 rim decoding shirt image money, implement the distortion of the compensation code The partial decoded image indicated by the partial image signal is output as a reference image to the frame memory 8. The frame memory 8 is a RAM for storing the reference horse image output from the loop chopper 7 ( A recording medium such as a random access memory or a random access memory. The encoding mode information output from the compression unit 4 and output from the motion compensation prediction unit 1 (block type/ The sub-block set type, the motion vector, and the reference image identification (entr〇Py) code, and generate a bit stream (program) indicating the result of the encoding to implement the bit stream output. The processing of the variable length coding unit 9 constitutes a variable length coding means. i_ can also directly encode the motion vector information. For example, 322764 9 201143452 can also be used as H.264/AVC. Way to use a coded tile set The motion vector generates a prediction vector and encodes the difference between the prediction vector and the prediction vector. When predicting the motion vector of the non-rectangular tile, for example, the method as shown in Fig. 11 can also be considered. In Fig. 11, the arrow The number indicates the motion vector used for the derivation of the prediction vector. The prediction vector of the segmentation region indicated by the three motion vectors enclosed by 〇 is the central value of the three motion vectors enclosed by 〇 ( Fig. 2 is a view showing the configuration of the operation compensation predicting unit 1 of the video encoding apparatus according to the first embodiment of the present invention. In Fig. 2, the motion vector memory 11 stores the encoded image. The recording vector of the motion vector of the tile set (or the sub-block set obtained by dividing the tile set) is, for example, a recording medium such as a RAM. The motion vector search unit 12 receives the inter-frame mode (inter mode) when the coding mode is received. Information (for example, receiving information indicating the use of the inter-frame mode from the outside), and searching for the most suitable motion vector in the inter-frame mode, and performing the motion vector to the motion compensation processing unit 14 and the motion vector memory When the time direct vector generation unit 13 receives the information indicating that the encoding mode is the time direct mode, the encoded image of the block set of each encoding target is formed near the time of the set of the tile set. Among the set of tiles, a set of tiles that are spatially identical to the set of tiles is specified, and actions from, for example, four tiles (partitioned regions) located at the center of the tile set are directed to 10 322764 201143452 The wedge-shaped straight-forward vector vector.: 1^1 motion vector) generates the time directly to implement the round-h motion vector processing and the motion to the memory 11. Further, the time direct vector generation unit 13 The direct vector is used to generate the reference image of the 1 frame of the ship 8 and the actual image is inserted into the image to generate the predicted image. . The x motion compensation processing unit 14 has a configuration =. 3 is a solution of the video decoding device according to the first embodiment of the invention, in which the variable length decoding unit 21 rotates the bit stream (encoded data) that is output from the image of Fig. 1 . The compressed data (quantization coefficient module 'the bit string type / sub-block set type, motion vector, reference block set decoding 'the * shrink data to the prediction error solution two; the image of the wheel) The horse mode information is processed by the motion compensation compensation unit 23 to form a variable length decoding means. Further, the 'predictable# difference decoding unit 22 obtains the DCT coefficients from the variable length == sub-ization, and performs the calculation of the differential image signal by performing the pairing of '' and the inverse DCT processing (equivalent to the first image). The coding mode determination unit 3:=: 322764 11 201143452 The prediction error signal of the sub-signal) constitutes an inverse quantization means_. Prediction error decoding unit

Motion compensation_Zou 23 is stored in the reference image above the frame memory 2 frame and reads the reference miscellaneous portion from the variable length decoding unit 21, and the block from the variable length such as = The set type/sub-block set type indicates that when the inter-frame mode is used, the motion compensation prediction process is performed by using the operation output from the variable length decoding unit 21 to describe the reference picture, thereby realizing the expansion. Image processing. & generates a prediction, and the output type/sub-block set type of the S variable from the variable length decoding unit 21 indicates that the direct mode is used, and the motion compensation prediction unit i is set by the image coding of the image. The same ^ ^ ^ directly vector ' and (4) time direct vector and the above-mentioned Shen Zhao F B action compensation deletion processing, secret face ying _1. The adder 24 adds the 22-difference knife image generated by the motion compensation prediction unit 23, and generates a solution 'decoding loop chopper 25 indicating a local decoded image equivalent to the output from the i: ::= r. The coding distortion Μ contained in the decommissioned video signal generated by the adder 24 is compensated, and the decoded image decoded by the compensation code is stored as a dump image and stored in the stick memory 26' and Real scale The processing of decoding the image to the outside. Further, the adder 24 and the loop chopper 25 constitute an image adding means. 322764 12 201143452 The frame money 'body 26 series stores the recording medium such as the RAM and the image output from the loop filter. 、", 〜4 Figure 4 is a configuration diagram of the video decoding device operation compensating prediction unit 23 according to the first embodiment of the present invention. The device is shown in Fig. 4, and the operation is performed in the memory 31. A set of tiles constituting the decoded m 圄 ( (or a recording medium such as RAM, which is a block 隹 丄 丄 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 直接 直接 直接 直接 直接 直接 直接 直接 直接 直接Set type/sub-circumference on 1 i/Μ tile type indicates that when using direct mode culvert, for each block set that is a decoding pair. ^ Τ, from the block located in the block =:::::= In the _, the tile set at the position of the tile 丨J, from the motion vector of the four 堍r 八八块 (divided region) located at the center of the tile ( (the action of the memory vector memory 31 & θ ^ ^ θ 勒 篁 篁 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生 产生

In the case of the motion vector, the motion compensation processing unit 33 and the motion & β & F function are output to the memory 31. Further, the time direct vector generation unit constitutes a direct vector generation means. The motion compensation processing unit 33 reads the reference code indicated by the identification number output from the variable length decoding unit 21 from the i-picture 4 image stored in the frame memory: ", ~, and The tile set type/sub-block set type Pear Table Μ "Letter" uses the motion vector output from the variable length damper unit 21 and the upstream stream L ^ .

Thereby, the predicted image is generated; the I w豕 'field tile set type/sub-block set type representation makes 13 322764 201143452 use the direct mode meaning time' to use the action vector output from the time direct vector generation unit 32 The reference image is subjected to motion compensation prediction processing, thereby performing processing for generating a predicted image. Further, the motion compensation processing unit 33 constitutes a predictive video generating means. In the first drawing, the operation compensation pre-compensation unit 1, the subtractor 2, the encoding mode determination unit 3, the compression unit 4, the local decoding unit 5, the adder 6, and the loop filter are assumed to be constituent elements of the video encoding device. The device 7 and the variable length 9 9 are each composed of a dedicated hardware (for example, a semiconductor semiconductor circuit with a CPU mounted or a single-chip microcomputer), but when the image coding device is composed of a computer, The memory storage used in the computer includes a motion compensation detection unit 1, a subtractor 2, an encoding mode determination unit 3, a compression unit 4, a local decoding unit 5, an adder 6, a loop filter 7, and a variable length. The program for processing the content of the encoding unit 9 and the program stored in the memory by the CPU of the computer. Fig. 5 is a flow chart showing the processing contents of the video encoding apparatus according to the first embodiment of the present invention. In the third diagram, the variable length decoding unit 21, the prediction error decoding unit 22, the motion compensation prediction unit 23, the addition unit 24, and the loop filter 25, which are constituent elements of the video decoding device, are each dedicated. The hardware (for example, a semiconductor integrated circuit mounted with a CPU, or a single-chip microcomputer or the like) constitutes a 'however' the image encoding device is constituted by a computer, and the memory storage of the computer may be described as having a variable length decoding unit 21 The program of the prediction error decoding 邡22, the motion compensation prediction unit 23, the addition unit 24, and the loop filter 25 is executed by the CPU of the computer to execute the program stored in the memory 322764 14 201143452. Fig. 6 is a flow chart showing the processing contents of the video decoding device according to the first embodiment of the present invention. Next, the operation will be described. First, the processing contents of the video encoding apparatus of Fig. 1 will be described. When the motion compensation video signal indicating the input image is input, the motion compensation prediction unit 1 divides each frame of the motion picture signal into a tile set unit (or a sub-block set unit). The motion compensation prediction unit 1 is configured to be used in the motion compensation prediction reference image stored in the frame of the frame memory 8 when the motion picture signal is divided into tile set units (or sub-block unit units). Select the reference image of the 1 frame, and perform motion compensation prediction processing for each color component in units of tile sets (or sub-block sets) to generate the action of the tile set (or sub-block set) of the encoding target. The vector produces a predicted image. The motion compensation prediction unit 1 outputs a predicted image to the subtractor 2 when the motion vector of the tile set (or the sub-block set) to be encoded is generated to generate the predicted image, and is used for the predicted image. The generated motion vector, tile set type/sub-block set type (for example, containing the coding mode used in the tile set (or sub-block set) is either an inter-frame mode or a direct mode The information) and the identification number of the reference image are output to the encoding mode determination unit 3. Hereinafter, the processing content of the motion compensation prediction unit 1 will be specifically described. However, here, for convenience of explanation, it is assumed that a motion vector is generated in units of a tile set to generate a predicted image. 15 322764 201143452 The motion vector search unit 12 of the full-motion complementation prediction unit 1 receives information indicating that the code pattern is the inter-frame mode (for example, information from the use of the inter-frame mode (4)) (5th) In step ST1) of the figure, ^ = mode search 4 is the most appropriate motion vector, and the motion vector is output to the motion compensation portion Z 14 (step ST2). The process of exploring the most appropriate motion vector in the n mode is itself a known technique, and thus a detailed description thereof will be omitted herein. The information of the direct vector generation unit of the motion compensation prediction unit 1 (4) The information of the meaning of the connection mode 1), the vicinity of the existing: tile set, which is spatially located when the block set is located Among the tile sets, a tile set that is at the same position as the tile set is specified (step ST3). ^ p,., ^ The tile set of the field editing object is shown in Fig. 12 as the image B2 has been circled MB1 'from the time of the composition block MB1 near the block set MB1; π image P3 block Concentration, specific tile placements that are spatially co-located with the tile set brain. The time is directly assigned to the spatial group 13 in the case where the tile set MB2 at the same position as the tile j of the encoding object is specified, as in the middle, from the encoded code stored in the motion vector memory 11. The motion vector is pure to the four rectangular tiles in the center of the tile set view (the segmentation is called the motion vector MVi (i, 2, 3, 4) (step (10). Λ μ / in the example of Figure 7 The motion vector of the rectangular block on the upper left side is set such that the motion vector of the rectangular block in the lower right is set to ·, and the number is assigned in the order of raster scan. 322764 16 201143452 The time direct vector generation unit 13 obtains four When the motion vector MVi (i = 1, 2, 3, 4) of the rectangular block is as shown in the following formula (2), the set of the coding target is generated by obtaining the addition average of the four motion vectors MVi. The time direct mode of the MB1 direct vector MVdiwt (step ST5).

MVdircct = - (2) When the time direct vector generation unit 13 generates the time direct vector MVdirect, the time direct vector MVdirect is output as the motion vector to the motion compensation processing unit 14 and the motion vector memory 11. The operation compensation processing unit 14 of the motion compensation prediction unit 1 is configured to use the motion vector and the image stored in the frame memory 8 when the motion vector is received from the motion vector search unit 12 when the coding mode is the inter-frame mode. The frame performs motion compensation prediction processing with reference to the video, thereby generating a predicted image (step ST6). On the other hand, when the encoding mode is the direct mode, when the time direct vector generating unit 13 receives the time direct vector MVdirm as the motion vector, the time direct vector MVdirect and the reference of the 1 frame stored in the frame memory 8 are used. The image performs motion compensation prediction processing, thereby generating a predicted image (step ST6). Further, since the motion compensation prediction processing of the motion compensation processing unit 14 is well known, the detailed description thereof will be omitted. When the motion compensation prediction unit 1 generates a predicted image, the subtracter 2 calculates a difference image between the predicted image and the input image, and outputs a predicted difference signal indicating the difference image to the encoding mode determining unit 3 (step ST7). 17 322764 201143452 The coding mode determination unit 3 evaluates the prediction efficiency of the prediction difference signal every time the prediction difference signal is received from the subtractor 2, and selects the prediction from at least one or more prediction difference signals output from the subtractor 2. The most efficient predictive differential signal. The process of evaluating the prediction efficiency of the differential signal by the coding mode determination unit 3 is a conventional technique, and thus detailed description thereof will be omitted. The coding mode determination unit 3 selects the action vector and the tile set type used in the generation of the predicted image with respect to the predicted difference signal, after selecting the prediction difference signal having the highest prediction efficiency. /Sub-block set type (for example, information indicating whether the encoding mode used in the tile set is the inter-frame mode or the direct mode), and the encoding mode information of the reference image identification number to the variable length encoding Part 9 output. Further, the coding mode determination unit 3 outputs the prediction difference signal having the highest prediction efficiency to the compression unit. However, when the encoding mode is the inter-frame mode, the encoding mode determining unit 3 includes the operation vector for generating the predicted image in the encoding mode information, and the encoding mode information including the motion vector is transmitted to the variable length encoding unit 9. When the encoding mode is the direct mode, the encoding mode information is not included in the encoding mode information, and the encoding mode information not including the motion vector is output to the variable length encoding unit 9. When receiving the prediction difference signal from the coding mode determination unit 3, the compression unit 4 performs DCT processing on the prediction difference signal, thereby calculating the DCT coefficient and quantizing the DCT coefficient (step ST8). 18 322764 201143452 The compression unit 4 outputs the compressed data of the quantized DCT coefficients to the local decoding unit 5 and the variable length coding unit 9. ▲ Bureau. When the compressed data is received from the compression unit 4, the p-decode 5 is obtained by dequantizing the compressed data to obtain a DCT coefficient, and performing the DCT-based: inverse DCT processing on the track to calculate the output from the coding mode determining unit 3. The _ error signal of the predicted differential signal. 6 Μ Local Decoding 5 Calculate the prediction error: The measurement error signal and the pre-: = prediction signal generated by the motion compensation compensation are added to generate a representation 邛 decoded video signal. The horse's local loop filter 7 is iπ # τ A is output from the adder 6: the person's 'code processing is prepared' and will be compensated, and the compensation code image signal towel contains the missing part of the mother. The decoding is called "variable length editing shown by the material code image signal ==: memory 8. The compression data and the slave motion compensation prediction unit 1: material 'is = ΓΓ tile set type, motion vector (when = = , according to the identification number of 4) Entropy coding, production flow, the bit stream is rounded out (steps of the wide sea' W 3rd image decoding device processing variable length solution Ma Department 21 when input from the first When the bit of the output of the graph is input, the device is: from the bit: the mode information (the tile set type/the sub-block set type, the action direction is the frame handle), and the reference image朗 钱 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ When the variable length is decompressed, (4) the pressure solution is solved. ^ 21 receives the inverse DCT ^ * > CT coefficient of the DCT coefficient, and the pre- The error signal (corresponding to the coding mode determination unit 3 that outputs the prediction difference signal indicating the differential image from the above-mentioned code), the operation mode compensation unit 3 is rotated (step ST12). The length decoding unit 21 receives the reference video received on the target, and performs: ===, the frame is additionally operated, and the prediction unit is operated; the image is read. The received tile set type/sub-block set type; :=: The code part 21 is connected to the sub-block set type and discriminates the], ', 参', ,, the tile set type/form or the direct mold handle to compose the image ^; the device money uses the inter-module ^ ra1 ^ ^ ^ ^ 11 ^ f ^ ® ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ The action vector and the above reference image 21 round out to produce the predicted image. Also the action complementer predictive processing, the car on the other hand, when the first picture When it is in the coding mode, it is the same as the first diagram: 'The direct use of the direct mode is the same as the first phase, and the motion compensation compensation prediction process 2 of the code device: _ time direct vector implementation I, The processing content of the motion compensation prediction unit 23 will be described in two parts. The time vector of the motion compensation unit 23 is generated from two 322,764 2 . 0 201143452 The tile set type/sub-block set type output by the variable length decoding unit 21 indicates the time when the direct mode is used, and the block set for each decoding object is composed at the time of the tile set. Among the tile sets of the decoded image in the vicinity, a tile set spatially located at the same position as the tile set is specified (step ST14). For example, when the tile set of the decoding target is as shown in FIG. When it is the tile set MB1 of the image B2, it is specified from the tile set of the decoded image P3 which is temporally located near the tile set MB1, and the tile set which is spatially identical to the tile set MB1 is specified. MB2. When the temporal direct vector generation unit 32 specifies the tile set MB2 that is spatially identical to the tile set MB1 to be decoded, as shown in FIG. 7, the decoded motion vector stored in the motion vector memory 31 is decoded. The motion vector 1 丨 (丨 = 1, 2, 3, 4) of the four rectangular tiles (divided regions) located at the center of the tile set MB2 is obtained (step 81 '15). When the time direct vector generation unit 3 2 obtains the motion vector MVi (i = 1, 2, 3, 4) of the four rectangular blocks, as shown by the above equation (2), four motion vectors MVi are obtained. The addition is averaged, and the time direct mode MVdirect of the time direct mode of the tile set MB1 of the decoding target is generated (step ST16). The time direct vector generation unit 32 is connected to the motion compensation processing unit 33 and the motion vector memory 31 when the time direct vector MVdirect is generated as the motion vector. When receiving the identification number of the reference video from the variable length decoding unit 21, the operation compensation processing unit 33 of the motion compensation prediction unit 23 performs the storage of the reference image from the frame 1 of the frame memory 26 stored in 21 322764 201143452. Read the reference image shown by the identification number. The flat motion compensation processing unit 33 uses the slave variable length decoding unit 21 when the tile set type/subblock set type that is rotated from the variable length decoding unit 21 indicates the use of the two culverts of the inter-frame mode. The output motion vector and the motion compensation prediction processing described above generate a prediction shadow ST17). (On the other hand, when the tile type/sub-block set is output from the variable length decoding unit 21 When the type indicates the meaning of using the direct mode, if the time direct vector MVdi(10) is received from the time 接=the vector generation unit 32 as the dynamic amount, the direct vector MVdirect and the reference image are used as the dynamic expansion. The predicted image is generated by the compensation prediction process (step ST17). The motion recognition technique: the motion compensation prediction process of the T motion compensation processing unit 33 is a conventional technique, and thus detailed description thereof will be omitted. The adder 24 is used when the motion compensation prediction is performed. The portion 23 generates a translation error image section 22 which adds the image to the posterior difference image and generates a decoded image corresponding to the decoded image of the local decoded image outputted from the adder 6 of the I: code set of Fig. 1 . Image signal Step ST18) The private loop filter 25 is used to compensate for the t-subtraction of the decoded image when the adder 24 generates the missing version, and the decoded image signal after the distortion is represented by: =:) The frame memory 26' and the image is output to the outside

S 322764 22 201143452 In the first embodiment, according to the first embodiment, the time direct vector production and the weighting are configured to be attached to the image set of the image from the block set that is temporally located in the encoding target. The set of tiles corresponding to the encoding object "the action direction of the solid rectangular block = pull from the center of the set of the tile set, so the time of the direct mode can be generated directly in the time of the direct connection mode: : The time series of the motion vector of the block set motion vector is the image of the r code efficiency, which can improve the composition of the code, and the composition of the code is L (four), the time direct vector generation unit 32 ^ @ ^, and the set of decoding object blocks. The solution in the vicinity: = the set of blocks in the block set 'specifically the axis object in the space rectangle' and the time from the 4 in the center of the set of tiles to generate the time direct mode directly Vector, which is similar to the time of the block set motion vector of the decoding object. The direct vector direct result 'results' is obtained by decoding the stream encoded by the image encoding apparatus capable of decoding the object: Like the effect of the solution In addition, in the first embodiment, it is exemplified that the time direct vector parts 13 and 32 are generated from the actions of four rectangular blocks located at the center of the tile set = ΜΜι - 1, 2' 3, 4) The time direct vector of the time direct mode 'but the block located at the center of the tile set is not limited to four. As long as the number of tiles located is less than 3 or more than 5, the time can also be generated from the (four) of the three deniers. The time of the direct mode is directly to M MVdirect ° 322764 23 201143452. In addition, although the first embodiment of the bear-vector is used as the motion vector, but fτ can be exemplified as the generation time directly (second embodiment), the vector is used as the prediction vector. In the first embodiment of the first embodiment, the parts 13 and 32 are obtained by directly determining the operation of the block (decoding target) in which the square is generated as a time direct vector. The addition of modulo 1 ^12, 3, 4) corresponds to the browning object (decoding object) = direct vector, -, but also the size of the tile, which is divided into a plurality of divided regions within the tile set. , thereby generating a time direct mode weighted phase (four) The vector-two figure 8 shows that there is no corresponding square of the direct vector I corresponding to the m-th order, and each of the blocks is located near the tile set in the first embodiment. Set, from the set of time sets, the set of tiles that are associated with the tile set in the decoded image. The tile at the same position on the second frame is the image of the coded object (the object to be dehydrated) The block set is as shown in Fig. 12 - the block set MM of the image B2. The coded near the MB1 of the picture can be located at the block set from the block set and the block set MB1 is located at the decoded image P3. The tile set is a set of legs at the same position as the direct vector generation unit. The tile set MB1 of the (decoding target) is stored in the motion vector memory 11 as shown in FIG. 8 when the tile set 322764 24 201143452 MB2 located at the same position as the encoding object 2 is specified. Among the encoded motion vectors (31), the motion vector MVi (i = 1, 2, 3, 4) of the four tiles (divided regions) located at the center of the tile set MB2 is obtained. In the example of Fig. 8, the motion vector of the upper left tile is set to MVi, and the motion vector of the lower right tile is set to MV4, and the numbers are assigned in the order of raster scan. Further, the time direct vector generation units 13 and 32 acquire the distance di (i = 1, 2, 3, 4) from the center of the tile set MB2 to the center of the four tiles. Further, the distance from the center of the tile set MB2 to the center of the four tiles is proportional to the size of the tile. When the time direct vector generation units 13 and 32 acquire the distance ώ from the center of the tile set MB2 to the center of the four tiles, the time vector generation unit 13 and 32 obtain the distance ώ from the center of the tile set MB2 to the center of the four tiles. The weights of the four motion vectors MVi are added by the distance di corresponding to the center of the four tiles, thereby generating the temporal direct vector MVdirect of the temporal direct mode. MVdircct = 1 —, (3)

(4) As described above, according to the second embodiment, the temporal direct vector generation units 13 and 32 are configured to weight-add the motion vector MVi of the plurality of divided regions in accordance with the tile size of the plurality of divided regions. When this occurs, 25 322764 201143452: the direct vector MVdi of the straight ϊΓ — is compared with the above ith real MV plus; the motion vector ML of the four rectangular blocks located at the center of the tile set is averaged and the time is directly generated. _2 ==" to reach the boostable time direct vector-precision ί (third embodiment) In the above-described first embodiment, (4) training is obtained by finding the coded vector: : a rectangular square, vector =? 2) = average, and the addition of the temporal direct mode, 2, 3, 4) can correspond to the compiled object (decoding = to the second I' but also the difference between the fields (four) as a vector, Wei is divided into several partitions, and the motion vector MVi of the °° domain is generated by MV_. The time direct mode is below the direct vector, and the time is straight. Method of production. Time directly MVd 卞 卞 卞 卞 卞 的 的 的 的 的 的 的 的 的 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 说明 MV 说明 MV MV MV MV MV MV MV MV MV MV MV MV In the set, the set of tiles that have been resolved with the tile is specified. 4 blocks located at the same position, for example, when the tile set of the object (the object to be solved) is 322764 26 201143452 as shown in Fig. 12 as the block set of the image B2, the MB1 can be constructed from the time. The coded image P3 in the vicinity (the decoded tile = time direct vector generation sections 13 and 32) when the tile set ΜΒ1 of the (decoding target) is spatially upper and the coding target ΜΒ2 is as shown in FIG. From the coded motion vector stored in the set of motion blocks 31 (the decoded motion vector, the four tiles in the center of the block 11 and the block set ( 2 are obtained in the figure = 1, 2, 3,4)0 #(6) The motion vector MVi(i) In the example of Fig. 9, the motion vector of the lower right block of the upper left block is set to MV1, according to σ set 5, ". In addition, the difference between the four blocks of the center of the time direct vector generation unit 13, the buckle, and the MB2 is obtained in the θ system. The time vector generation unit 13 and 32^^MVdl are located in the tile set (1 = 1,2, 3, 4). The difference between 4 and 4 tiles is shown in the direction of 4mv=4 tiles, corresponding to 4 tiles. The differential action to the sun is weighted by the vector MVi as shown in the following equation (5), thereby generating Q MVdi and 4 movements to '5 MVdirect. 曰 Direct mode time directly' gives a fixed value, but when When the differential action vector MVdi is 〇, it is weighted. ' Σ^Μνΐ (Srect

R (5) 322764 27 (6) 201143452 (6) 201143452 &lt;Pi MVd; The difference between the difference ==vector_l indicates the degree of isolation of the vector, and the weighted leaf 丄: the larger the size, the greater the isolation may be Therefore, the reciprocal of the formula (6), which is the equivalent of the crowning method, can be weighted by the (6) sickle motion vector read 1 size and the isolated vector. According to the third embodiment, the time direct vector 曰TM is configured to weight-add the motion vector points of the plurality of divided regions corresponding to the difference operation direction of the plurality of divided regions, and the time of the direct mode is directly The vector MWt is compared to the above-described first real (four) state '彳 to achieve the effect of improving the precision of the direct vector MVdi(10). (Fourth Embodiment) In the second embodiment, the motion vector M1 of the plurality of divided regions is weighted and added in accordance with the tile size of the plurality of divided regions. In the third embodiment, the third embodiment is illustrated. The motion vector MVi of the plurality of divided regions is weighted and added for the differential motion vector MVdi ' of the plurality of divided regions, but may also be corresponding to the coding mode of the plurality of divided regions, and the motion vector MVi of the plurality of divided regions is weighted plus. For example, when the divided area belonging to the encoded tile (decoded tile) is an intra b ock, a method of culling the segmented region from the calculation object to generate the temporal direct vector MVdirect may also be considered. Specifically, in FIG. 8, for example, when the upper left tile (divided area) is an internal tile (in the eighth figure, although the upper left tile is marked with an action vector 322764 28 201143452 MVi, but it is an internal image In the case of a block, there is no information of the motion vector. By subtracting the upper left tile from the calculation object, the motion vectors MVi(i = 2, 3, 4) of the remaining three tiles are weighted and added. The time direct vector MVd irect is generated. Since the inner block does not have the information of the motion vector, by eliminating the inner block from the calculation object, it is possible to avoid unnecessarily making the time direct vector a zero vector and raising the time. The precision of the direct vector. (Fifth Embodiment) In the first to third embodiments described above, the time direct vector MVdirect is generated by using vectors of four tiles, but the motion vectors of all the tiles in the tile set may be used to generate time. Direct vector MVdirect. Fig. 10 is an explanatory diagram showing a method of generating a temporal direct vector MVdir^t using motion vectors of all tiles in a tile set. If it is assumed that the motion vector of k tiles in the tile set MB2 which is spatially located at the same position as the tile set MB1 of the encoding object (decoding object) is MVi (OSiSk), the size of k tiles is Si (OSiSk) ), by calculating the following formula (7), the time direct vector MVdirect of the time direct mode can be generated. Σ MV MV = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = 29 322764 201143452 [Simple description of the diagram] Figure 1 shows the composition of the figure. The second embodiment of the video encoding device according to the first embodiment of the present invention is shown in the third embodiment of the video encoding device according to the first embodiment. Make up the picture. Fig. 4 of the first embodiment of the present invention is an operation of the image decommissioning device according to the first embodiment of the present invention. The video editing device according to the first embodiment performs 5:=== 1 Implementation form - an illustration of the method of direct (4) direct king method for the village day (four). 1 MVdirect time 2'' shows an explanatory diagram of a method of generating an icMVdirect by weighted addition of the corresponding tile size. The =9 figure is an explanatory diagram showing a method of generating the direct vector MWt by the addition of the corresponding differential action vector. </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> ^ 11 The picture shows the action of non-rectangular blocks. Figure 12 of the prediction process of the ghost action inward is a schematic diagram showing a method 322764 30 201143452 for generating an action vector in a time direct mode. Fig. 13 is an explanatory diagram showing an operation vector MV used when the tile set mb2 of the image p3 is divided into two or more rectangular tiles. [Description of main component symbols] 1 Motion compensation prediction unit 2 Subtractor (quantization means) 3 Encoding mode determination unit (quantization means) 4 Compression unit (quantization means) 5 Local decoding section 6 Adder 7 Loop filter 8 Frame memory 9 variable length coding unit (variable length coding means) 11 motion vector memory 12 motion vector search unit 13 temporal direct vector generation unit (direct vector generation means) 1 motion compensation processing unit (predictive video generation means) Worry length depletion unit (variable length decoding means) Predictive error decoding unit (anti-quantization means) 23 Motion compensation prediction unit 24 adder (image addition means) 25 loop chopper (image addition means) 26 Frame memory motion vector memory 322764 31 31 201143452 32 Time direct vector generation unit (direct vector generation means) 33 Motion compensation processing unit (predictive video generation means) 32 322764

Claims (1)

201143452 ^ VII. Application for patents: 1. Image coding device, which has: direct vector generation means, for each of the input image_blocks, (4) mixed with the coded image located near the block The action vector generates the time of the direct mode, and the method of predicting the image money is performed by using the time direct vector generated by directly generating the hand-draw and the motion compensation pre-processing to generate the predicted image; a method for lining up the difference between the predicted image generated by the predicted image generating means and the input image, and outputting the quantized coefficient of the difference image; and the variable length worming means 'to be quantized by the above means The output sub-coefficient is variable-length coded, and the coded output of the above-mentioned quantized coefficient is outputted, and the above-mentioned direct vector generation means is a coded image from the constituent time=upper and the flat code object block. In the tile, the tile corresponding to the handle object tile is spatially located at the same position, and the plurality of tiles are located at the center of the tile. Cutting motion vector generation area and the time of the temporal direct mode, the direct vectors. 2. The video coding apparatus according to claim w, wherein the direct=quantization generating means weights the motion vectors of the plurality of divided regions corresponding to the tile size of the plurality of divided regions, thereby generating time Direct vector time direct vector. 3. The image coding apparatus according to claim 2, wherein the direct vector generation means weight-adds the motion vectors of the plurality of divided regions corresponding to the differential motion vectors of the plurality of divided regions, thereby generating the time Direct vector time direct vector. 322764 1 201143452 4. The image coding apparatus according to claim 1, wherein the direct vector generation means corresponds to the coding modes of the plurality of divided regions and the action vectors of the plurality of divided domains are weighted by the direct force mode. Time direct vector. $ 5. A f-image decoding device having: variable-length code data to decode a quantization coefficient; and an anti-quantization means to dequantize a quantized coefficient decoded by the above-described Ϊ variable length access method; direct vector The method of generating the time-directed mode by the action vector of the decoded image temporally located near the decoding target tile is directly directed to the 篁, ^ (4) generating means by using the direct vector generated by the hand The generated time direct vector implements the motion compensation prediction process, and the image and the image addition means add the predicted image generated by the predicted image to the difference image indicating the anti-quantization means: t2, ° fruit' And obtaining a decoded image corresponding to the input image of the image encoding device; and the direct vector generating means is configured to 'specify and decode the target block from the block that constitutes the decoded image temporally located near the decoding target block. In space, the same block:: the tile 'generates the direct mode from the action vector of multiple partitions = at the center of the above tile Time direct vector. 6·如申, the image unwrapping device described in the fifth item of the patent Fan Gu, wherein the straight 2 means corresponds to the block size of the plurality of divided regions, and the actions of the ancient scanning/knife cutting regions are weighted and added. This produces a time direct vector of the time direct mode. 7. The image unwinding device described in claim 5, wherein the straight 322764 2 201143452 2 vector generating means weights the motion vectors of the plurality of divided regions corresponding to the differential action of the plurality of divided regions to t Add time direct mode time direct vector. 8. The image-resolving stone device according to claim 5, the centering, the direct-quantity generating means is corresponding to the coding modes of the plurality of divided regions, and the motion vectors of the plurality of divided regions are weighted and added. Generates a direct vector of time direct mode. 9. The image encoding method has the following steps: a direct vector generation processing step, and a direct vector generation means to hide the motion of the encoded image located near the tile in the time of the transmission time. (4) direct vector; pre-step, the system _ image generation means by using the time direct vector generated by the direct vector rationalization to implement the motion compensation prediction, the raw prediction image; the quantization processing step, is based on the quantization means The predicted image generated by the one-step tying image generation processing step and the differential image of the == image are quantized, and the straightening coefficient of the differential image is rotated; and the variable length encoding processing step 'the variable hand 1 is again The quantum variable length outputted by the above quantization processing step is called the encoding processing for outputting the above-mentioned quantization coefficient. The direct vector generation means implements the direct vector generation of the near-two coding system, which is temporally located in the horse-drawn object map. In the block of the i-picture, the block that is specific to the block of the encoding object is in the space of the eight treasures. The complex number of the heart is the action vector of the π domain and the time of the direct mode is generated. 322764 3 201143452 The vector is connected. The image decoding method has a variable length decoding processing step, wherein the variable length decoding means decodes the quantization coefficient from the encoded data; and the inverse quantization processing step is performed by the variable length decoding method. The quantized coefficient dequantization decoded by the processing step; the direct vector generation processing step is a direct vector generation method that generates a temporal direct vector of time directly from the motion vector of the decoded image temporally located near the decoding target tile. a predictive image generation processing step of predicting a shirt image generation means by performing motion compensation prediction processing using a time direct vector generated by the direct vector generation processing step, and generating a pre-_#^ step, the image addition means will show The predicted image generated by the inverse quantization step is added to the table to obtain a differential image of the inverse quantization result of the phase step; and 'Sheng Cheng: the decoded image of the input image of the second image encoding device 5 ^ ' I direct vector generation means to implement direct vector generation of sound v FIG secret code image is located in the vicinity of the decoding target soil FIG position = the same, and the solution Yun limb like tiles in a space divided area: _ a complex vector located at the center of the tile. ° Straight and time direct mode time directly 322764 4