JP2006074520A - Motion vector detecting device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a motion vector detecting device capable of greatly reducing the calculation amount of motion retrieval by narrowing down a block size beforehand. <P>SOLUTION: In this motion vector detecting device for dividing a target block into prescribed sub-blocks in size to detect at least one motion vector and outputting a division type when the target block is divided into the sub-blocks and the motion vector of the divided sub-blocks as motion information, a division type extracting part (41) extracts division types of a just above adjacent block and a just left adjacent block which are already encoded with respect to the target block and estimates a division type of the target block from the extracted division types of the just above adjacent block and just left block, and a target block motion vector detecting part detects the motion vector of the target block by the estimated division type. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a motion vector detection apparatus in moving picture coding technology.

  First, an overall configuration example (Patent Document 1) of a typical moving image compression encoding apparatus (encoder) used in MPEG (Moving Picture Expert Group) will be briefly described with reference to FIG. The moving image compression coding apparatus includes a transform unit (DCT) 13 that transforms image data into a frequency domain, a quantizer (Q) 14 that quantizes transform coefficients obtained by the transform unit 13, and a quantized signal. A variable length coding unit (VLC) 15 for entropy coding the transform coefficient, a buffer 17 for supplying compressed image data with a variable bit rate to the transmission line according to the transmission rate, a decoder 16, and a motion prediction unit 19 With.

  As shown in FIG. 10, first, image data 10 is input. The difference unit 11 calculates a difference value between the image data 10 and the motion compensated image 12. The motion compensated image 12 is obtained by decoding an already encoded image and performing motion compensation based on the motion prediction data. This is performed by the decoder 16 corresponding to image coding.

The decoder 16 reverses the encoding procedure, that is, the difference data decoded by the inverse quantization unit (Q ⁻¹ ) 20 and the inverse transform unit (IDCT) 21 is converted into a buffer 22 and motion compensation. The unit 23 and an adder 24 for adding the motion compensated images generated by the unit 23 generate a previous image similar to that obtained on the decoding side.

  In motion compensation coding, the motion compensated image of the target image (corresponding to the image data 10) is the target image and the decoded image (the decoded image stored in the buffer 22 corresponds to the reference image). Is generated from the image data of the corresponding decoded image based on the prediction of motion between and.

  The predicted value of motion is expressed by a two-dimensional motion vector indicating the displacement of the pixel between the decoded image and the target image. Usually, motion prediction is performed in units of blocks. That is, in the decoded image, a block having the highest correlation with the block of the target image is set as a motion compensated image. A motion prediction unit 19 that performs this motion prediction and a motion compensation unit 23 that generates a motion compensated image from an image decoded corresponding to the motion vector are incorporated in the encoder.

The moving image compression encoding apparatus shown in FIG. 10 operates as follows. The image of the image signal 10 is usually divided into a certain number of small block groups called macroblocks. For example, the picture (target image) 30 shown in FIG. 11 is divided into a plurality of macroblocks 31. Each macroblock typically has a 16 × 16 pixel size.
Further, the picture is divided into a certain number of slices 32. Each slice is composed of a plurality of macroblocks, and is a unit of synchronous recovery when data is lost. Note that the arrangement of macroblocks constituting one slice does not have to be composed only of macroblocks in the same row as shown in FIG. 11, and includes macroblocks in a plurality of rows or in the middle of rows. Different slices may be provided.

  When video image data is encoded only by reducing spatial redundancy in the image, the resulting picture is called an I picture. An I picture is encoded with reference to only pixel values in the picture. Since the encoded I picture cannot use temporal information for reducing the amount of data, the size of the encoded data is large.

In order to perform efficient compression by taking advantage of temporal redundancy that exists between successive pictures,
Based on motion prediction and compensation, predictive coding is performed between consecutive pictures. When the reference picture selected in motion prediction is one picture that has already been encoded and decoded, it is called a P picture. In addition, when there are two reference pictures (usually, the display order is front and rear with respect to the target picture), it is called a B picture.

  MPEG-4 AVC (Advanced Video Coding), which is the latest video compression encoding system, or H.264. In a standard called H.264, motion compensation in each 16 × 16 macroblock can be performed using different block sizes. Individual motion vectors can be determined for block sizes of 4 × 4, 4 × 8, 8 × 4, 8 × 8, 8 × 16, or 16 × 16 pixels, which are motion compensation block sizes. .

  In the motion compensation process, prediction based on the determined motion vector is performed based on the result of the motion prediction process. Information included in the prediction error block obtained from the predicted block is converted into a conversion coefficient by the conversion unit 13. In general, two-dimensional DCT (discrete cosine transform) is often used. The obtained transform coefficient is quantized, and finally, variable length coding unit 15 performs entropy coding (VLC). Note that the motion vector calculated by the motion prediction unit 19 is used for motion compensation, and is incorporated into the compressed image data 18 via the variable length encoding unit 15 and the buffer 17 and output.

  The transmission stream of the compressed image data 18 is transmitted to a decoder (moving image decoding apparatus), where an encoded video image sequence is reproduced based on the received data. The configuration of the decoder corresponds to the configuration of the decoder 16 included in the moving image compression encoding apparatus shown in FIG. As described above, in the MPEG-4 AVC of the latest moving image compression encoding system, a block having a size smaller than that of a macro block can be selected as a motion compensation block size. This is to improve the coding efficiency by detecting the movement of an object smaller than a macro block.

  FIG. 12 shows the block size of motion compensation in MPEG-4 AVC, and shows how to divide a macroblock having a size of 16 pixels vertically × 16 pixels horizontally. There are four main division methods, which are shown in FIGS.

  FIG. 12A shows a case where motion compensation is performed without dividing a macroblock (division type 1), and FIG. 12B shows that the macroblock is divided into two blocks of 16 vertical pixels × 8 horizontal pixels. FIG. 12C shows a case where motion compensation is performed by dividing a macroblock into two blocks of 8 vertical pixels × 16 horizontal pixels (division type 3). FIG. 12D shows a case where motion compensation is performed by dividing a macroblock into four blocks of 8 vertical pixels × 8 horizontal pixels (division type 4).

  Further, when the division method of division type 4 shown in FIG. 12D is selected, a method of dividing each of the 4 blocks of 8 vertical pixels × 8 horizontal pixels shown in FIGS. 12E to 12H. You can choose from. The division methods shown in FIGS. 12E, 12F, 12G, and 12H are divided into division type 4, division type 5, division type 6, and division type 7, respectively. It corresponds.

  As described above, in MPEG-4 AVC, encoding efficiency can be improved by selecting a plurality of different block sizes as the block size for motion compensation. However, in order to determine which division type is optimal for motion compensation as a macroblock division type, it is necessary to obtain motion vectors for all division types and to determine the optimum division type based on prediction errors in the motion vectors. is there. Therefore, the calculation amount of motion vector search increases.

  For this reason, in MPEG-4 AVC, the amount of calculation required for the above-described multi-partition type motion vector search is reduced as follows. For example, first, a motion vector having a smaller block size is obtained, and a motion vector having a larger block size is estimated from the direction and size of the motion vector having the smaller block size, thereby reducing the number of block sizes for motion search. The amount of calculation is reduced (Non-Patent Document 1).

  As shown in FIG. 13A, when the motion vectors of the sub-block (division type 4) obtained by dividing the block into four are very close, it is estimated that the motion vector approximates the entire block. Therefore, motion search is performed on the entire block (partition type 1) composed of sub-blocks as a block size to be motion compensated.

On the other hand, as shown in FIG. 13B, when the motion vectors of the four sub-blocks (division type 4) vary, it does not indicate one motion vector as a whole block. Therefore, the motion vectors of the four sub-blocks (division type 4) are used as they are as the block size for motion compensation of the four sub-blocks (division type 4).
JP-A-5-236456 Tomoyuki Shimizu, Ikuo Yoneyama, Hiromasa Yanagihara, Yasuhiro Nakajima: “A Study on Speeding Up of Motion Prediction in H.264 Coding” Information Processing Society of Japan (IPSJ SIG Technical Report) 2003-AVM-42 (4) 2003 / 10/3 pp. 23-28

  As described above, in MPEG-4 AVC, motion search using a plurality of block sizes (partition types) can be used to increase the efficiency of motion compensation. However, there is a problem that the calculation amount increases accordingly. On the other hand, as a means for reducing the increase in the calculation amount, there is a method of estimating a motion vector having a larger block size from a motion vector having a smaller block size (for example, division type 4) and specifying the block size. . However, even in that case, a motion vector having a smaller block size must be obtained. In addition, since the smaller block size has a larger number of divided blocks, the amount of calculation for motion search increases, and the reduction is not significantly reduced.

  Therefore, in the present invention, the block size (division type) of the macroblock is estimated from the block sizes (division type) of the already encoded peripheral macroblocks, and the motion search is performed by narrowing down the block size (division type) in advance. An object of the present invention is to provide a motion vector detection device capable of greatly reducing the amount of computation of the above.

When calculating a motion vector of a target block in video encoding, the motion vector detection device according to the present invention divides the target block into a predetermined sub-block size to detect at least one motion vector, A motion vector detection device that outputs a division type when a target block is divided into sub-blocks and a motion vector of the divided sub-blocks as motion information,
Division type extraction means for extracting the division type of the immediately adjacent block and the immediately adjacent block that have already been encoded with respect to the target block;
Division type estimation means for estimating the division type of the target block from the division type of the immediately adjacent block and the immediately adjacent block extracted by the division type extraction means;
Target block motion vector detection means for detecting a motion vector of the target block according to the division type estimated by the division type estimation means.

  According to the motion vector detection device of the present invention, the block size for motion search is estimated by estimating in advance the block size type of the macroblock to be motion-predicted from the block size types of the already-encoded peripheral macroblocks. Can be significantly narrowed down, and the amount of motion search computation can be greatly reduced.

  Before describing each embodiment in detail, first, the basic concept of the motion vector detection device according to the present invention will be described. The motion vector detection apparatus according to the present invention includes a motion information recording unit that records information on a division type and a motion vector of a block that has already been encoded with respect to a target block to be motion-predicted, and the motion information recording unit An adjacent block division type extraction unit that extracts a division type of an immediately adjacent block immediately above and a left adjacent block from the target block, a division type estimation unit that estimates a division type of the target block from the adjacent block division type extraction unit, and the division The motion prediction unit that detects the motion vector of the target block with the division type estimated by the type estimation unit is set as a minimum component.

  An adjacent block motion vector extraction unit that extracts the motion vector of the immediately adjacent block and immediately adjacent block of the target block from the motion information recording unit is added to the minimum component described above, and the target detected by the motion prediction unit If the motion vector of the adjacent block extracted by the adjacent block motion vector extraction unit with respect to the block motion vector is not within a predetermined range, the partition type estimation unit also includes means for changing the previously estimated partition type. Become.

  By configuring in this way, the most likely division type of the target block is estimated from the division type of the adjacent block with respect to the target block, and only motion vectors in the division type are detected, thereby calculating motion search. The amount can be greatly reduced. In addition, when the estimated division type motion vector is significantly different from the motion vector of the adjacent block as a result, the motion vector is detected again after changing to a division type different from the previously estimated division block, If it is determined that the encoding efficiency is higher (the prediction error and the motion vector information are smaller), the changed division type and motion vector can be corrected.

(First embodiment)
The motion vector detection apparatus according to the first embodiment of the present invention will be described with reference to FIGS. 2, 3, 4, 5, 6, 7, 7, 9, and 15. . FIG. 2 is a block diagram conceptually showing the overall configuration of a moving image compression encoding apparatus including the motion vector detection apparatus of the present invention. FIG. 3 is an explanatory diagram for explaining the motion vector detection of the present invention. is there. 4 to 9 are diagrams specifically explaining the division type estimation means that is the core of the present invention. And FIG. 15 is explanatory drawing of the division | segmentation block estimation order in a motion vector detection apparatus. The motion vector detection apparatus according to the present embodiment is realized with the above-described minimum components.

  As shown in FIG. 2, the moving image compression coding apparatus according to the present embodiment has a motion search control unit 25a added to the conventional typical moving image compression coding apparatus shown in FIG. It is configured. Therefore, the features unique to the present invention will be described in detail with emphasis on the motion search control unit 25a added in the present invention. In FIG. 2, first, the image data 10 is divided into macroblock units (16 vertical pixels × 16 horizontal pixels are often used). Thereafter, encoding processing is performed in units of macroblocks. The image data 10 divided into macroblocks (same as what is called the target block above) is differentiated from the motion compensated image 12 by the differentiator 11.

  The difference data obtained by the difference is orthogonally transformed by the transforming unit 13 and quantized by the quantizing unit 14. The orthogonal transform coefficients of the quantized difference data are entropy encoded by the variable length encoding unit 15, temporarily stored in the buffer 17, and output as a bit stream. Further, the orthogonal transform coefficient of the quantized difference data is inversely quantized by the inverse quantization unit 20, and inversely orthogonally transformed by the inverse transform unit 21 and decoded as difference data. Note that the difference data at this time is not the original difference data because irreversible processing is performed by quantization for reducing the amount of data.

  The decoded difference data is added by the adder 24 to the motion compensated image that has been motion compensated by the motion compensation unit 23. The added image is temporarily stored in the buffer 22 as a local decoded image in the encoder as a reference image for later motion compensation or motion prediction. The motion prediction unit 19 calculates a motion vector between the image data 10 (target block) divided into macroblocks from which a motion vector is to be detected and the reference image stored in the buffer 22.

  The motion vector is calculated as follows. For example, for the target block, block matching is performed with an arbitrary block within a certain range of the reference image, the sum of absolute differences between blocks is obtained, and the reference block when the sum of absolute differences is minimized Is a two-dimensional (horizontal, vertical) pixel value size (however, it can be calculated in units of 1/2 pixel or 1/4 pixel using interpolation between pixels). Possible). At this time, as the evaluation value of motion vector calculation, the sum of absolute difference or sum of squares of difference, or an evaluation value obtained by loading the magnitude of the motion vector to these values, or the code when the inter-block difference data is encoded after Hadamard transform In some cases, the minimum length is used as the evaluation value. The motion vector calculated by the motion prediction unit 19 is supplied to the motion compensation unit 23, and an image of the block of the reference image shifted by the motion vector is extracted from the buffer 22 in which the reference image is accumulated by the motion compensation unit 23, and the motion compensation image is extracted. It becomes.

In the embodiment according to the present invention, particularly when the motion vector is calculated by the motion prediction unit 19, the block size of the target block is not fixed, and a plurality of types can be selected as described above with reference to FIG. At this time, the block size is selected by the motion search control unit 25a.
FIG. 3 shows a detailed configuration of the motion search control unit 25a. The motion search control unit 25 a includes a motion information recording unit 40, an adjacent block division type extraction unit 41, and a division type estimation unit 42. In the motion information recording unit 40, an already-encoded macroblock is divided into division type numbers (1 to 7) that are divided into sub-blocks as shown in FIG. 12, and motion vectors ( Horizontal shift size and vertical shift size) are recorded in advance. The range of the already-encoded macroblocks recorded in the motion information recording unit 40 is, for example, as shown in FIG. 4, the macroblock immediately to the left of the macroblock B immediately above the target block to be motion-predicted. 1 slice up to A.

  The adjacent block division type extraction unit 41 extracts the division type numbers of the macroblock immediately above the target block and the macroblock immediately on the left from the already encoded macroblock recorded by the motion information recording unit 40.

  The division type estimation unit 42 estimates the division type of the target block from the division type numbers of the macroblock immediately above the target block and the macroblock immediately to the left extracted from the adjacent block division type extraction unit 41. Specifically, the division type of the target block shown in FIG. 5 is estimated from the combination of the division type number of the macroblock B immediately above the target block and the division type number of the macroblock immediately to the left. This is made a reference table in advance from the spatial division pattern of the target block and the adjacent block.

  For example, when “1” is estimated as the division type of the target block, five types (a), (b), (c), (d), and (e) in FIG. 6 are conceivable. In this case, the division type of the sub-block immediately above the target block and the division type of the sub-block immediately to the left are (1, 1) in the case of FIG. 6 (1, 1), (3, 1) in (b), ( In the case of c), (1, 2), in the case of (d) (3, 1), and in the case of (e), (1, 2). Therefore, as a reference table, the selection of the target block division type 1 is (division type of the immediately above subblock, division type of the subblock immediately to the left) = (1, 1), (1, 2), (3, 1) 3 types.

  Similarly, when “2” is estimated as the division type of the target block, (a) and (b) in FIG. 7 are considered. In this case, the division type of the sub-block immediately above the target block and the division type of the sub-block on the left are (division type of the sub-block immediately above, division of the sub-block on the left) = ((2, 2) 2).

  Similarly, when estimating “3” as the division type of the target block, (a) and (b) in FIG. 8 are considered. In this case, the division type of the sub-block immediately above the target block and the division type of the sub-block on the left are (division type of the sub-block just above, division type of the sub-block on the left) = (3, (3, 3 )).

  When “4” is estimated as the division type of the target block, (a), (b), (c), (d), (e), (f), (g), and (i) in FIG. Conceivable. However, when “4” is designated, a reference table is not necessary. If the division type of the target block is other than “1” to “3”, it may be estimated as “4”.

  In the above example, a case where a 16 × 16 macroblock is divided into a minimum of 8 × 8 is shown. However, when 8 × 8 is selected as the division type, it is desired to specify a division type up to a minimum of 4 × 4. In this case, the estimation may be performed by replacing 16 × 16 in the above example with 8 × 8. That is, the selection of the target block division type 4 can be made in three ways: (division type of the immediately above subblock, division type of the subblock immediately to the left) = (4, 4), (4, 5), (6, 4) It is. Then, the selection of the target block division type 5 is one of (division type of the immediately above subblock, division type of the subblock immediately to the left) = ((5, 5), 5). Further, the selection of the target block division type 5 is one of (division type of the immediately above subblock, division type of the subblock immediately to the left) = (6, (6, 6)). The minimum 4 × 4 block division type 7 is selected in other cases.

  However, in order to estimate a division type whose target block division type is smaller than 4, the division type estimation in the macroblock is executed in the order shown in FIG. In addition, since there is no adjacent block used for division type estimation in the macro block located at the top and left end of the picture, the estimation value of motion vector calculation is obtained for all division types without using the division type estimation. The smallest division type is selected, or the division type is processed as “1”, for example. The division type estimation unit 42 supplies the estimated division type of the target block to the motion prediction unit 19, and the motion prediction unit 19 calculates a motion vector of the block size.

  As described above, in the first embodiment, the division type of the target block can be estimated and narrowed down from the division type of the adjacent block, so that the amount of motion vector calculation can be greatly reduced. In the above embodiment, the division type estimation may be realized on software. In the embodiment, the size of the macroblock and the block size to be divided are specified, but other sizes and different division methods may be used.

(Second Embodiment)
The motion vector detection device according to the second embodiment determines whether or not the estimated division type is appropriate in addition to the function of the motion vector detection device according to the first embodiment described above. When it is determined that the type is not valid, the motion vector is calculated again by changing to another division type. For this purpose, the motion search control unit 25a (FIG. 3) is replaced with a motion search control unit 25b (FIG. 14). Specifically, as shown in FIG. 14, the motion search control unit 25 b has a signal line directly from the motion prediction unit 19 to the division type estimation unit 42 in the motion search control unit 25 a. Therefore, in the motion vector detection device according to the second embodiment, the description of the parts common to the motion vector detection device according to the first embodiment described above will be omitted, and different parts, particularly the motion search control unit 25b. Only will be described.

  In the motion search control unit 25b, the division type of the adjacent block is extracted from the motion information recording unit 40 by the adjacent block division type extraction unit 41, supplied to the division type estimation unit 42, and the division type estimation unit 42 performs the first implementation. The division type of the target block is estimated as in The division type estimated by the division type estimation unit 42 is supplied to the motion prediction unit 19, and the motion vector of the division type (subblock) estimated by the motion prediction unit 19 is calculated.

  A calculated evaluation value used for evaluation of motion vector calculation of the division type (subblock) estimated by the motion prediction unit 19, for example, the difference absolute value sum or the difference square sum of squares, or the value of the motion vector in these values. The division type estimation unit 42 extracts the weighted evaluation value or the code length when the inter-block difference data is encoded after Hadamard transform as a motion vector calculation evaluation value, and is estimated when a predetermined value is exceeded It is determined that the division type is not optimal, and a division type different from the previously estimated division type is estimated and supplied to the motion prediction unit 19 again.

  As the predetermined value, for example, a value obtained by adding a predetermined value to the motion vector calculation evaluation value of the immediately left block of the same division type as the previously estimated division type is set. When there is no same division type in the right block, for example, in the case of (c) in FIG. 6, it is assumed that the size is the same as the previously estimated division type. The motion vector calculation evaluation value of the block is used. For example, if the previous division type is “1”, “2” or “3”, “4” is selected, and if it is “4”, “1” is selected.

  As described above, in the present embodiment, it is determined whether or not the estimated division type is valid. If it is determined that the division type is not valid, the coding efficiency is calculated by changing to another division type and calculating the motion vector again. Can be raised. If the motion vector is calculated with the changed division type and the calculated evaluation value at that time is extracted again by the division type estimation unit 42 and becomes larger than the first motion vector calculation evaluation value, the first division type A motion vector may be selected. In this case, a memory for holding the first division type motion vector and the motion vector calculation evaluation value is separately provided. In the above embodiment, the division type estimation may be realized on software.

(Third embodiment)
In the third embodiment, similarly to the second embodiment, it is determined whether or not the estimated division type is valid. When it is determined that the division type is not valid, another division is performed. It has a function of changing to a type and calculating a motion vector again. For this purpose, the motion search control unit 25b (FIG. 14) is replaced with a motion search control unit 25c (FIG. 1). Specifically, as illustrated in FIG. 1, the motion search control unit 25c includes an adjacent block motion vector extraction unit 43 between the motion information recording unit 40 and the division type estimation unit 42 in the motion search control unit 25b. ing. Therefore, in the motion vector detection device according to the third embodiment, the description of the parts common to the motion vector detection device according to the second embodiment described above will be omitted, and different parts, in particular, adjacent block motion vector extraction will be omitted. Only the unit 43 will be described.

  In the motion search control unit 25 c, the adjacent block division type is extracted by the adjacent block division type extraction unit 41 from the motion information recording unit 40, supplied to the division type estimation unit 42, and the division type estimation unit 42 performs the first implementation. Similar to the form, the division type of the target block is estimated.

  The division type estimated by the division type estimation unit 42 is supplied to the motion prediction unit 19, and the motion vector of the division type (subblock) estimated by the motion prediction unit 19 is calculated. An adjacent block motion vector extraction unit 43 extracts an adjacent block motion vector from the motion information recording unit 40. The adjacent block motion vector extracted by the adjacent block motion vector extraction unit 43 is supplied to the division type estimation unit 42. At the same time, in the connection from the motion prediction unit 19 to the division type estimation unit 42, the motion vector of the estimated division type (subblock) calculated by the motion prediction unit 19 is estimated to the division type estimation unit 42. (Estimated split type motion vector) is fed back.

  The division type estimation unit 42 compares the estimated division type motion vector fed back from the motion prediction unit 19 with the adjacent block motion vector supplied from the adjacent block motion vector extraction unit 43, and determines the estimated division type. When the motion vector and the adjacent block motion vector supplied from the adjacent block motion vector extraction unit 43 are not within the predetermined range, it is determined that the estimated partition type is not optimal, and the partition is different from the previously estimated partition type. The type is estimated and supplied to the motion prediction unit 19 again.

  As the predetermined range, when the direction and / or size is greatly different from the adjacent block motion vector supplied from the adjacent block motion vector extraction unit 43, for example, the direction and / or size is different by 90 degrees or more from any of the adjacent block motion vectors. When the difference is two times or more, the division type is changed and supplied to the motion prediction unit 19. For example, if the previous division type is “1”, “2”, or “3”, “4” is selected, and “4” is selected as “1”.

  As described above, in the present embodiment, it is determined whether or not the estimated division type is valid. If it is determined that the division type is not valid, the motion vector is calculated again after changing to another division type. The encoding efficiency can be increased. If the result of detecting the motion vector again with the changed partition type and feeding it back to the partition type estimation unit 42 is different from the motion vector of the adjacent block rather than the motion vector of the first partition type, the first partition is performed. A type of motion vector may be selected. In this case, a memory for holding the first division type motion vector is separately provided. In the above embodiment, the division type estimation may be realized on software.

  As described above, when calculating a motion vector of a target block in moving image coding, the motion vector detection device according to the present invention divides the target block into a predetermined sub-block size and obtains at least one motion vector. It has a function of detecting and outputting, as motion information, the division type when the target block is divided into sub-blocks and the motion vectors of the divided sub-blocks. Then, the motion vector detection device according to the first embodiment includes a division type extraction unit that extracts a division type of an immediately adjacent block and an immediately adjacent block that have already been encoded with respect to the target block, and a division type A division type estimation unit that estimates the division type of the target block from the division type of the immediately adjacent block and the immediately adjacent block extracted by the extraction unit, and a motion vector of the target block according to the division type estimated by the division type estimation unit And a target block motion vector detecting unit for detecting. Thereby, since the division type of the target block can be narrowed down, the effect of greatly reducing the motion vector calculation amount can be obtained.

  The motion vector detection device according to the second embodiment is different from the motion vector detection device according to the first embodiment in that the division type estimation unit is a motion vector of the target block detected by the target block motion vector detection unit. If the target block motion vector calculation evaluation value used in the calculation evaluation of the target block motion vector and the extracted evaluation value exceeds the predetermined value, the division type previously estimated by the division type estimation unit Is further provided with a division type estimation change unit that can change the value to a predetermined division type. Thereby, when the motion vector detected by the division type previously estimated by the division type estimation unit has a motion prediction error larger than a predetermined magnitude, the coding efficiency is improved by changing to a better division type. be able to.

  The motion vector detection device according to the third embodiment is the same as the motion vector detection device according to the second embodiment. Motion vector extraction unit that extracts the motion vector of the target block, and the division type change unit extracts the motion vector of the detected target block instead of extracting the target block motion vector calculation evaluation value from the target block motion vector detection unit If the motion vector of the target block and the motion vector of the adjacent block are out of a predetermined range, the partition type estimation that can change the partition type previously estimated by the partition type estimation unit to the predetermined partition type. And a changing unit. As a result, if the motion vector detected by the partition type previously estimated by the partition type estimation unit is significantly different from the motion vector of the adjacent block, the coding efficiency can be improved by changing to a better partition type. it can.

  INDUSTRIAL APPLICABILITY The present invention facilitates practical use because motion search can be greatly reduced as a motion vector detection apparatus and method for MPEG-4 AVC, particularly MPEG-4 AVC. It can be used as a motion vector detection device for video encoding of equipment that must be suppressed.

The block diagram which shows the structure of the motion search control part in the motion vector detection apparatus which concerns on the 3rd Embodiment of this invention. 1 is a block diagram showing a configuration of a moving picture compression encoding apparatus incorporating a motion vector detection apparatus according to a first embodiment of the present invention. The block diagram which shows the structure of the motion search control part in the motion vector detection apparatus which concerns on the 1st Embodiment of this invention. 2 is an explanatory diagram of a spatial arrangement example in a picture of a motion prediction target block and an adjacent block used for motion prediction in the motion vector detection device shown in FIG. Explanatory drawing of the spatial arrangement | positioning relationship of the target block of motion prediction in the motion vector detection apparatus, and the adjacent block used for motion prediction Explanatory drawing in the case of estimating the block type of the object block of motion prediction from the block type of an adjacent block in the motion vector detection apparatus shown in FIG. Explanatory drawing in the case of estimating the block type of the object block of motion prediction from the block type of an adjacent block in the motion vector detection apparatus shown in FIG. Explanatory drawing in the case of estimating the block type of the object block of motion prediction from the block type of an adjacent block in the motion vector detection apparatus shown in FIG. Explanatory drawing in the case of estimating the block type of the object block of motion prediction from the block type of an adjacent block in the motion vector detection apparatus shown in FIG. Block diagram showing a configuration of a conventional typical moving image compression encoding apparatus Explanatory drawing about the structural unit of the picture in the conventional moving image compression encoding shown in FIG. Illustration of motion compensation block size in MPEG-4AVC Explanatory drawing of a method for estimating a motion vector of a larger block size from a motion vector of a smaller block size when the target block can perform motion prediction with a plurality of block sizes The block diagram which shows the structure of the motion search control part in the motion vector detection apparatus which concerns on the 2nd Embodiment of this invention. Explanatory drawing of the division | segmentation block estimation order in the motion vector detection apparatus which concerns on the 3rd Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Image data 11 Differentiator 12 Motion compensation image 13 Orthogonal transformation part 14 Quantization part 15 Variable length encoding part 16 Decoder 17 Buffer 18 Bit stream 19 Motion prediction part 20 Inverse quantization part 21 Inverse orthogonal transformation part 22 Buffer 23 Motion compensation Unit 24 adders 25a, 25b, 25c motion search control unit 30 picture 31 macroblock 32 slice 40 motion information recording unit 41 adjacent block division type extraction unit 42 division type control unit 43 adjacent block motion vector extraction unit

Claims (4)

When calculating the motion vector of the target block in moving image encoding, when the target block is divided into a predetermined sub-block size and at least one motion vector is detected, and the target block is divided into sub-blocks A motion vector detection device that outputs the division type and the motion vector of the divided sub-block as motion information,
Division type extraction means for extracting the division type of the immediately adjacent block and the immediately adjacent block that have already been encoded with respect to the target block;
Division type estimation means for estimating the division type of the target block from the division type of the immediately adjacent block and the immediately adjacent block extracted by the division type extraction means;
A motion vector detection device comprising: target block motion vector detection means for detecting a motion vector of the target block according to the division type estimated by the division type estimation means.   The division type estimation unit extracts a target block motion vector calculation evaluation value used for calculation evaluation of the motion vector of the target block detected by the target block motion vector detection unit, and calculates the extracted target block motion vector The movement according to claim 1, further comprising: a division type changing unit capable of changing a division type previously estimated by the division type estimating unit to a predetermined division type when the evaluation value exceeds a predetermined value. Vector detection device. Adjacent block motion vector extraction means for extracting motion vectors of the immediately adjacent block and the immediately adjacent block already encoded with respect to the target block;
Instead of extracting the target block motion vector calculation evaluation value from the target block motion vector detection unit, the division type changing unit extracts a motion vector of the detected target block, and the motion vector of the target block and the adjacent 2. A division type estimation change unit capable of changing a division type previously estimated by the division type estimation unit to a predetermined division type if the motion vector of the block is outside a predetermined range. And a motion vector detection device according to claim 2.   The motion vector detection device according to claim 1, wherein the predetermined sub-block size is a size equal to or smaller than the target block.

Images