JP3597107B2 - Motion vector detection circuit and motion vector detection method - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a motion vector detection circuit and a motion vector detection method for detecting a motion amount and a direction in an image signal, that is, a motion vector.
[0002]
[Prior art]
The motion vector is Movement The motion vector indicates the motion amount and direction of the object. The motion vector is used, for example, when improving the inter-frame coding efficiency in the high-efficiency coding of a video signal (for example, a television signal), or in the television system conversion. It is used to reduce the discontinuity of motion due to the conversion of the number of fields.
[0003]
The method of detecting a motion vector is a method of subdividing one image (frame, field, etc.) into blocks of s pixels × t lines (s and t are integers), and then detecting each block.
[0004]
As a method of detecting such a motion vector, a method generally called a pattern matching method of detecting a motion vector using similarity of a signal pattern between successive images (Japanese Patent Laid-Open No. 55-162683). And a method generally referred to as an iterative gradient method for estimating a motion amount based on a physical correspondence between a signal gradient within one image and a signal difference value between images (Japanese Patent Application Laid-Open No. Sho 55-162684). No. 60-158786, etc.) are well known.
[0005]
Among the iterative gradient methods, there is a method using an initial displacement vector in order to improve the detection accuracy of a motion vector (Japanese Patent Laid-Open No. 62-206980). In the iterative gradient method using the initial displacement vector, one motion vector is selected as an initial displacement vector from the motion vectors of a plurality of blocks that have been detected, and the initial displacement vector is set as a reference point. This is a method of detecting a motion deviation amount of a plurality of blocks to detect a true motion vector.
[0006]
Hereinafter, a method of obtaining the true motion vector of the current field using the initial displacement vector with reference to the previous field will be described with reference to FIGS. Here, FIG. 3 is a functional block diagram of a motion vector detection circuit according to an iterative operation method of executing an iterative gradient operation twice.
[0007]
As shown in FIG. 2, for a detected block (m1, n1) for which a true motion vector is to be obtained, an optimal motion vector is selected from motion vectors detected earlier in time, The selected motion vector is set as an initial displacement vector V0 = (α0, β0). That is, the initial deviation vector selection unit 11 performs block matching of the previous field signal SC and the current field signal SB from the detected motion vectors of the blocks near the detected block stored in the detected vector memory 10. Is selected based on the initial deviation vector V0. A vector V1 corresponding to a deviation from the initial deviation vector V0 is obtained by a gradient method operation by the first gradient method operation unit 12 using the initial deviation vector V0, the previous field signal SC and the current field signal SB, and added. The vector is corrected by the unit 13 (see FIG. 4). Further, the vector V0 + V1, the previous field signal S C and current field signal S A vector V2 corresponding to a deviation from the vector V0 + V1 is obtained by a gradient method operation by the second gradient method operation unit 14 using B, and this vector is corrected by the addition unit 15 (see FIG. 4). The motion vector is output as a motion vector V (= V0 + V1 + V2), and is also provided to the detected vector memory 10 and stored.
[0008]
Note that the gradient method calculation and the like by the iterative gradient method are described in, for example, MPEG of the Institute of Image Information and Television Engineers, and the detailed description thereof will be omitted.
[0009]
In the motion vector detection circuit according to the iterative gradient method, the selection of the initial deviation vector V0 has a great effect on the detection accuracy of the motion vector. That is, when the initial displacement vector V0 is appropriately selected, the gradient calculation can be performed from a value close to the true motion vector of the detected block, and the detection accuracy is improved.
[0010]
Conventionally, the initial displacement vector V0 is determined by pattern matching based on the detected vectors of the blocks around the detected block (normally, eight blocks around the detected block).
[0011]
[Problems to be solved by the invention]
However, in the above-described related art, when selecting an initial deviation vector from a plurality of candidate vectors (detected vectors of peripheral blocks), the initial deviation vector is selected independently for each detected block. Therefore, even if it is originally desired that the initial deviation vectors in the respective detected blocks be the same, such as when the entire image is skewed, the initial deviations in the detected blocks may vary. In some cases, a position vector is selected, and as a result, detection accuracy may be degraded.
[0012]
For example, when the screen pans, the entire screen moves at the same speed in the same direction. In this case, if the image is a flat image, the detected motion vectors of a plurality of blocks around the detected block have the same value, and the pixel values when they are displaced have the same value. When selecting a position vector, a plurality of candidate vectors may have similar values. Therefore, even though the entire screen is moving at the same speed in the same direction, the initial deviation vector selected for each detected block may vary, and such an initial deviation vector may be different. When the gradient calculation is performed, errors will vary for the motion vectors obtained for each detected block.
[0013]
Not only in the case of panning, but also in the case of an image with little motion, the initial deviation vector selected for each detected block may vary.
[0014]
That is, conventionally, without considering the motion of the entire image, since the initial deviation vector is selected for each detected block, for example, it is detected as a vertical motion vector in a horizontal motion image, In some cases, an image of a vertical motion is detected as a horizontal motion vector.
[0015]
Therefore, even when there is a biased motion in the entire image, a motion vector detection circuit and a motion vector detection method capable of appropriately selecting an initial displacement vector (initial vector) for each detected block and improving detection accuracy Is desired.
[0016]
[Means for Solving the Problems]
In order to solve such a problem, a first aspect of the present invention is to divide one image into a plurality of blocks of a predetermined size, and to detect a motion vector from signals of at least two consecutive images for each block. In the vector detection circuit, (1) a detected vector storage unit for storing a motion vector detected for each block for a predetermined time; and (2) a motion vector related to a block around a detected block from which a motion vector is to be detected. candidate vector set switching means for selecting, as a candidate vector set, N (N <n) motion vectors according to given set information from the n detected motion vectors; The closest motion vector to the motion vector of the detected block from the N motion vectors in the candidate vector set One Initial vector selecting means for selecting a motion vector as an initial vector; (4) vector correcting means for performing vector correction on the selected initial vector to detect a final motion vector for the detected block; 5) Obtain statistics of the relative position of the selected block related to the initial vector with respect to the detected block , Based on the statistics obtained, Set information forming means for forming set information to be given to the candidate vector set switching means.
[0017]
According to a second aspect of the present invention, there is provided a motion vector detecting method for dividing one image into a plurality of blocks having a predetermined size, and detecting a motion vector from signals of at least two consecutive images for each block. (1) A motion vector detected for each block is stored for a predetermined period of time, and (2) a motion vector is calculated from n detected motion vectors related to blocks around a detected block to be detected. , N (N <n) motion vectors according to the set information are selected as a candidate vector set, and (3) the N motion vectors in the selected candidate vector set are used as the motion vectors of the detected block. the nearest One A motion vector is selected as an initial vector. (4) Vector correction is performed on the selected initial vector to detect a final motion vector for the detected block. Obtain statistics of the relative position of the block with respect to the detected block , Based on the statistics obtained, It is characterized in that set information defining the candidate vector set is formed.
[0018]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an embodiment of a motion vector detection circuit and a motion vector detection method according to the present invention will be described in detail with reference to the drawings.
[0019]
(A) Overview of the embodiment
In this embodiment, when selecting the initial deviation vector, a plurality of sets of eight candidate peripheral blocks are selected from a wide range of peripheral blocks, for example, 24 blocks as shown in FIG. prepare. Every time the initial displacement vector for the detected block is selected, the relative positions of the peripheral blocks related to the selected initial displacement vector in the eight candidates with respect to the detected block are counted. Perform only during Based on the relative position of the initial displacement vector most frequently selected in one field, one block set suitable for selecting the initial displacement vector is selected from a plurality of block sets (vector sets). It is determined and selected, and the detected vectors of the eight peripheral blocks belonging to the selected block set are used for selecting the initial deviation vector. That is, the set of candidate vectors is switched according to the selection information in the field preceding the initial deviation vector, and the continuity of the image on the time axis in the field direction can be reflected.
[0020]
In this embodiment, the range of the block to be evaluated for motion detection is expanded from the number of candidate vectors for the initial displacement vector, and the relative position selected as the initial displacement vector for one field before the detected block is set. Counts the number of blocks having the same relative position, and preferentially uses the block having the same relative position at the time of selecting an initial deviation vector (block) of each detected block. Even in the case of a signal having a uniform motion as a whole, a motion vector can be detected with high accuracy.
[0021]
(B) Configuration of the embodiment
FIG. 1 is a block diagram showing a configuration of a motion vector detection circuit according to this embodiment. The same reference numerals as in FIG. 3 denote the same or corresponding parts.
[0022]
In FIG. 1, a motion vector detection circuit according to this embodiment includes a detected vector memory 10, an initial deviation vector selection unit 11, a first gradient method operation unit 12, an addition unit 13, In addition to the calculation unit 14 and the addition unit 15, a candidate vector set switching unit 20, an eight candidate counter 21, a set switching determination unit 22, and a one-field holding unit 23 are provided. Although omitted in FIG. 3 according to the related art, FIG. 1 also illustrates a one-field delay unit 16 provided in the input stage.
[0023]
1-field delay unit 1 6 is , The input video signal SA is delayed by one field to form a field signal SC (hereinafter referred to as a previous field signal) one field before the current field signal SB.
[0024]
In this embodiment, a motion vector is detected using the previous field signal SC as a reference field and the current field signal SB as a search field. The motion vector is detected using a block of 8 pixels × 8 lines.
[0025]
The detected vector memory 10, the initial displacement vector selection unit 11, the first gradient method operation unit 12, the addition unit 13, the second gradient method operation unit 14, and the addition unit 15 are the same as those in the related art.
[0026]
That is, based on the block matching of the previous field signal SC and the current field signal SB from the detected motion vectors near the detected block stored in the detected vector memory 10 by the initial deviation vector selection unit 11. The initial deviation vector SG is selected, and the initial deviation vector SG is calculated by the gradient method operation by the first gradient method operation unit 12 using the initial deviation vector SG, the previous field signal SC and the current field signal SB. A vector SH corresponding to the deviation is obtained, and the adder 13 corrects the vector SH. Further, a vector SJ corresponding to a deviation from the vector SI is obtained by a gradient method operation by the second gradient method operation unit 14 using the vector SI (= SG + SH), the previous field signal SC and the current field signal SB, and This corrected vector is corrected by 15 and output as a final motion vector SK (= SI + SJ), and is also provided to the detected vector memory 10 and stored.
[0027]
The candidate vector set switching unit 20 added in this embodiment converts eight candidate vectors near the position of the detected block at the time pointed by the set information SN given from the one-field holding unit 23, It is taken out from the detected vector memory 10 and given to the initial displacement vector selection unit 11 as a candidate vector set SM. An example of the candidate vector set SM will be clarified in the operation description described later (see FIG. 8).
[0028]
Note that the candidate vector is not limited to the one related to the field before the detected block belongs, as long as the candidate vector is already detected, and may be the one related to the field to which the detected block belongs. For example, when the detected block is to be switched in a raster scan in the horizontal direction, for example, a motion vector has already been detected for a block above the detected block.
[0029]
For example, as shown in FIG. 6, the candidate vector set switching unit 20 sets eight candidate vectors from a set SL of n (n is greater than 8) candidate sets from the one-field holding unit 23. It can be realized by a selector circuit that selects according to the information SN.
[0030]
For example, as shown in FIG. 5 described above, out of a total of 25 blocks of 5 × 5 in length and width around the detected block, vectors related to a total of 24 peripheral blocks excluding the detected block are defined as candidate vectors. In this case, n described above is 24. In addition to the vectors related to the surrounding blocks, the position of the detected block in the previous field and the average vector of the detected vectors related to the eight surrounding blocks can also be used as the candidate vector. In this case, n The candidate vector includes its average vector.
[0031]
Here, it is needless to say that the number of candidate vectors in the candidate vector set is not limited to eight. In order to speed up the motion vector detection process, there is already an initial displacement vector selection unit 11 having a hardware configuration that obtains evaluation values for eight candidate vectors in parallel.
[0032]
As described above, the initial displacement vector selection unit 11 selects the initial displacement vector SG from the plurality of candidate vectors and provides the selected initial displacement vector SG to the first gradient method calculation unit 12. In the case of this embodiment, selection information SD defining the relative position of the candidate vector selected as the initial deviation vector SG is also formed and given to the eight candidate counters 21. In other words, the selection information SD is, for example, information indicating the relative position in the block set of the block related to the selected initial deviation vector SG.
[0033]
The 8 candidate counter 21 sets the relative position in the 8 candidate block indicated by the selection information SD every time the initial deviation vector SG is selected, in other words, every time the selection information SD is given, over one field. (The count value is cleared to 0 when the field is switched), and when the detection of all the motion vectors for the field for which the current motion vector is to be detected ends, the count value becomes This is to obtain the most frequent candidate SE indicating the most relative position and give it to the set switching determination unit 22.
[0034]
FIG. 7 shows a detailed configuration example of the eight candidate counter 21. The eight candidate counter 21 includes a candidate decoding unit 30, eight counters 31 to 38, and a magnitude comparing unit 39.
[0035]
The selection information SD from the initial deviation vector selection unit 11 is given to the candidate decoding unit 30, and the candidate decoding unit 30 decodes the selection information SD and sets the candidate information to the relative position indicated by the selection information SD. A trigger signal SDX (X is any of A to H and corresponds to x) for incrementing the counter 3x (x is any of 1 to 8) by 1 is made significant.
[0036]
The eight counters 31 to 38 are provided for each of the relative positions of the eight candidate vectors (see FIG. 8 to be described later). In the example of FIG. ,..., “H candidate”. Each of the counters 31,..., 38 increments by one when the trigger signal SDA,. Although not shown in FIG. 7, when the field to be detected is switched (for example, during a blanking period), the count value is supplied from a timing generation circuit (not shown) to all the counters 31 to 38. Is reset to zero.
[0037]
Although not shown in FIG. 7, the size comparison unit 39 is supplied with a comparison enable signal (not shown) when the field to be detected is switched (for example, during a blanking period). When the comparison enable signal is given, the magnitude comparing section 39 takes out the count values SEA to SEH from all the counters 31 to 38, recognizes the largest count value, and determines the relative position related to the largest count value as the most candidate SE. To the set switching determination unit 22.
[0038]
The set switching determination unit 22 forms the set information (switching signal) SO in the future processing field from the given maximum number of candidate SEs and the set information SN held so far by the one-field holding unit 23, and generates 1 This is given to the field holding unit 23. Note that an example of a method of forming the switching signal SO by the set switching determining unit 22 will be clarified in an operation description described later.
[0039]
The one-field holding unit 23 holds the set information (switching signal) SO given from the set switching determination unit 22 only for a period of a processing field from now on, and stores the held set information SN in the candidate vector set switching unit 20 and This is given to the set switching determination unit 22.
[0040]
(C) Operation of the embodiment
Hereinafter, the operation (motion vector detection method) of the motion vector detection circuit of the embodiment whose configuration is shown in FIG. 1 will be described.
[0041]
By passing the input video signal SB through the one-field delay unit 16, a current field signal SB and a previous field signal SC necessary for detecting a motion vector are prepared. The current field signal SB and the previous field signal SC are provided to an initial displacement vector selection unit 11, a first gradient method operation unit 12, and a second gradient method operation unit 14.
[0042]
When the detected block whose motion vector is to be detected in the current field is new, the candidate vector set switching unit 20 sets a set SM of eight candidate vectors around the detected block and indicated by the set information SN. Is retrieved from the detected vector memory 10 and supplied to the initial displacement vector selection unit 11.
[0043]
The initial deviation vector selection unit 11 selects one of the eight candidate vectors selected by the candidate vector set switching unit 20 as described above based on block matching of the previous field signal SC and the current field signal SB. The initial deviation vector SG is selected, and the initial deviation vector SG is supplied to the first gradient method operation unit 12 and the addition unit 13 and the candidate vector (candidate block) selected as the initial deviation vector SG is selected. Selection information SD indicating a relative position with respect to the detected block is provided from the initial deviation vector selection unit 11 to the eight candidate counters 21.
[0044]
In the first gradient method operation unit 12 to which the initial deviation vector SG is given, the gradient method operation using the initial deviation vector SG, the previous field signal SC and the current field signal SB is executed, and the initial deviation vector A vector SH corresponding to a deviation from SG is obtained, and the adder 13 corrects the vector SH. Further, a vector SJ corresponding to a deviation from the vector SI is obtained by a gradient method operation by the second gradient method operation unit 14 using the vector SI (= SG + SH), the previous field signal SC and the current field signal SB, and The corrected vector is corrected by 15 and is output as the final motion vector SK (= SI + SJ) for the detected block to be processed at the same time, and is given to the already detected vector memory 10 to provide one image (one field). Is stored as the detected vector at that block position in the.
[0045]
On the other hand, in the eight candidate counter 21 to which the selection information SD is given, the count value corresponding to the relative position indicated by the selection information SD is incremented by one.
[0046]
Such a process of detecting a motion vector in units of a block is executed for all blocks in a field to be currently processed.
[0047]
The counters (see 31 to 38 in FIG. 7) corresponding to the respective relative positions of the eight candidate counters 21 are cleared to 0 when a new field is to be processed, and when the field to be processed is completed, each counter is cleared. The number of selections of the relative position selected as the initial deviation vector for the detection block is the count value. When the field to be processed is completed, the most frequent candidate SE indicating the relative position where the number of selections is the largest is supplied from the eight candidate counter 21 to the set switching determination unit 22.
[0048]
In the set switching determination unit 22, the set information (switching signal) SO in the future field is formed from the given maximum number of candidate SEs and the set information SN held so far in the one-field holding unit 23, and 1 The set information (switching signal) SO is provided to the field holding unit 23, and the set information (switching signal) SO is held by the one-field holding unit 23 only for the period of the next field. It is provided to the switching determination unit 22.
[0049]
As a result, in the future fields, eight candidate vector sets determined by the held set information SN are transmitted to the candidate vector set switching unit 20. Therefore Will be selected.
[0050]
Hereinafter, several examples of how to switch the candidate vector set, in other words, what criteria are used for switching will be described.
[0051]
As described above, if the candidate vector is a detected vector, the candidate vector may be related to a field before the detected block belongs, or may be related to a field to which the detected block belongs.
[0052]
As one of the criteria for the method of switching the candidate vector set, when the most frequent candidate is a relative position with respect to the candidate vector (initial deviation vector) related to the previous field, a candidate vector set with more candidate vectors in the previous field is set. If the most frequent candidate is a relative position with respect to a candidate vector (initial deviation vector) related to the current field, a candidate vector set in which the number of candidate vectors of the current field is increased is set. It is also possible to switch the set by combining any number of vectors.
[0053]
At this time, continuity in the field direction (field time axis) may be used as a switching parameter.
[0054]
For example, when the case where the most frequent candidate is related to the candidate vector (initial deviation vector) of the previous field is continuous for a predetermined number of fields, a candidate vector set in which the number of candidate vectors of the previous field is increased, and When the case related to the (initial deviation vector) continues for a predetermined number of fields, a candidate vector set in which the number of candidate vectors for the current field is increased can be used.
[0055]
In addition, as a criterion for the switching method of another vector set, if the most likely candidate of the relative position of the block related to the candidate vector selected as the initial deviation vector is the upper block of the detected block, the upper block Is switched to the candidate vector set with more candidate vectors, and if the most frequent candidate is the lower block of the detected block, the candidate vector of the lower block is switched to the candidate vector set with more candidate vectors. In the case of the block on the left side of the detected block, the candidate vector of the left block is switched to a candidate vector set with more candidate vectors, and in the case where the most candidate is the block on the right of the detected block, the candidate vector of the right block is increased. Switch between top, bottom, left, and right vector sets, such as switching to a set of candidate vectors It is also possible to so that.
[0056]
At this time, as described above, the continuity in the field direction may be used as a switching parameter.
[0057]
FIG. 8 shows an example of a candidate vector set in each of the up, down, left, and right directions when an area including a block that can be a set element is 5 × 5 blocks in the vertical and horizontal directions. In addition, the reference vector set is not biased.
[0058]
In FIG. 8, the candidate vector sets in the up, down, left, and right directions are respectively represented by character strings “upward set”, “downward set”, “leftward set”, and “rightward set”. The reference vector set is represented by a character string “reference set”. In FIG. 8, "*" indicates a detected block. Furthermore, although FIG. 8 shows a set of five types of candidate vectors, it goes without saying that there may be other sets of candidate vectors. For example, there may be a candidate vector set in an upper left direction, an upper right direction, a lower left direction, and a lower right direction. Also, for example, even if the pointing is performed in the same direction, there may be a plurality of candidate vector sets depending on the distance or the like.
[0059]
In FIG. 8, as an initial state, the reference set shown in FIG. 8A is selected, and detection of an initial deviation vector is performed using A to H as candidate vectors. In addition, A to H represent candidate vectors, and also represent relative positions (block names) of blocks related to the candidate vectors.
[0060]
Here, the candidate vectors A, B, C, D, and H are candidate vectors whose relative positions to the detected block are fixed among a plurality of candidate vector sets, and the candidate vectors E, F, and G are the respective candidate vector sets. , Is a candidate vector whose relative position to the detected block changes depending on the most candidates in the previous field.
[0061]
Also, in the case of FIG. 8, as described above, both the detected vector of the field before the field to which the detected block belongs and the detected vector of the current field to which the detected block belongs are included in the candidate vector set. Can be.
[0062]
The relative positions of the fixed candidate vectors (block names) A, B, C, D, and H are as follows. It is assumed that the switching of the detected block is in accordance with the raster shape in the left-right direction.
[0063]
A: Left of current field 2 blocks
B: One block above the current field
C: Right one block before the previous field
D: One block below the previous field
H: average vector of the previous field (average of a total of 9 blocks of the detected position block and surrounding blocks in the previous field)
If the candidate vector H, F, or G whose relative position changes in the vector set becomes the largest candidate as a result of the selection of the initial deviation vector in the reference set, the top, bottom, left, and right cannot be specified. Is used to select an initial displacement vector.
[0064]
An example of a block configuration of the candidate vector H, F or G whose relative position changes will be described below.
[0065]
If the relative position B is the most candidate as a result of the selection of the initial displacement vector in the reference set for one field, in the next field, the initial displacement using the upward set shown in FIG. Make a vector selection.
[0066]
If the relative position D is the largest candidate as a result of the selection of the initial displacement vector in the reference set for one field, in the next field, the initial displacement using the downward set shown in FIG. Make a vector selection.
[0067]
If the relative position A or E is the most candidate as a result of the selection of the initial displacement vector in the reference set for one field, the next field uses the initial set using the leftward set shown in FIG. Select a displacement vector.
[0068]
If the relative position C is the most candidate as a result of the selection of the initial displacement vector in the reference set for one field, in the next field, the initial displacement using the rightward set shown in FIG. Make a vector selection.
[0069]
When the relative position B, E, F, G, or H is the most candidate as a result of the selection of the initial displacement vector in the upward set for one field, the initial field using the upward set again in the next field. Select a displacement vector.
[0070]
If the result of one field selection of the initial displacement vector in the upward direction set indicates that the candidates other than the relative positions B, E, F, G, and H are the most candidates, the next field uses the initial set using the reference set. Select a displacement vector.
[0071]
As in the case of the upward direction set, the result of one field selection of the initial deviation vector in the downward direction set, the leftward direction set, and the rightward direction is the candidate having the largest relative position in the direction from the position of the detected block. In the next field, the initial deviation vector is selected again using the set of the direction in the next field. On the other hand, when the other relative position becomes the most candidate, the next field uses the reference set. Is used to select an initial displacement vector.
[0072]
As described above, for each field, the initial set is set between the reference set, the upper set, the lower set, the left set, and the right set. Deviation The candidate vector set of the vector transitions.
[0073]
Here, the temporal continuity of the largest number of candidates may be used as a parameter for the transition of the candidate vector set, and a transition may be made when the same direction is repeated a plurality of times.
[0074]
The candidate vector set can be determined by verifying data obtained by simulating various moving images in each candidate vector set.
[0075]
(D) Effects of the embodiment
As described above, according to the embodiment, a plurality of candidate vector sets for selecting an initial deviation vector are provided, and the number of selections of the initial deviation vector is counted by one field at each relative position of the eight candidates. By comparing the count values, the most frequent candidate among the eight candidates is detected, and from the characteristics of the most frequent candidate, an optimal candidate vector set used for selecting an initial deviation vector in the next field is selected. Since the motion vector detection is performed by the iterative gradient method using the initial deviation vector detected in the set, even if there is a motion that is biased in the entire image, the initial deviation vector for each detected block can be appropriately set. A motion vector detection circuit and a motion vector detection method which can be selected and which can improve detection accuracy can be realized.
[0076]
That is, in the case of this embodiment, the direct selection of the initial deviation vector is performed from eight candidate vectors, and the selection of the set of eight candidate vectors is also performed from a plurality of sets. As a result, the number of candidate vectors to be used for selecting the initial displacement vector is more than 8 candidates. In other words, the initial displacement vector can be selected from a wide range of blocks. It is possible to select and improve the detection accuracy.
[0077]
In addition, since the combination of the candidate vector set of the initial deviation vector is flexibly changed in accordance with the nature of the motion of the image for each field, that is, the result of selecting the initial deviation vector of the previous field is Since the combination of the initial deviation vector candidate sets is flexibly changed by capturing the characteristics of the motion, it is possible to improve the detection accuracy of the motion vector even when there is a motion that is biased in the entire image.
[0078]
By the way, it is conceivable to improve the detection accuracy by increasing the number of direct candidate vectors for determining the initial deviation vector, but the configuration for selecting the initial deviation vector becomes large, The amount may increase. In this regard, the above embodiment is more preferable.
[0079]
(E) Other embodiments
In the description of the above-described embodiment, various other embodiments obtained by modifying the above-described embodiment have been referred to, but other embodiments as exemplified below can be further included.
[0080]
In the above-described embodiment, an example has been described in which an image signal used for detection of a motion vector follows the interlace method, and a motion vector is detected by using a signal between successive fields. The present invention is not limited to this.
[0081]
For example, the fields used for detecting a motion vector may be two or more fields apart, and the present invention can be applied to the detection of a motion vector of an image signal according to a non-interlace method without a field concept.
[0082]
Further, the various numerical values in the description of the above embodiment are merely examples. For example, the number of candidate vectors for direct selection of initial candidate vectors may be more or less than eight. Further, for example, the size of the region defining the candidate vector set may be larger or smaller than the region of 5 × 5 blocks in length and width. Furthermore, for example, the size of the block , It may be larger or smaller than 8 pixels × 8 lines.
[0083]
Furthermore, in the above embodiment, the candidate vector set is reviewed for each field, but may be reviewed in a longer or shorter period.
[0084]
Furthermore, in the above embodiment, the eight candidate vectors in the same set have no selection priority, but when a new set is determined, the selection priority in each candidate vector in the set is determined. A degree may be determined, or a selection priority may be given in advance to each candidate vector in the set. For example, when the value of the most frequent information for switching to the upward set or maintaining the upward set is very large, the relative position in the upward set selects the upward candidate vector from other candidate vectors. Weighting may be performed so as to facilitate the operation.
[0085]
In the above embodiment, the candidate vector set is determined based on the most information, but the candidate vector set is determined based on the other information or based on the most information and other information. You may do it. For example, the most frequent information may be compared with the next most frequent information, and when the difference is equal to or smaller than a predetermined threshold, the reference set may be selected in the next field. Also, instead of counting the relative position of each block, a plurality of blocks, such as a plurality of blocks in an upward direction, a plurality of blocks in a downward direction, and the like, are set and selected as an initial candidate vector for each set of blocks. The calculated relative positions may be counted.
[0086]
Further, in the above embodiment, the candidate vector set Although the switching unit 20 is shown, when reading from the detected vector memory 10, eight candidate vectors indicated by the set information may be read and provided to the initial displacement vector selection unit 11. Even in this case, functionally, the candidate vector set This is equivalent to the presence of the switching unit 20, and the "candidate vector set switching means" in the claims includes a case where eight candidate vectors indicated by the set information are read from the detected vector memory 10. Shall be.
[0087]
Note that the initial deviation vector constitutes the initial detection vector of the motion vector, and in the above embodiment, the initial motion vector is corrected by applying the iterative gradient method to the final motion vector. Was getting. As is apparent from the above description, the feature of the present invention is until the initial displacement vector (initial detection vector) is obtained, and the vector correction method after obtaining the initial displacement vector (initial detection vector) is an iterative gradient. It is needless to say that the present invention is not limited to the method, and another method such as a block matching method may be applied.
[0088]
【The invention's effect】
As described above, according to the present invention, even when there is a biased motion in the entire image, a motion vector detection circuit and a motion vector detection method capable of appropriately selecting an initial vector for each detected block and improving detection accuracy A vector detection method can be realized.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration of a motion vector detection circuit according to an embodiment.
FIG. 2 is an explanatory diagram showing a correspondence between a current field and a previous field and blocks related to an initial displacement vector.
FIG. 3 is a block diagram illustrating a configuration of a conventional motion vector detection circuit.
FIG. 4 is an explanatory diagram showing a relationship between a motion vector and a detection stage by an iterative gradient method using an initial displacement vector.
FIG. 5 is an explanatory diagram showing the relationship between the size of a search range for an initial deviation vector and a set of initial deviation vector candidate vectors according to the embodiment.
FIG. 6 is a block diagram illustrating a selection configuration of a candidate vector set switching unit according to the embodiment.
FIG. 7 is a block diagram illustrating a detailed configuration example of an eight candidate counter according to the embodiment;
FIG. 8 is an explanatory diagram illustrating an example of a set of candidate vectors according to the embodiment;
[Explanation of symbols]
10: detected vector memory, 11: initial displacement vector selection unit, 12: first gradient method calculation unit, 13, 15: addition unit, 14: second gradient method calculation unit, 16: 1 field delay unit, 20: candidate vector set switching unit, 21: 8 candidate counters, 22: set switching determination unit, 23: 1 field holding unit.

Claims (4)

A motion vector detection circuit for dividing one image into a plurality of blocks of a predetermined size, and for each block, detecting a motion vector from signals of at least two consecutive images,
Detected vector storage means for storing a motion vector detected for each block for a predetermined time,
N (N <n) motion vectors according to the given set information are selected from n detected motion vectors of blocks around a detected block from which a motion vector is to be detected. Candidate vector set switching means for selecting as a candidate vector set;
Initial vector selection means for selecting one motion vector closest to the motion vector of the detected block from the N motion vectors in the selected candidate vector set as an initial vector;
Vector correction means for performing vector correction on the selected initial vector to detect a final motion vector for the detected block,
Set information forming means for obtaining a statistic of a relative position of the block relating to the selected initial vector with respect to the detected block, and forming set information to be given to the candidate vector set switching means based on the obtained statistic. A motion vector detection circuit characterized by the following. The set information forming means converts the set information so that the N motion vectors provided to the initial vector selection means include a large number of motion vectors already detected in the direction from the detected block at a relative position having a large statistic. The motion vector detection circuit according to claim 1, wherein the motion vector detection circuit is formed. In a motion vector detection method of dividing one image into a plurality of blocks of a predetermined size, and detecting a motion vector from signals of at least two consecutive images for each block,
The motion vector detected for each block is stored for a predetermined time,
N (N <n) motion vectors according to the set information are selected as a candidate vector set from n detected motion vectors of blocks around a detected block from which a motion vector is to be detected. And
Of N motion vectors in the selected candidate vector set, a single motion vector closest to the motion vector of the detected block is selected as the initial vector,
Vector correction is performed on the selected initial vector to detect the final motion vector for the detected block,
A motion vector detection method comprising: obtaining a statistic of a relative position of a block relating to the selected initial vector with respect to a detected block; and forming set information defining the candidate vector set based on the obtained statistic. Method. Set information is formed such that motion vectors that have been detected in a direction from a detected block at a relative position having a large statistic are included in the N motion vectors used for the selection of the initial vector in large numbers. The motion vector detection method according to claim 3.

Images