KR101242560B1 - Device and method for adjusting search range - Google Patents

Abstract

An apparatus and method for adjusting a search area is disclosed. The encoding apparatus includes: a motion estimation unit for performing motion estimation on a default search region using a block size preset for each reference frame; And a search region adjusting unit which adjusts the size of the search region using the motion vector value calculated by the motion estimation. According to the present invention, the video encoding time can be shortened in all devices using the H.264 / AVC encoder.

Description

Device and method for adjusting search range

The present invention relates to an encoder, and more particularly, to an apparatus and a method for adjusting a search region.

Due to the development of various multimedia devices and the generalization of digital multimedia broadcasting service, the demand for high quality video service is increasing.

To this end, efficient image compression technology is required for processing high-definition video, and standardization work on video compression technology has been carried out by organizations such as Moving Picture Experts Group (MPEG) and International Telecommunications Union-Telecommunication Standardization Sector (ITU-T). It's going on. Video compression standards established by these organizations to date include MPEG-1, MPEG-2, MPEG-4, H.261, H.263, and H.264 / AVC.

Here, H.264 / AVC defines the base technologies for high-definition video compression in next-generation multimedia devices, and shows superior performance compared to existing compression standards. For example, it has a superior compression ratio of 50% compared to MPEG-4 simple profile and 70% compared to H.263, and enables various multimedia services to be provided even with a small amount of data transmission. Thus, H.264 / AVC is a video compression standard that can meet the industry's desire to export more data in high quality with limited bandwidth.

However, H.264 / AVC requires up to 20 times more computation than the existing compression standards, making it difficult to process video in real time.

FIG. 1 illustrates a variable block size for inter prediction, FIG. 2 illustrates a plurality of reference frames for inter prediction, and FIG. 3 illustrates a search range for inter prediction. Drawing.

As shown in FIGS. 1 and 2, H.264 / AVC performs motion estimation by dividing a macroblock into various variable block sizes when encoding a macroblock, and performing motion estimation. Multiple reference frames are used for forward prediction and backward prediction at the time.

Motion estimation is to obtain the position change of the current image and the reference image due to the movement of the object in the video or the movement, enlargement and reduction of the camera in the form of a motion vector. In order to obtain a motion vector, it is first necessary to determine whether the motion estimation is to be performed in pixels or in blocks. In motion picture compression, motion estimation is mainly used in block units. Accordingly, a block matching algorithm may be referred to as a technique for estimating motion vectors of a current picture and a reference picture in units of blocks.

The range of the search range in the reference frame at the time of motion estimation is determined by the user at the time of encoding. In this case, the defined range may be referred to as a default maximum search range as shown in FIG. 3.

The block matching algorithm finds the block most similar to the macroblock of the current picture within a predetermined search area of the designated reference picture, that is, the default search area that is maximum. That is, it estimates how much the current image has moved from the designated reference image in macroblock units, and the size of the moved position is a motion vector.

However, the variable block size and multiple reference frames of H.264 / AVC, which are used during inter prediction to increase coding efficiency, cause enormous complexity of the encoder. There is this.

In addition, in the case of full search, there is a problem that the complexity is seriously increased due to a comparison operation with blocks located at all pixel positions within the default maximum search range.

An object of the present invention is to provide an apparatus and a method for adjusting a search region, which can reduce a video encoding time in all devices using an H.264 / AVC encoder.

In addition, the present invention is to provide an apparatus and method for adjusting the search region that can significantly reduce the computational complexity by resetting the size of the search range to perform the motion estimation of the macroblock small; will be.

Other objects of the present invention will be readily understood through the following description.

According to an aspect of the present invention, there is provided an encoding apparatus, comprising: a motion estimator for performing motion estimation on a default search region using a block size preset for each reference frame; And a search region adjusting unit for adjusting the size of a search range by using the motion vector value calculated by the motion estimation, wherein the motion estimating unit is a sub-partition of the adjusted search region. There is provided an encoding apparatus, which performs motion estimation on a block size corresponding to n).

에 의해 결정된 영역 경계값(RB)을 이용하여 상기 탐색 영역의 크기를 조정할 수 있다.The search area adjusting unit is

The size of the search region may be adjusted using the region boundary value RB determined by.

The adjusted size of the search area r may be a value within a range from 0 (zero) to the size d of the default search area.

If the reference frames are n (random natural numbers) and the current frame is the t-th frame, the search region adjustment unit performs one search region adjustment and 8x8 according to the motion estimation of the 16x16 block for the reference frame that is the t-1th frame. Four search area adjustments are performed according to the motion estimation of the block, and the search area adjustment unit adjusts each one search area according to the motion estimation of the 16x16 block for each of the reference frames that are frames t-2 to tn. Only can be done.

The search region adjustment unit may perform one search region adjustment for each of the n (arbitrary natural numbers) reference frames according to the motion estimation of the 16x16 block.

For each of the n (arbitrary natural numbers) reference frames, the search region adjustment unit may perform one search region adjustment according to the motion estimation of the 16x16 block and four search region adjustments according to the motion estimation of the 8x8 block, respectively. .

If the reference frames are n (random natural numbers) and the current frame is the t-th frame, the search region adjustment unit performs one search region adjustment based on the motion estimation of the 16x16 block with respect to the reference frame that is the t-1 th frame. The search region adjusted by the search region adjustment may be commonly applied to reference frames that are frames t-2 to tn.

The size of the search region adjusted for each of the reference frames that are the t-th to t-nth frames may not be consistent.

The encoding device may be an H.264 / AVC encoder.

According to another aspect of the present invention, in the search region adjusting method performed in the encoding apparatus, performing motion estimation on a default search region using a block size preset for each reference frame (a ); (B) adjusting a size of a search range by using a motion vector value calculated by the motion estimation; And (c) performing a motion estimation on each block size corresponding to a sub-partition with respect to the adjusted search region.

에 의해 결정된 영역 경계값(RB)을 이용하여 조정될 수 있다.The size of the search area is

It can be adjusted using the region boundary value RB determined by.

The adjusted size of the search area r may be a value within a range from 0 (zero) to the size d of the default search area.

If the reference frames are n (random natural numbers), and the current frame is the t-th frame, in step (a), one search area adjustment is performed according to the motion estimation of the 16x16 block with respect to the t-th frame. Four search region adjustments may be performed according to the motion estimation of the 8x8 block, and search region adjustments according to the motion estimation of the 16x16 block may be performed on each of the reference frames that are the t-2 th frame to the tn th frame.

In step (a), one search region adjustment may be performed for each of the n (arbitrary natural numbers) reference frames according to the motion estimation of the 16 × 16 block.

For each of the n (arbitrary natural numbers) reference frames, one search region adjustment according to the motion estimation of the 16x16 block and four search region adjustments according to the motion estimation of the 8x8 block may be performed in step (a), respectively. have.

If the reference frames are n (random natural numbers) and the current frame is the t-th frame, in step (a), one search area adjustment is performed according to the motion estimation of the 16x16 block for the t-th frame. The search region adjusted by the search region adjustment may be commonly applied to reference frames that are frames t-2 to tn.

The encoding device may be an H.264 / AVC encoder.

Other aspects, features, and advantages other than those described above will become apparent from the following drawings, claims, and detailed description of the invention.

According to the embodiment of the present invention, there is an effect of reducing the video encoding time in all devices using the H.264 / AVC encoder.

In addition, it is possible to significantly reduce the computational complexity by resetting the size of the search range to make the motion estimation of the macroblock small.

1 illustrates a variable block size for inter prediction.
2 illustrates a plurality of reference frames for inter prediction.
3 is a diagram illustrating a search range for inter prediction according to the prior art;
4 is a diagram schematically showing a configuration of a motion estimation unit according to an embodiment of the present invention.
FIG. 5 is a diagram illustrating an adjusted search range for inter prediction according to an embodiment of the present invention. FIG.
6 is a diagram illustrating a motion estimation concept for each frame unit for adjusting a search area according to an embodiment of the present invention.

The present invention is capable of various modifications and various embodiments, and specific embodiments are illustrated and described in the drawings. However, this is not intended to limit the present invention to specific embodiments, it should be understood to include all transformations, equivalents, and substitutes included in the spirit and scope of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

The terms first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

In addition, the terms “… unit”, “… group”, “… unit”, “… module”, “… block”, etc. described in the specification mean a unit that processes at least one function or operation. It may be implemented in software or a combination of hardware and software.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. Referring to the accompanying drawings, the same or corresponding components are denoted by the same reference numerals, .

4 is a diagram schematically illustrating a configuration of a motion estimation unit according to an embodiment of the present invention, and FIG. 5 is a diagram illustrating an adjusted search range for inter prediction according to an embodiment of the present invention. 6 is a diagram illustrating a motion estimation concept for each frame unit for adjusting a search area according to an embodiment of the present invention.

Referring to FIG. 4, the motion estimation unit 400 may include a motion estimation unit 410 and a search region adjustment unit 420.

The motion estimator 410 performs motion estimation on a default maximum search range 510 for each reference frame.

The search region adjusting unit 420 newly defines a search region for the subdivided mode by using the motion estimation result value of the motion estimation unit 410. The newly defined search area 520 may be the default search area 510 that is 0 to the maximum, as shown in FIG. 5, and the size of the search area 520 may be expressed as 0 ≦ r ≦ d. have.

When the search area is adjusted by the search area adjuster 420, the motion estimator 410 targets a newly defined search area 520 instead of the maximum default search area 510. Motion estimation is performed for 16x8, 8x16, and so on.

Hereinafter, a process in which the motion estimation unit 400 readjusts a search range for more effective processing will be described with reference to related drawings.

The motion estimation unit 400 according to the present embodiment generates a motion estimation area by reducing a range of a search range corresponding to each reference frame for a macro block. It suggests ways to reduce the complexity.

In this case, due to the temporal correlation between the plurality of reference frames and the current frame, a reference frame that is temporally adjacent to the current frame is frequently selected as the final reference frame. A discrimination technique is applied when adjusting the search region between reference frames. For example, if five reference frames are applied, the t-1th frame and other reference frames (that is, the t-2th frame to the t-5th frame) based on the tth frame, which is the current frame, are applied. The adjustment method of the search area is applied differently.

A selection ratio for five reference frames in a foreman sequence is shown as a table below. Although the case where the number of reference frames is five is described by way of example herein, it is natural that the reference frames may be set to n (any natural number).

The selection ratio of five reference frames in the Foreman sequence Qp Reference frame number ( t -1) -th ( t -2) -th ( t -3) -th ( t -4) -th ( t -5) -th 22 64.96 13.77 10.51 5.44 5.32 28 75.43 10.37 7.83 3.21 3.15 34 88.12 5.64 3.81 1.12 1.31 Average 76.17 9.93 7.38 3.26 3.26

As shown in Table 1, it can be seen that the t-1 th frame, which is a reference frame temporally adjacent to the current frame, has the highest selection frequency of about 76%. Qp in Table 1 represents a quantization parameter.

In addition, in the H.264 / AVC standard, when encoding one macroblock, motion estimation is performed by dividing each macroblock into 16x16, 16x8, 8x16, 8x8, 8x4, 4x8, and 4x4 sizes, as shown in Table 2 below. In the case of non-contiguous reference frames (that is, frames t-2 to t-5), the proportion of 16x16 divisions to be selected as the final encoding mode is significantly high, on average about 64%.

The selected ME block type distribution in the Foreman sequence (for the (t-2) -th through the (t-5) -th reference frames) Qp ME block type 16x16 16 x 8 8 x 16 P8x8 22 50.86 15.31 16.83 17.00 28 63.54 12.92 14.99 8.55 34 76.07 9.88 10.83 3.22 Average 63.49 12.71 14.22 9.59

On the other hand, in the case of the t-1 th frame that is adjacent to the current frame in time, as shown in Table 3 below, when the low quantization parameter Qp is encoded, the selection ratio of the 8x8 mode increases. For example, if Qp is 22, the selection ratio of 8x8 mode is about 27%, which is relatively higher than other modes.

The selected ME block type distribution in the Foreman sequence (for the (t-1) -th reference frame) Qp ME block type SKIP 16x16 16 x 8 8 x 16 P8x8 22 20.44 26.60 12.37 13.40 27.19 28 41.46 26.54 9.96 10.69 11.36 34 60.58 23.72 6.12 6.13 3.44 Average 40.83 25.62 9.48 10.07 14.00

By using the above-described statistical characteristics of Tables 2 and 3, as shown in FIG. 6, the reference frames ranging from the t-2 th frame to the t-5 th frame are subjected to the motion estimation result of the 16x16 block in the subdivided mode. Take advantage of. However, the motion estimation result of the 16x16 block is applied to the 16x8 and 8x16 partition modes and the motion estimation result of the 8x8 block is applied to the 8x4, 4x8, and 4x4 partition modes, respectively.

The motion estimation result value for each reference frame is used to adjust a search range for each reference frame, and the adjusted search region 520 is used as a search region for motion estimation in the subdivided mode.

Therefore, motion estimation of a 16x16 block is performed on each reference frame. Since the motion estimation is performed separately for each reference frame, the motion estimation result for adjusting the search region for each reference frame may be different for each reference frame. In addition, the size of the search region determined by this may be different for each reference frame.

As described above, the range of the search region that can be adjusted differently for each reference frame may be determined as any value (that is, 0 ≦ r ≦ d) from the size of the default search region 510 which is zero to maximum.

At this time, the adjusted search area is redefined by the range boundary (RB) as described in Equation 1 below, and the area boundary value RB is inferred from the motion estimation result value.

In order to obtain the region boundary value (RB), first, a motion vector (MV) having a minimum rate-distortion (RD) cost in the default search region 510 that is maximum for a given block for each reference frame is obtained. .

In this case, the RD function used may be the same as a function commonly used in H.264 / AVC as illustrated in Equation 1, and when MV _x and MV _y having a minimum cost function are determined, The region boundary value is determined as the maximum value of the absolute values. MV _x and MV _y may have negative values according to the direction of travel, but in order to obtain region boundary values, the absolute values thereof are calculated and applied.

When the region boundary value is determined for the reference frame, the motion estimation process for the remaining blocks is performed in the newly adjusted search region 520. In FIG. 5 an adjusted search area 520 is shown. 5 d represents the size (eg, 16) of the search area designated or defaulted by the user, and r represents the size of the search area adjusted by the above-described process. If we assume that the motion vector value of the 16x16 block is (0,0), the size r of the adjusted search area will be 0, which means that motion estimation is performed for one pixel which is the origin.

Hereinafter, a method of estimating a motion and determining a size of a search area according to the relationship between the current frame and each reference frame will be described.

First, in the case of a reference frame corresponding to the t-2 th to t-5 th frames, motion estimation for a 16x16 block is performed for each reference frame. In this case, the range of the search area may be determined by a value specified by the user (eg, d = 16).

When motion estimation is completed for the corresponding search area, the most similar block may be found and the position (ie, the motion vector value) of the corresponding block may be identified. As shown in Equation 1, the region boundary value RB is determined using the motion vector value, and the range of the search region is adjusted using the range determination value. The adjusted range r of the search area is adjusted to a value of 0 ≦ r ≦ d by the area boundary value.

Subsequently, motion estimation is performed on sub-partitions 16x8, 8x16, P8x8, and the like for the search region adjusted according to the corresponding reference frame. At this time, since the motion estimation is performed only for the range of the previously adjusted search region, the complexity of motion estimation can be significantly reduced.

Next, in the case of the reference frame corresponding to the t-1 th frame, as described above, the sub-partitions 16x8 and 8x16 within the range of the search region adjusted using the motion estimation result value for the 16x16 block as described above. Perform motion estimation for. Then, for 8x8 blocks, motion estimation is performed on the search area (that is, the maximum default search area 510) determined by a user-specified value (for example, d = 16), and the resulting value is used. Apply to 8x4, 4x8 and 4x4 blocks that are sub-partitions within the range of the adjusted search area.

As described above, in the case of the t-2 th frame to the t-5 th frame, one search area is adjusted as a result of the motion estimation for the 16x16 block for each frame. Since the motion estimation is performed for each frame, the size of the search area determined for each frame may be different.

However, in the case of the t-1 < th > frame, up to 5 search area adjustments can be made. That is, one search region is adjusted as a result of motion estimation for 16x16 blocks, and four search region adjustments are made as a result of motion estimation for 8x8 blocks in consideration of the case where the quantization parameter is low. This is because four 8x8 blocks exist in one macroblock as shown in FIG.

As described above, in the method for adjusting the motion search region according to the present embodiment, the region boundary value RB is determined using the motion vector value of the 16x16 or 8x8 block. Table 4 below shows the block types for motion estimation selected to determine region boundary values.

The selected ME block type for determining the RB value Reference
frame number Block partition type 16 × 16 16 × 8 8 × 16 P8 × 8 8 × 8 8 × 4 4 × 8 4 × 4 (t-2) -th to (t-5) -th 16 × 16ME with J _{16 × 16 ME} 16 × 16 ME with J _{16 × 16 ME} (t-1) -th 16 × 16ME with J _{16 × 16 ME} 8 × 8 ME with J _{8 × 8 ME}

As can be seen from Table 4, the position of the motion vector (MV) of the 16x16 block is used to calculate the region boundary values for the divided blocks of the block.

However, in the case of the t-1th reference frame, the motion vector of the 8x8 block calculates the region boundary values of 8x8 partition blocks (for example, 8x4, 4x8, and 4x4). It is additionally used to. The pseudo code according to this embodiment is shown in Table 5.

The procedure of the proposed adaptive search range method Algorithm  One adaptive search range method for each reference frame do
if the reference frame is the (t-1) -th frame
for each MB do
if ME is 16 × 16ME
maximum_search_range = [(2 × SR + 1) × (2 × SR + 1)]
MV_position = integer ME (maximum_search_range)
16 × 16_range_boundary = compute_range boundary (MV_position)
else if ME is 16 × 8ME or 8 × 16ME
maximum_search_range = 16 × 16_range boundary
MV_position = integer ME (maximum_search_range)
else if ME is 8 × 8ME
maximum_search_range = [(2 × SR + 1) × (2 × SR + 1)]
MV_position = integer ME (maximum_search_range)
8 × 8_range boundary = compute_range boundary (MV_position)
else
maximum_search_range = 8 × 8_range boundary
MV_position = integer ME (maximum_search_range)
end if
end
else
for each MB do
if ME is 16 × 16ME
maximum_search_range = [(2 × SR + 1) × (2 × SR + 1)]
MV_position = integer ME (maximum_search_range)
16 × 16_range boundary = compute_range boundary (MV_position)
else
maximum_search_range = 16 × 16_range boundary
MV_position = integer ME (maximum_search_range)
end if
end
end if
end

In order to evaluate the apparatus and method for adjusting a search region according to the present embodiment, the present inventors have implemented and experimented with the technical idea of the H.264 / AVC encoder. In this experiment, full search was used as the motion estimation (ME) method, but it is natural that other motion estimation methods using the search area SR such as hexagon can be adjusted according to the present embodiment.

For the experiment, four quantization parameters and five reference frames were used, and the size d of the default maximum search range was set to ± 16 in the horizontal and vertical directions, respectively.

For objective evaluation, the average number of search points for one macroblock is calculated and shown in Table 6. In Table 6, N _sp represents the average value of the number of search points in the search region, and S _SR represents the size of the search region.

The comparison of average N _SP and S _SR in the Foreman sequence Qp Reference frame number Proposed method FSR method A / B
(%) N _SP- (A) S _SR N _SP- (B) S _SR 22 (t -1) -th 169.15 6.00 1089 16 15.53 ( t -2) -th 172.29 6.06 1089 16 15.82 ( t -3) -th 247.74 7.37 1089 16 22.75 ( t -4) -th 293.71 8.07 1089 16 26.97 ( t -5) -th 331.55 8.60 1089 16 30.45 28 ( t -1) -th 166.62 5.95 1089 16 15.30 ( t -2) -th 175.80 6.13 1089 16 16.14 ( t -3) -th 251.69 7.43 1089 16 23.11 ( t -4) -th 297.97 8.13 1089 16 27.36 ( t -5) -th 337.00 8.68 1089 16 30.95 34 ( t -1) -th 165.05 5.92 1089 16 15.16 ( t -2) -th 181.90 6.24 1089 16 16.70 ( t -3) -th 253.69 7.46 1089 16 23.30 ( t -4) -th 299.65 8.16 1089 16 27.52 ( t -5) -th 338.53 8.70 1089 16 31.09 Average 22.54

As shown in Table 6, the method according to the present embodiment shows a reduction of about 78% in the number of average search points against the full search range (FSR). In addition, the size of the search region is also significantly reduced, so that the superiority of the efficiency can be predicted.

In addition, to evaluate the overall performance of the encoder, we measured the bit increase and decrease (B-%), PSNR increase and decrease (PSNR-dB), encoding time increase and decrease (ET-%), and ME time increase and decrease (MET-%). Is shown in Table 7.

Encoder performance evaluation of the proposed method compared to the FSR method in the AVC / H.264 JM encoder Sequence △ B (%) △ PSNR (dB) △ ET (%) △ MET (%) Akiyo -0.05 0.00 -47.51 -61.64 Bus 1.00 -0.02 -39.35 -44.30 Coastguard 0.24 -0.01 -52.49 -59.03 Container 0.03 -0.02 -55.23 -66.63 Foreman 0.26 -0.01 -47.84 -56.30 Hall 0.83 -0.02 -49.37 -62.32 Mobile 0.22 -0.01 -62.16 -72.50 Mother & daughter 0.13 -0.02 -49.21 -60.63 News 0.01 -0.02 -49.21 -61.97 Silent 0.40 0.00 -49.76 -62.16 Stefan 0.52 -0.01 -45.26 -52.25 Average 0.33 -0.01 -49.76 -59.98

As shown in Table 7, the encoder using the method according to the present embodiment increased the encoding time by about 50% with 0.33% bit increase and 0.01 dB PSNR decrease, compared to the conventional H.264 / AVC encoder. The 60% ME time was also reduced.

In addition, the results of further performing the comparative experiment of Table 7 for Full HD data are shown in Table 8.

Encoder performance evaluation of the proposed method compared to the FSR method in the AVC / H.264 JM encoder (Full HD) Sequence △ B (%) △ PSNR (dB) △ ET (%) △ MET (%) Basketballdrive 0.31 -0.03 -65.21 -66.50 BQTerrace -0.04 -0.02 -68.02 -69.57 Cactus 0.38 0.00 -66.72 -68.37 Parkjoy 0.32 0.00 -53.62 -69.93 Parkscene 0.17 0.00 -67.72 -69.16 Peopleonstreet 0.14 -0.02 -68.64 -70.32 Traffic 0.01 0.00 -67.58 -69.71 Average 0.18 -0.01 -65.36 -69.08

In comparison with the Full HD data, the results were superior to those shown in Table 7, such as a reduction of 66% in encoding time and 69% in ME time.

As described above, the encoder implementing the search region adjusting method according to the present embodiment not only works well with a moving video such as mobile, but also has an advantage of more effectively functioning in a Full HD system.

Up to now, when the current frame is the t-th frame, the search region adjustment unit 420 performs one search region adjustment based on the motion estimation of the 16x16 block and four times the motion estimation of the 8x8 block with respect to the reference frame that is the t-1 th frame. The search region adjustment unit 420 performs search region adjustment, and the search region adjustment unit 420 determines each of the reference frames that are t-2 th to tn th frames among n (arbitrary natural numbers) individual reference frames according to the motion estimation of the 16x16 block. A description has been given focusing on an embodiment of performing one search area adjustment. This is summarized in Table 9 below.

16x16 16 x 8 8 x 16 8x8 8 x 4 4x8 4x4 t-1 MSR - ASR1 ASR1 MSR - ASR2 ASR2 t-2 MSR - ASR3 ASR3 t-3 MSR - ASR4 ASR4 t-4 MSR - ASR5 ASR5 t-5 MSR - ASR6 ASR6

That is, as shown in Table 9, after performing motion estimation within a maximum search range (MSR, Maximum Search Range) range maximum for each reference frame, and within the adjusted search range (ASR, Adaptive Search Range) range Perform motion estimation on other blocks. However, for the reference frames immediately adjacent in time, search area adjustments according to motion estimation of 8x8 blocks are performed again.

For example, as shown in Table 9, in the case of the t-2 th reference frame, the search area ASR2 adjusted by the motion estimation of the 16x16 block is commonly applied to the other blocks, but t-1 In the case of the first reference frame, the search area adjustment (ASR1) according to the motion estimation of the 16x16 block and the search area adjustment (ASR2) according to the motion estimation of the 8x8 block are performed, respectively, and the motion of the block size corresponding to the sub-partition is performed. Used as a basis for estimation.

In another embodiment, when the current frame is the t-th frame, the search region adjusting unit 420 may perform each one search region adjustment according to the motion estimation of the 16x16 block in each of the n (arbitrary natural numbers) individual reference frames. It may be. In this case, in the above-described Table 9, the motion estimation of the block size corresponding to the sub-partition of one search area ASR1 adjusted in the same manner as the t-2 th reference frame also applies to the t-1 th reference frame. Will be used as a standard for This is shown in Table 10 below.

16x16 16 x 8 8 x 16 8x8 8 x 4 4x8 4x4 t-1 MSR - ASR1 ASR1 t-2 MSR - ASR3 ASR3 t-3 MSR - ASR4 ASR4 t-4 MSR - ASR5 ASR5 t-5 MSR - ASR6 ASR6

In another embodiment, when the current frame is the t-th frame, the search region ASR is determined by performing one search region adjustment based on the motion estimation of the 16x16 block only for the t-1 th reference frame, and ASR is determined. It may also be used as a coordinated search area in common for all other reference frames. In this case, the ASR is used as a common search region for motion estimation of each block size corresponding to the t-th reference frames to the t-nth reference frames and sub-partitions. This is shown in Table 11 below.

16x16 16 x 8 8 x 16 8x8 8 x 4 4x8 4x4 t-1 MSR - ASR

ASR t-2
ASR t-3 t-4 t-5

In another embodiment, for each of the n reference frames, the search region adjusting unit 420 performs one search region adjustment according to the motion estimation of the 16x16 block and four search region adjustments according to the motion estimation of the 8x8 block, respectively. You may. This is shown in Table 12.

16x16 16 x 8 8 x 16 8x8 8 x 4 4x8 4x4 t-1 MSR - ASR1 ASR1 MSR - ASR2 ASR2 t-2 MSR - ASR3 ASR3 MSR - ASR4 ASR4 t-3 MSR - ASR5 ASR5 MSR - ASR6 ASR6 t-4 MSR - ASR7 ASR7 MSR - ASR8 ASR8 t-5 MSR - ASR9 ASR9 MSR - ASR10 ASR10

In addition, it is natural that methods and criteria for adjusting the search region for motion estimation of each reference frame and each sub-partition may vary.

It is apparent that the above-described search region adjusting method may be performed by an automated procedure according to time series order by a software program or the like embedded in the encoder. The codes and code segments that make up the program can be easily deduced by a computer programmer in the field. In addition, the program is stored in a computer readable media, and read and executed by a computer to implement the method. The information storage medium includes a magnetic recording medium, an optical recording medium, and a carrier wave medium.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the invention as defined in the appended claims. And changes may be made without departing from the spirit and scope of the invention.

400: motion estimation unit
410: motion estimation unit
420: search area adjustment unit

Claims (16)

In the encoding device,
A motion estimation that performs motion estimation on a default search area for a block size of one mxm size preset for any macroblock included in the current frame by referring to each of n reference frames. government; And
A search region adjusting unit configured to set an adjusted search region corresponding to each reference frame, the adjusted search region being equal to or smaller than the size of the default search region using the motion vector value calculated by the motion estimation,
The motion estimator may refer to the n reference frames, respectively, to the one or more block sizes corresponding to the sub-partition for the mxm-sized block size for the adjusted search area corresponding to each reference frame. Perform motion estimation for
And n and m are arbitrary natural numbers, respectively. The method of claim 1,
The search area adjusting unit is

And a size of the adjusted search region is set using a region boundary value (RB) determined by.
delete The method of claim 1,
If the current frame is the t (random natural number) frame,
For the t-1 th frame, which is a reference frame that is the temporally closest reference frame to the current frame among the n reference frames, the motion estimation unit performs motion estimation on a block size of pxp size for a default region,
The search region adjusting unit resets the adjusted search region equal to or smaller than the size of the default search region by using the motion vector value calculated by the motion estimation performed on the t-1th frame.
The motion estimator performs motion estimation on at least one block size corresponding to a sub-partition with respect to the block size of the pxp size with respect to the reset adjusted search area with reference to the t-1 th frame. Do it,
And p is any natural number smaller than m.
The method of claim 1,
And m is 16.
5. The method of claim 4,
P is 8, characterized in that the encoding device.
delete delete The method of claim 1,
And the encoding device is an H.264 / AVC encoder.
In the search region adjustment method performed in the encoding apparatus,
Performing a motion estimation on the default search region for a block size of one mxm size preset for any macroblock included in the current frame with reference to any reference frame;
Setting an adjusted search region corresponding to the reference frame that is equal to or smaller than the size of the default search region by using the motion vector value calculated by the motion estimation; And
Performing motion estimation on at least one block size corresponding to a sub-partition with respect to the block size of mxm in the adjusted search area corresponding to the reference frame with reference to the reference frame Including but not limited to:
M is an arbitrary natural number.
The method of claim 10,
The size of the adjusted search region is

The search area adjusting method, characterized in that the set using the area boundary value (RB) determined by.
delete The method of claim 10,
If the current frame is the t (random natural number) frame,
Performing a motion estimation on a default region for a block size of pxp size with respect to a t−1 th frame, which is a reference frame closest in time to the current frame among the arbitrary reference frames;
Resetting an adjusted search region equal to or smaller than the size of the default search region by using the motion vector value calculated by the motion estimation performed on the t-1th frame; And
Performing motion estimation on at least one block size corresponding to a sub-partition with respect to the block size of the pxp size with respect to the reset adjusted search area with reference to the t-1th frame; Include more,
And p is any natural number smaller than m.
The method of claim 10,
M is 16;
The method of claim 13,
And p is 8.
delete

Images