JP2004349756A - Moving picture coding apparatus, method and moving picture coding program - Google Patents

Abstract

Kind Code: A1 Abstract: A moving image coding for performing a search with a high motion search accuracy and a small motion compensation unit on an input video is realized at low cost.
An input image is enlarged by M times in a horizontal direction and by N times in a vertical direction. The resulting enlarged image is encoded by the pre-processing encoding unit 11. The main encoder 13 receives the motion vector information obtained in the pre-processing encoding unit 11, reduces the size of the motion vector to 1 / M in the horizontal direction and 1 / N in the vertical direction, and The input video is encoded while performing motion compensation using the obtained vector.
[Selection diagram] Fig. 1

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
According to the present invention, an input image is enlarged at a stage prior to an encoder, a motion vector search or encoding with a motion vector search is performed on the enlarged image, and a motion vector obtained from the result is used to convert the input image into an input image. On the other hand, the present invention relates to a moving picture coding apparatus for performing motion compensation with higher pixel accuracy in units of finer blocks.
[0002]
[Prior art]
In many moving image codings, redundancy between frames is eliminated by motion compensation in block units, and information is compressed. For example, in MPEG-1 and MPEG-2, which are international standards for video coding, motion compensation is performed for each macroblock which is a rectangular area of 16 × 16 pixels, and the encoder performs motion compensation for each macroblock. A search is performed to find a block having a high degree of coincidence with the target macroblock from within the search range of the reference image, and a vector indicating the position is encoded as a motion vector.
[0003]
A 16 × 16 pixel rectangular area image indicated by the motion vector is used as a prediction image, DCT is applied to a prediction error image which is a difference image between the prediction image and the original image, and its coefficients are quantized and encoded. . According to these international standards, the accuracy of this motion vector is 0.5 pixel accuracy. A reference image at a position of one pixel or less is generated by interpolating data at an integer pixel position, and a pixel difference between the generated reference image and the target macroblock is obtained.
[0004]
On the other hand, in a newer coding method, the coding efficiency is improved by performing motion compensation in units of finer blocks according to the content of the video. The newer international standard, H.264. In H.264, etc., motion compensation can be performed in units of rectangular regions of 16 × 8 pixels, 8 × 8 pixels, 8 × 4 pixels, and 4 × 4 pixels in addition to 16 × 16 pixels. If it is more efficient to perform motion compensation in smaller blocks for each block, the coding efficiency can be improved by performing coding while performing motion compensation in smaller block units. In addition, the accuracy of the motion vector is becoming finer than before, such as 0.25 pixel accuracy.
[0005]
On the other hand, if attention is paid to the mounting technology of the encoders of these moving picture coding methods, encoder LSIs corresponding to the coding methods such as MPEG-1 and MPEG-2 are already widely used, and are very small and real-time processing is not possible. Possible hardware modules can be used relatively inexpensively. In addition, there is a hardware module capable of encoding an HDTV size larger than a normal TV size also for an image size.
[0006]
On the other hand, the newer H. Encoders compatible with encoding systems such as H.264 have not yet been developed into LSIs. In general, the amount of computation required for motion vector search and motion compensation is very large in a video coding method using motion compensation, but these coding methods require more motion vector search and motion compensation than conventional methods. Due to further advancement, the required amount of calculation processing is dramatically increased, and it is difficult to implement an encoder capable of real-time encoding.
[0007]
Note that there is the following non-patent document 1 as a reference document regarding MPEG-1 encoding. Further, there is Non-Patent Document 2 below as a reference regarding MPEG-2 encoding. H. References relating to H.264 encoding include Non-Patent Document 3 below.
[0008]
[Non-patent document 1]
ISO / IEC 11172, "Information Technology-Coding of Moving Picture and Associated Associated Audio for Digital Storage Media at Up to about 19 Mbit.
[Non-patent document 2]
ITU-T Rec. H. H.262 / ISO / IEC 13818-2 MPEG-2 International Standard: Video part
[Non-Patent Document 3]
ITU-T Rec. H. H.264 / ISO / IEC 11496-10, "Advanced Video Coding", Final Committee Draft, Document JVT-E022, September 2002
[0009]
[Problems to be solved by the invention]
As described above, motion compensation is performed in units of blocks smaller in size than the blocks used in the conventional encoding system, and pixel accuracy is smaller than the accuracy used in the conventional encoding system. To realize an encoder that performs real-time encoding in a coding scheme that performs motion compensation, the amount of computation required for motion search is enormous, and developing hardware that performs such processing is costly. There is a problem.
[0010]
SUMMARY OF THE INVENTION It is an object of the present invention to solve the above-mentioned problems and to realize, at a low cost, moving picture coding using a search having a high motion search accuracy and a small motion compensation unit.
[0011]
[Means for Solving the Problems]
The present invention has been made to solve the above problems, and is a motion search with a low pixel accuracy and a large motion compensation unit, which is created for MPEG-1, MPEG-2, etc., which are conventional international standards for video coding. Using the module or the encoding module, the main encoder performs encoding with motion search or motion search on the enlarged video of the input video, and based on the resulting reduced motion vector, the main encoder On the other hand, the most main feature is to perform coding with motion compensation.
[0012]
By the above means, the motion search accuracy is high and the motion vector information with a small motion compensation unit can be obtained by the enlargement rate of the input image. Therefore, motion compensation using this vector information can be performed or the vector information can be obtained. It can be used to reduce the amount of calculation for motion search. As a result, it is possible to use a conventional module that performs a search with a lower motion search accuracy and a larger motion compensation unit, and to perform video coding using a search with a higher motion search accuracy and a smaller motion compensation unit. It can be performed at lower cost.
[0013]
BEST MODE FOR CARRYING OUT THE INVENTION
[Embodiment 1]
FIG. 1 shows an apparatus configuration diagram according to the first embodiment of the present invention. FIG. 2 is an explanatory diagram of an input image enlarging method according to the first embodiment. FIG. 3 is a vertical double of an input image (progressive image) according to the first embodiment (the same applies to later-described fourth and fifth embodiments). It is explanatory drawing of the enlargement method. FIG. 4 is a processing flowchart of the first embodiment.
[0014]
It is assumed that the input video in the first embodiment is a progressive video. In the image enlargement unit 10, the input image is enlarged twice in the horizontal direction and twice in the vertical direction in frame units as shown in FIG. 2 (S1 in FIG. 4). 2A shows a pixel group of one frame of the input video, and FIG. 2B shows a pixel group of an enlarged image obtained by enlarging the input video. Here, in the vertical direction, as shown in FIG. 3, a line is generated by interpolation using a linear sum of pixel values of several upper and lower pixels (lines) in the frame at an intermediate position of each line, and in the horizontal direction, each line is generated. A pixel is generated at an intermediate position between pixels by interpolation of a linear sum of pixel values of several pixels on the left and right.
[0015]
The pre-processing encoding unit 11 encodes the enlarged image generated by the image enlarging unit 10 using MPEG-2 (S2 in FIG. 4). The pre-coding unit 11 obtains a frame prediction motion vector with a precision of 0.5 pixel for each macroblock of 16 × 16 pixels by motion search. In encoding, frame prediction using this vector is performed. The quantization parameter is fixed to a small value of 5 or less so that a vector shift due to coding distortion of the reference image is unlikely to occur. The obtained motion vector for each macroblock is transmitted to the main encoder 13.
[0016]
The delay buffer 12 is a memory that temporarily holds input video data to be supplied to the main encoder 13 during processing of the input video by the image enlarging unit 10 and the pre-processing encoding unit 11.
[0017]
In the main encoder 13, the motion compensator 14 reduces the horizontal and vertical components of the motion vector obtained from the pre-processing encoder 11 to （(S3 in FIG. 4), and obtains a vector with 0.25 pixel accuracy. Frame prediction is performed for each 8 × 8 pixel block using the vector. The sum of absolute differences between the prediction error image generated using this vector and the original image of the same block is compared. If the former is smaller, the vector and the DCT coefficients of the prediction error component are encoded as a non-intra coded block. . If the latter is small, the DCT coefficient of the original image is encoded as an intra-encoded block. In this way, by performing the encoding, it is possible to realize the motion compensation by the frame prediction of the 0.25 pixel accuracy and the 8 × 8 pixel block (S4 in FIG. 4).
[0018]
[Embodiment 2]
The device configuration diagram of the second embodiment of the present invention is the same as that of the first embodiment shown in FIG. FIG. 5 is an explanatory diagram of a method of vertically expanding an input video (interlaced video) by a factor of 2 in the second embodiment. FIG. 6 is a diagram illustrating a change in a shift of a vector vertical component between fields of different parities in the second embodiment. FIG. FIG. 7 is a processing flowchart of the second embodiment.
[0019]
The input video in the second embodiment is an interlaced video. In the image enlargement unit 10, the input image is enlarged twice in the horizontal direction and twice in the vertical direction in field units (S10 in FIG. 7). Here, in the enlargement in the vertical direction, as shown in FIG. 5, for the first field, a line is placed at an intermediate position of each line by interpolation using a linear sum of pixel values of several upper and lower pixels (lines) in the same field. For the second field, the line is interpolated by the linear sum of the pixel values of several upper and lower pixels (lines) in the same field at the positions of quarters and three-quarters of each line. Generate Further, in the horizontal direction, pixels are generated at intermediate positions of the respective pixels by interpolation of linear sums of pixel values of several right and left pixels.
[0020]
The pre-processing encoding unit 11 encodes the enlarged image generated by the image enlarging unit 10 using MPEG-2. The pre-processing encoding unit 11 obtains a frame prediction motion vector and a field prediction vector with a 0.5-pixel accuracy for each 16 × 16 pixel macroblock by motion search (S11 in FIG. 7). The encoding performs frame field adaptive prediction using these vectors. In addition, the quantization parameter is fixed to a small value of 5 or less so that the displacement of the vector due to the coding distortion of the reference image hardly occurs. The obtained motion vector and prediction mode for each macroblock are transmitted to the main encoder 13.
[0021]
In the main encoder 13, the horizontal and vertical components of the obtained motion vector are reduced to １ ／ to obtain a vector with a precision of 0.25 pixels. However, when the prediction mode is field prediction and prediction is performed between fields having different parities, 0.25 is added to the vertical component of the vector (S12 in FIG. 7).
[0022]
The interlaced frame image generated by alternately combining the first field enlarged image and the second field enlarged image generated by the image enlarging unit 10 for each line has a correct relative position of the image content, and is preprocessed. In the encoding unit 11, a normal frame prediction vector can be obtained. Also, in the prediction between fields of the same parity, the relative position of the image content is accurate between the fields of the same parity, and a normal field prediction vector can be obtained.
[0023]
However, in the prediction between fields of different parities, when field prediction is selected, correction is necessary because the prediction vector between fields of different parities is shifted by 0.5 pixel as shown in FIG. For example, when the first field is the top field and the second field refers to the first field, the vertical component of the motion vector indicating a position on one line in the frame image is zero. Therefore, a vector having a vertical component of 0 in a vertically doubled image corresponds to a motion vector indicating a position on a 1/2 line in a frame image in an input image before enlargement. In the field prediction, the vertical component of the motion vector pointing to the position on the 1/2 line in the frame image is 0.25. Therefore, the prediction vector between the fields of different parities obtained from the pre-coding unit 11 is different. Needs to be corrected by reducing the horizontal and vertical components by half and then adding 0.25 to the vertical components.
[0024]
The main encoder 13 performs frame prediction for each 8 × 8 pixel block according to the vector and the prediction mode thus obtained. The prediction error image generated using this vector is compared with the sum of absolute differences of the original image of the same block. If the former is small, the vector and the DCT coefficient of the prediction error component are encoded as a non-intra coded block. Become If the latter is small, the DCT coefficients of the original image are encoded as an intra-coded block.
[0025]
In this way, by performing the encoding, it is possible to realize the motion compensation by the frame prediction of the 0.25 pixel accuracy and the 8 × 8 pixel block (S13 in FIG. 7).
[0026]
[Embodiment 3]
FIG. 8 shows an apparatus configuration diagram of the third embodiment of the present invention. FIG. 9 is a diagram showing the image content position in the vertical double enlargement method in units of frame images in the third embodiment, and FIG. 10 is an explanatory diagram of double vertical enlargement in units of field images in the third embodiment. FIG. 11 is a diagram illustrating a change in a shift of a vector vertical component between fields having different parities in the third embodiment. FIG. 12 is a processing flowchart of the third embodiment.
[0027]
The input video in the third embodiment is an interlaced video. In the frame image enlargement unit 20, the input image is enlarged twice in the horizontal direction and twice in the vertical direction in frame units (S20 in FIG. 12). That is, in the vertical direction, a line is generated at the intermediate position of each line of the frame by interpolation using a linear sum of the pixel values of several upper and lower pixels (lines) in the same frame, and in the horizontal direction, Pixels are generated at intermediate positions of each pixel by interpolation of linear sums of pixel values of several pixels on the left and right. Here, in the image enlarged in this way, the image content is correctly enlarged in the still region, but in the moving region, the image contents of two temporally different fields are averaged to correctly generate an enlarged image. Note that this is not done.
[0028]
The frame preprocessing encoding unit 21 encodes the enlarged image generated by the frame image enlarging unit 20 by MPEG-2 (S21 in FIG. 12). The frame pre-coding unit 21 obtains a frame prediction motion vector for each 16 × 16 pixel macroblock with 0.5 pixel accuracy by motion search. Encoding is performed by frame prediction using this vector, and the quantization parameter is fixed to a small value of 5 or less, so that the displacement of the vector due to the encoding distortion of the reference image hardly occurs. The obtained motion vector for each macroblock is transmitted to the main encoder 25.
[0029]
At the same time, the field image enlargement unit 22 enlarges the input video twice in the horizontal direction and twice in the vertical direction in units of fields (S22 in FIG. 12). That is, as shown in FIG. 5, the linear sum of the pixel values of several upper and lower pixels (lines) in the same field is independently provided in the first field and the second field at an intermediate position of each line of the field in the vertical direction. , A line is generated by horizontal interpolation, and a pixel is generated by interpolating a linear sum of pixel values of several left and right pixels at an intermediate position of each pixel in the horizontal direction.
[0030]
The field preprocessing encoding unit 23 encodes the enlarged image generated by the field image enlarging unit 22 using MPEG-2 (S23 in FIG. 12). The field pre-coding unit 23 obtains a field prediction motion vector for each macroblock of 16 × 16 pixels with a 0.5 pixel precision by motion search. The encoding uses field prediction using this vector, and the quantization parameter is fixed to a small value of 5 or less, so that the displacement of the vector due to the encoding distortion of the reference image hardly occurs. The obtained motion vector for each macroblock is transmitted to the main encoder 25.
[0031]
The delay buffer 24 supplies the input video to be supplied to the main encoder 25 during the processing of the input video by the frame image enlarging unit 20 and the frame pre-processing encoding unit 21 and the processing by the field image enlarging unit 22 and the field pre-processing encoding unit 23. To temporarily hold the data.
[0032]
As described above, the frame image of the interlaced video generated by alternately combining the first field enlarged image and the second field enlarged image generated by the field image enlarging unit 22 for each line has the original image position. Note the misalignment. However, since the relative positions of the pixel contents between the fields of the same parity do not shift, the components of the field motion vector between the fields of the same parity do not shift. In addition, although there exists a shift in the motion vector between different parities, as shown in FIG. 11, since the position of the field motion vector between the different parities in the main encoder 25 matches the positional shift, the shift of the motion vector is a problem. It does not become. That is, in the main encoder 25, when the second field refers to the first field, when the vertical component of the motion vector is 0, and when the image is viewed as a frame, it indicates a position shifted by one line. This is because a similar relationship holds for the two enlarged fields.
[0033]
In the main encoder 25, the horizontal and vertical components of the frame motion vector obtained from the frame pre-coding unit 21 are reduced to １ ／ to obtain a vector with 0.25 pixel accuracy (S24 in FIG. 12). At the same time, the horizontal and vertical components of the field motion vector obtained from the field pre-coding unit 23 are reduced to １ ／ to obtain a vector with 0.25 pixel accuracy. The motion compensator 26 performs motion compensation for each 8 × 8 pixel block using one of these vectors. The prediction error image generated using the field vector and the prediction error image generated using the frame vector are compared with the sum of absolute differences between the original images of the same block, and encoding is performed in a mode having the smallest value.
[0034]
In this way, by performing the encoding, it is possible to realize the motion compensation by the frame prediction of the 0.25 pixel accuracy and the 8 × 8 pixel block (S25 in FIG. 12).
[0035]
[Embodiment 4]
FIG. 13 shows an apparatus configuration diagram of the fourth embodiment of the present invention. FIG. 14 is a processing flowchart of the fourth embodiment.
[0036]
It is assumed that the input video in the fourth embodiment is a progressive video. In the image enlargement unit A30, the input image is enlarged twice in the horizontal direction and twice in the vertical direction in frame units (S30 in FIG. 14). That is, in the vertical direction, a line is generated at the intermediate position of each line of the frame by interpolation using a linear sum of pixel values of several upper and lower pixels (lines), and in the horizontal direction, several left and right pixels are located at the intermediate position of each pixel. Pixels are generated by interpolating the linear sum of the pixel values of minutes.
[0037]
The pre-processing encoding unit A31 encodes the enlarged image generated by the image enlarging unit A30 using MPEG-2 (S31 in FIG. 14). The pre-processing encoding unit A31 obtains a frame prediction motion vector with a 0.5-pixel accuracy for each macroblock of 16 × 16 pixels by motion search. In encoding, frame prediction using this vector is performed. In addition, the quantization parameter is fixed to a small value of 5 or less so that the displacement of the vector due to the coding distortion of the reference image hardly occurs. The obtained motion vector for each macroblock is transmitted to the main encoder 37. In the main encoder 37, the horizontal and vertical components of the obtained motion vector are reduced to 1/2 (S36 in FIG. 14), and a vector with horizontal and vertical 0.25 pixel accuracy in units of 8 × 8 pixels is obtained.
[0038]
In the image enlargement unit B32, the input image is enlarged twice in the horizontal direction and one time in the vertical direction in frame units (S32 in FIG. 14). That is, a pixel is generated by interpolating a linear sum of pixel values of several pixels on the left and right at an intermediate position of each pixel in the horizontal direction.
[0039]
The pre-processing encoding unit B33 encodes the enlarged image generated by the image enlarging unit B32 using MPEG-2 (S33 in FIG. 14). The preprocessing encoding unit B33 obtains a frame prediction motion vector for each macroblock of 16 × 16 pixels with 0.5 pixel accuracy by motion search. In encoding, frame prediction using this vector is performed. In addition, the quantization parameter is fixed to a small value of 5 or less so that the displacement of the vector due to the coding distortion of the reference image hardly occurs. The obtained motion vector for each macroblock is transmitted to the main encoder 37. In the main encoder 37, the horizontal component of the obtained motion vector is reduced by half (S36 in FIG. 14), and a vector having a horizontal 0.25 pixel precision and a vertical 0.5 pixel precision of 8 × 16 pixel units is obtained. obtain.
[0040]
In the image enlargement unit C34, the input image is enlarged by 1 in the horizontal direction and twice in the vertical direction in frame units (S34 in FIG. 14). That is, in the vertical direction, a line is generated at an intermediate position of each line of the frame by interpolation using a linear sum of pixel values of several upper and lower pixels (lines).
[0041]
The preprocessing encoding unit C35 encodes the enlarged image generated by the image enlarging unit C34 by MPEG-2 (S35 in FIG. 14). The preprocessing coding unit C35 obtains a frame prediction motion vector with a 0.5-pixel accuracy for each macroblock of 16 × 16 pixels by motion search. In encoding, frame prediction using this vector is performed. In addition, the quantization parameter is fixed to a small value of 5 or less so that the displacement of the vector due to the coding distortion of the reference image hardly occurs. The obtained motion vector for each macroblock is transmitted to the main encoder 37. In the main encoder 37, the vertical component of the obtained motion vector is reduced by half (S36 in FIG. 14), and a vector of horizontal 0.5 pixel precision and vertical 0.25 pixel precision of 16 × 8 pixel unit is converted. obtain.
[0042]
The delay buffer 36 temporarily holds input video data to be supplied to the main encoder 37 during the above-described process of image enlargement and pre-processing encoding of the input video.
[0043]
The main encoder 37 performs a motion search in the vicinity of these vectors using the locally decoded image in the main encoder 37 as a reference image with a block size corresponding to each (S37 in FIG. 14). That is, the neighborhood of the vector obtained from the pre-coding unit A31 has the size of 8 × 8 pixel blocks, the neighborhood of the vector obtained from the pre-coding unit B33 has the size of 8 × 16 pixel blocks, and the neighborhood of the vector obtained from the pre-coding unit C35. A motion search is performed with a 16 × 8 pixel block size and a horizontal and vertical 0.25 pixel precision for the neighborhood of the vector. Apart from this, in the main encoder 37, a frame motion vector search of 16 × 16 pixels is performed in the internal motion compensation unit 38 using the locally decoded image in the main encoder 37 as a reference image.
[0044]
The main encoder 37 compares, for each 16 × 16 pixel block, the sum of absolute differences of the original image of the same 16 × 16 pixel block and the prediction error image generated using the above-described vector having the same block size but a different block size. At this time, the sum of the absolute values of the prediction error images for 4 for the 8 × 8 pixel block size, 2 for the 16 × 8 pixel block, 8 for the 8 × 16 pixel block, and 1 for the 16 × 16 pixel block, respectively Compare.
[0045]
Among these, the mode with the smallest sum of absolute differences is selected. That is, if the sum of the absolute values of the prediction error images for four 8 × 8 pixel blocks is small, the non-intra encoding of the 16 × 16 pixel block is performed using four 8 × 8 pixel block size motion vectors. If the sum of absolute values of the prediction error images for two 16 × 8 pixel blocks is small, non-intra encoding of the 16 × 16 pixel block is performed using two 16 × 8 pixel block size motion vectors. If the sum of the absolute values of the prediction error images for two 8 × 16 pixel blocks is small, the non-intra encoding of the 16 × 16 pixel block is performed using two 8 × 16 pixel block size motion vectors. If the sum of absolute values of the prediction error image of the 16 × 16 pixel block is small, non-intra coding of the 16 × 16 pixel block is performed using a motion vector of a 16 × 16 pixel block size. If the sum of the absolute values of the original images of the 16 × 16 pixel block is small, the 16 × 16 pixel block is intra-coded.
[0046]
In this way, by performing encoding, motion compensation by adaptive frame prediction of 0.25 pixel accuracy, 8 × 8 pixel block, 8 × 16 pixel block, 16 × 8 pixel block, and 16 × 16 pixel block is realized. (S38 in FIG. 14).
[0047]
[Embodiment 5]
FIG. 15 shows an apparatus configuration diagram according to the fifth embodiment of the present invention. FIG. 16 is a processing flowchart of the fifth embodiment. The input video of the fifth embodiment is assumed to be a progressive video.
[0048]
In the image enlargement unit 40, the input video is enlarged twice in the horizontal direction and twice in the vertical direction in frame units (S40 in FIG. 16). That is, in the vertical direction, a line is generated at the intermediate position of each line of the frame by interpolation using a linear sum of pixel values of several upper and lower pixels (lines), and in the horizontal direction, several left and right pixels are located at the intermediate position of each pixel. Pixels are generated by interpolating the linear sum of the pixel values of minutes.
[0049]
The preprocessing encoding unit 41 encodes the enlarged image generated by the image enlarging unit 40 using MPEG-2 (S41 in FIG. 16). The pre-coding unit 41 obtains a frame prediction motion vector with a precision of 0.5 pixel for each macroblock of 16 × 16 pixels by motion search. In encoding, frame prediction using this vector is performed. In addition, the quantization parameter is fixed to a small value of 5 or less so that the displacement of the vector due to the coding distortion of the reference image hardly occurs. The coded stream coded in this way is output from the pre-processing coding unit 41 (S42 in FIG. 16) and input to the motion vector information extraction unit 42.
[0050]
In the motion vector information extraction unit 42, motion vector information for each macroblock is extracted (S44 in FIG. 16) and transmitted to the main encoder 44.
[0051]
The delay buffer 43 temporarily holds input video data to be supplied to the main encoder 44 during processing of the input video by the image enlargement unit 40, the pre-processing encoding unit 41, and the motion vector information extraction unit 42.
[0052]
In the main encoder 44, the horizontal and vertical components of the obtained motion vector are reduced to ４５ (S45 in FIG. 16). A motion search of an 8 × 8 pixel block near this vector is performed using the locally decoded image of the main encoder 44 as a reference image to obtain a frame vector (S46 in FIG. 16). The sum of absolute differences between the prediction error image generated using this vector and the original image of the same block is compared. If the former is smaller, the vector and the DCT coefficients of the prediction error component are encoded as a non-intra coded block. . If the latter is small, the DCT coefficient of the original image is encoded as an intra-encoded block. In this way, by performing the encoding, it is possible to realize the motion compensation by the frame prediction of the 0.25 pixel accuracy and the 8 × 8 pixel block (S47 in FIG. 16).
[0053]
In the fifth embodiment, the pre-processing encoder 41 does not specifically output the motion vector information, but simply outputs a stream as a normal operation as an encoder. There is an advantage that it is easy.
[0054]
The present invention can be implemented by combining the above embodiments. Further, the processing related to the overall control in each of the above embodiments can also be realized by a computer and a software program, and the program can be provided by being recorded on a computer-readable recording medium or provided through a network. It is also possible.
[0055]
【The invention's effect】
As described above, according to the present invention, based on a vector obtained by enlarging an input video, performing motion search or encoding with motion search using an inexpensive encoder, and reducing a resultant motion vector. , The main encoder performs coding with motion compensation on the input image, so that it is possible to realize video coding using search with higher motion search accuracy and smaller motion compensation unit at lower cost than before. it can.
[Brief description of the drawings]
FIG. 1 is an apparatus configuration diagram of Embodiments 1 and 2 of the present invention.
FIG. 2 is an explanatory diagram of a method of enlarging an input image according to the first embodiment.
FIG. 3 is an explanatory diagram of a method of vertically enlarging an input video (progressive video) twice in Embodiments 1, 4, and 5;
FIG. 4 is a processing flowchart according to the first embodiment.
FIG. 5 is an explanatory diagram of a method of vertically enlarging an input video (interlaced video) by a factor of 2 in the second embodiment.
FIG. 6 is a diagram showing a change in a shift of a vector vertical component between fields of different parities in the second embodiment.
FIG. 7 is a processing flowchart according to the second embodiment.
FIG. 8 is an apparatus configuration diagram according to a third embodiment of the present invention.
FIG. 9 is a diagram showing an image content position in a method of enlarging a frame image unit twice in the vertical direction in the third embodiment.
FIG. 10 is an explanatory diagram of a field image unit vertically enlarged twice in Embodiment 3;
FIG. 11 is a diagram illustrating a change in a shift of a vector vertical component between fields of different parities in the third embodiment.
FIG. 12 is a processing flowchart of the third embodiment.
FIG. 13 is an apparatus configuration diagram according to a fourth embodiment of the present invention.
FIG. 14 is a processing flowchart according to the fourth embodiment.
FIG. 15 is an apparatus configuration diagram according to a fifth embodiment of the present invention.
FIG. 16 is a processing flowchart according to the fifth embodiment.
[Explanation of symbols]
10,30,32,34,40 Image enlargement unit
11, 31, 33, 35, 41 pre-processing encoding unit
12, 24, 36, 43 delay buffer
13,25,37,44 Main encoder
14,26,38,45 Motion compensation unit
20 Frame image enlargement unit
21 Frame Pre-Processing Encoding Unit
22 Field image magnifier
23 Field precoding unit

Claims (13)

In a video encoding device using motion compensation,
Image enlarging means for enlarging the input image by a factor of M (M ≧ 1) in the horizontal direction and N times (N ≧ 1, M × N> 1) in the vertical direction;
Preprocessing encoding means for encoding the resulting image;
Receiving the motion vector information obtained by the pre-processing encoding means, the size of the motion vector is reduced to 1 / M in the horizontal direction and 1 / N in the vertical direction, and the vector obtained for the input image is used. A moving image coding apparatus comprising: a coding unit having a motion compensating unit for performing motion compensation by using a motion compensation unit. The moving picture coding apparatus according to claim 1,
The encoding means receives the motion vector information obtained by the preprocessing encoding means, reduces the size of the motion vector to 1 / M in the horizontal direction and 1 / N in the vertical direction, A moving image encoding apparatus characterized by adding a component for correcting a motion vector to a motion vector. In a video encoding device using motion compensation,
Frame image enlarging means for enlarging the input image by a factor of M (M ≧ 1) in the horizontal direction and N times (N ≧ 1, M × N> 1) in the vertical direction;
Field image enlarging means for enlarging an input image by a factor of M in the horizontal direction and N times in the vertical direction,
Frame preprocessing encoding means for encoding an enlarged image obtained from the frame image enlarging means using frame prediction;
A field pre-processing encoding unit for encoding the enlarged image obtained from the field image enlarging unit by using field prediction;
A motion vector obtained from the frame pre-coding unit and the field pre-coding unit is reduced to 1 / M in the horizontal direction and 1 / N in the vertical direction, and one of the obtained vectors is used. A moving image encoding apparatus comprising: an encoding unit that encodes an input image while performing motion compensation. The moving picture coding apparatus according to claim 1,
The image enlarging means includes an image enlarging unit having a plurality of different enlarging ratios,
The pre-processing encoding unit includes a plurality of pre-processing encoding units that respectively encode the enlarged images obtained from the image enlarging units,
The encoding means reduces a motion vector obtained from each of the pre-processing encoding units according to a respective enlargement ratio, and encodes an input image while performing motion compensation using any of the obtained vectors. A video encoding device characterized by the above-mentioned. The moving picture coding apparatus according to any one of claims 1 to 4,
The encoding means receives the motion vector information obtained in the preprocessing encoding means, reduces the magnitude of the motion vector to 1 / M in the horizontal direction and 1 / N in the vertical direction, A moving picture coding apparatus characterized by performing a motion search around an obtained vector, determining a motion vector, and coding an input picture while performing motion compensation. The moving picture coding apparatus according to any one of claims 1 to 4,
Motion vector information extracting means for extracting motion vector information from the stream output from the preprocessing encoding means,
The encoding means receives the vector information obtained from the motion vector information extraction means, reduces the size of the motion vector to 1 / M in the horizontal direction and 1 / N in the vertical direction, and uses the obtained vector. A moving image encoding apparatus that encodes an input image while performing motion compensation by performing the motion compensation. In a video coding method using motion compensation,
An image enlarging step of enlarging the input image by a factor of M (M ≧ 1) in the horizontal direction and N times (N ≧ 1, M × N> 1) in the vertical direction;
A pre-processing encoding step for encoding the resulting enlarged image;
In response to the motion vector information obtained in the above preprocessing encoding step, the size of the motion vector is reduced to 1 / M in the horizontal direction and 1 / N in the vertical direction, and the vector obtained for the input image is used. A coding step of coding an input image while performing motion compensation by using the moving image coding method. The moving picture encoding method according to claim 7,
In the encoding step, upon receiving the motion vector information obtained by encoding the enlarged image, the magnitude of the motion vector is reduced to 1 / M in the horizontal direction and 1 / N in the vertical direction. A moving picture coding method characterized by adding a component for correcting displacement to a motion vector. In a video coding method using motion compensation,
A frame image enlarging step of enlarging the input image by a factor of M (M ≧ 1) in the horizontal direction and N times (N ≧ 1, M × N> 1) in the vertical direction;
A field image enlargement step of enlarging an input image by a factor of M in the horizontal direction and N times in the vertical direction;
A frame pre-processing encoding step of encoding the enlarged image obtained by the frame image enlarging step by using frame prediction;
A field pre-processing encoding step of encoding the enlarged image obtained by the field image enlarging step by using field prediction;
A motion vector obtained by the pre-processing coding step of the frame and the pre-processing coding step of the field is reduced to 1 / M in the horizontal direction and 1 / N in the vertical direction, and one of the obtained vectors is used. A coding step of coding an input image while performing motion compensation by using the moving image coding method. The moving picture encoding method according to claim 7,
In the image enlargement step, the input image is enlarged to a plurality of different enlargement ratios.
In the above-mentioned encoding step, each motion vector obtained as a result of encoding each enlarged image is reduced in accordance with each enlargement ratio, and encoding of the input image is performed while performing motion compensation using any of the obtained vectors. A moving picture coding method characterized by performing the following. In the moving picture coding method according to any one of claims 7 to 10,
In the above-described encoding step, the magnitude of the motion vector obtained by encoding the enlarged image is reduced to 1 / M in the horizontal direction and 1 / N in the vertical direction, and the magnitude of the motion vector around the vector obtained for the input image is reduced. A moving picture coding method characterized by performing a motion search, determining a motion vector, and coding an input picture while performing motion compensation. In the moving picture coding method according to any one of claims 7 to 10,
Extracting motion vector information from a stream output by encoding the enlarged image,
In the encoding step, the size of the motion vector is reduced to 1 / M in the horizontal direction and 1 / N in the vertical direction, and the input image is encoded while performing motion compensation using the obtained vector. A moving picture coding method characterized by the following. A moving image encoding program for causing a computer to execute the moving image encoding method according to any one of claims 7 to 12.

Images