JP2005236584A - Moving picture information conversion encoding apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a moving picture information conversion encoding apparatus allowing re-estimation, from a motion vector calculated by variably split blocks in a first moving picture encoding system employing the variably split blocks to the motion vector of a macro-block in a second moving picture encoding system. <P>SOLUTION: The moving picture information conversion encoding apparatus reuses the average Vm of two motion vectors calculated before conversion and having high correlation when the correlation value ρ of the motion vector before the conversion is greater than or equal to a threshold ρ1 as the motion vector after the conversion (Steps S4-S7); when the correlation value ρ is of a value between the thresholds ρ2 and ρ1, determines that an error is increased when the average Vm is used as the motion vector, and re-calculates the motion vector after the conversion with the point of the average Vm of the two motion vectors before the conversion as a center (Steps S5, S6, S8); when the correlation value ρ is lower than the threshold ρ2, the correlation is very low, thus, re-calculates the motion vector after the conversion in an ordinary search range (Steps S4, S9). <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a moving image information conversion encoding apparatus, and particularly when decoding encoded moving image information, re-encodes a moving image code string according to the intended use, and enables conversion of an encoding method. The present invention relates to a moving picture information conversion coding apparatus that performs motion vector prediction for converting a moving picture coding method with a small amount of information processing.

  MPEG-2 (Moving Picture Experts Group Phase 2) is used as a moving picture coding standard system in digital broadcasting and DVD (Digital Versatile Disk), and MPEG-4 is used in fields such as Internet streaming and mobile communication. Video coding standards are also used. In addition, a moving image coding standard such as H.263 intended for transmission at a low bit rate such as a video conference system is also used.

  Under such circumstances, standardization of an encoding method “H.264 / MPEG-4 AVC (hereinafter referred to as H.264)” for the purpose of further improving compression efficiency is being studied. This H.264 is compatible with a wide range of applications from video conferencing at low bit rates to HDTV (High Definition Television), and is about twice as high as conventional methods (MPEG-2, MPEG-4, etc.). This is a moving picture encoding method that can realize the above. Therefore, in order to use encoded moving image information according to various usages and encoding methods, compatibility of moving image information between the encoding methods described above is important.

  Here, a variable divided block system that does not exist in MPEG-2 or MPEG-4, which is a feature of the H.264 encoding system, will be described with reference to FIG. However, here, as an example, the description will be made assuming that the size of the macroblock is 16 × 16 pixels. In this block method, a macro block 510 having a size of 16 × 16 pixels in FIG. 8 is converted into a block 541 having an 8 × 8 pixel size, a block 542 having an 8 × 4 pixel size, and a 4 × 8 pixel depending on the complexity of the image area. Block 543, 4 × 4 pixel size block 544, 8 × 16 pixel size block 530, 16 × 8 pixel size block 520, 16 × 16 pixel size block 510, and so on. Is adopted.

  In FIG. 8, when the block pattern of an 8 × 8 pixel size block 540 is selected, four patterns as shown in blocks 541 to 544 can be further selected. This is called a sub macroblock type, and a motion vector can be calculated for each sub macroblock. Therefore, if there is a complex part and a flat part in the macro block 510 of 16 × 16 pixel size, the block size can be changed flexibly and used for prediction, so the generated code amount after quantization is reduced. This enables calculation of a motion vector with high accuracy.

  In order to use such an encoding method in which variably divided blocks are adopted according to various uses such as the MPEG-2 method in DVD, variably divided blocks are adopted. It is necessary to convert the encoding method into another conventional encoding method such as the MPEG-2 method. That is, re-encoding according to the conventional encoding method is required.

  Considering the processing amount of the re-encoding process, there is a method called block matching particularly in the process related to motion vector calculation. In this block matching method, a corresponding point in the second frame with respect to an arbitrary point in the first frame is determined based on pattern matching between the blocks, and at this time, as an evaluation of matching, a difference in gradation value of each point is determined. Since the sum is used, it can be said that the processing amount is large depending on the search range.

  Therefore, as a motion vector re-estimation method for the purpose of reducing the amount of processing at the time of re-encoding, the motion vector statistics (average, variance, etc.) are used when reducing the image size, and the code amount is high without reducing the quality. There has been known an image encoded data size conversion apparatus that performs re-estimation of a motion vector while suppressing the above (see, for example, Patent Document 1). In addition to the integration method using motion vector dispersion at the time of image size reduction as image complexity information, a moving image code string conversion device that performs a determination for converting a motion vector to a sub-macroblock in MPEG-4 has also been conventionally used. It is known (see, for example, Patent Document 2).

  In each of the conventional devices described in Patent Documents 1 and 2 above, the variance of motion vector information calculated with the conventional macroblock size of a moving image encoded when considering the reduction of the image size is used. This corresponds to a motion vector integration method when converting from −2 to MPEG-4, or use determination of a sub-macroblock mode which is a quarter size of a macroblock having a 16 × 16 pixel size.

JP 2002-344972 A JP 2003-153271 A

  However, in the conventional apparatuses described in Patent Documents 1 and 2, there is a problem that it is impossible to perform motion vector estimation at the time of moving image re-encoding using a motion vector obtained by a variably divided sub-macroblock. is there. Therefore, when converting an encoding method adopting a variably divided block mode, it is necessary to recalculate the motion vector after decoding all the encoded information once. However, there is a problem of requiring a calculation amount.

  In addition, the motion vector calculated in the sub-macroblock mode used in H.264 is more accurate than the conventional method, but the conventional devices described in Patent Documents 1 and 2 employ variably divided blocks. However, there is a problem that the recalculation by the conventional video coding method that has not been performed reduces the accuracy of the motion vector and degrades the image quality.

  The present invention has been made in view of the above points. From the first moving picture coding method that employs a variable divided block, the second moving picture coding that does not employ a variable divided block. It is an object of the present invention to provide a moving picture information conversion encoding apparatus capable of performing re-encoding that enables high-speed control of the amount of generated codes without degrading quality.

  In addition, another object of the present invention is to provide a second moving picture coding method from a motion vector calculated in a variable divided block in the first moving picture coding method that employs a variable divided block. It is an object of the present invention to provide a moving picture information transform coding apparatus that can re-estimate the motion vector of a macroblock.

  In order to achieve the above object, the present invention decodes a first moving image encoded signal encoded by a first encoding method based on a first motion vector for each variable first encoding unit. Moving picture information that re-encodes the decoded moving picture information for each fixed second coding unit based on the second motion vector by the second coding method to generate a second moving picture coded signal A transform coding apparatus, a motion vector information interpretation means for obtaining a first motion vector from a signal obtained by decoding a first moving image coded signal, and decoding a first moving image coded signal Image resolution conversion means for performing image resolution conversion based on the image resolution conversion coefficient, and motion vector information when the first encoding unit is smaller than the second encoding unit. Of the first motion vector obtained by the interpreting means A correlation value calculating means for calculating a function value, a motion vector estimating means for comparing the calculated correlation value with a preset threshold value, and estimating a second motion vector according to the comparison result, and a motion vector estimating means Based on the result of multiplying the average of the first motion vector and the image resolution conversion coefficient, the reuse or search range by the average of the first motion vector is changed according to the estimation result of the second motion vector obtained by Motion vector re-calculation or motion vector re-calculation in the normal range based on the multiplication result of the first motion vector and the image resolution conversion coefficient is selected to generate a second motion vector Based on the search means, the moving image information whose image resolution is converted by the image resolution conversion means, and the second motion vector generated by the motion vector search means, the second code It is obtained by a structure having an encoding means for generating a second video encoding signal, and then re-coded by the scheme.

  In the present invention, the first moving image encoded signal encoded by the first encoding method based on the first motion vector for each variable first encoding unit is converted into a fixed second encoding unit. When generating a second moving image encoded signal by re-encoding by the second encoding method based on the second motion vector every time, the correlation value of the first motion vector is compared with a preset threshold value. The second motion vector is estimated according to the comparison result, and the first motion vector based on the multiplication result of the average of the first motion vector and the image resolution conversion coefficient is determined according to the estimation result. Select either re-calculation by averaging or re-calculation of motion vectors with a changed search range, or re-calculation of motion vectors in the normal range based on the multiplication result of the first motion vector and the image resolution conversion coefficient, 2 motion vectors raw As a result, the amount of calculation is not required compared to the re-search of motion vectors in normal re-encoding, and the feature that the accuracy of motion vectors, which is a feature of the first encoding method, is excellent is ensured. it can.

  In order to achieve the above object, according to the present invention, the encoding means includes the second encoding unit generated by the motion vector search means when the first encoding unit is larger than the second encoding unit. Using the first motion vector obtained by the motion vector information interpreting means and the image resolution conversion coefficient in place of the motion vector, the moving image information whose image resolution is converted by the image resolution converting means is converted into the second code. It is characterized by re-encoding by the encoding method.

  In the present invention, when the first coding unit is larger than the second coding unit, the first motion vector and the second motion vector are regarded as having the same direction, and motion vector information interpretation means Since the first motion vector obtained by the above is used as it is, recalculation for generating the second motion vector can be made unnecessary.

  According to the present invention, the first encoded moving image having the first motion vector in the variable first encoding unit is not required in comparison with the motion vector re-search in the normal re-encoding. An image can be converted to a second encoded moving image having a second motion vector in a normal second encoding unit at high speed and with high accuracy.

  Further, according to the present invention, it is possible to ensure the feature that the accuracy of the motion vector, which is the feature of the first coding method having the first motion vector for each variable first coding unit, is excellent, A highly accurate motion vector can be estimated.

  Furthermore, according to the present invention, when the first coding unit is larger than the second coding unit, the first motion vector obtained by the motion vector information interpreting means is used as it is. Since the recalculation for generating the second motion vector is not necessary, the moving image conversion can be further speeded up. As described above, according to the present invention, it is possible to reduce the amount of calculation in the re-encoding process of the encoded moving image information, perform high-speed moving image conversion, and improve the encoding efficiency.

  Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a moving picture information transform coding apparatus according to the present invention. As shown in the figure, the moving image information conversion coding apparatus includes a moving image decoding unit 100, a moving image information conversion unit 200, and a moving image coding unit 300. The moving image decoding unit 100 includes a moving image information separation unit 101, a variable length decoding unit 102, a motion vector information interpretation unit 103, a motion vector compensation unit 104, an inverse quantization unit 105, an inverse coefficient conversion unit 106, an adder 107, and an image. A memory unit 108 is included. The moving image information conversion unit 200 includes an image resolution conversion unit 201 and a motion vector estimation unit 202. The moving image encoding unit 300 includes a motion vector search unit 301, a motion compensation unit 302, an image memory unit 303, a subtractor 304, a coefficient conversion unit 305, a quantization unit 306, an inverse quantization unit 307, an inverse coefficient conversion unit 308, An adder 309, a variable length encoding unit 310, and a moving image information combining unit 311, for example, encoded moving image information 41 encoded by the H.264 encoding method is converted into MPEG-1, MPEG- 2. Converted into encoded moving picture information 66 having a motion vector in a fixed macroblock unit such as MPEG-4.

  Next, the operation of the present embodiment will be described. The encoded moving image information 41 input from the encoded information input terminal 10, for example, encoded by the H.264 encoding method, is motion vector information 42 that is variable-length encoded by the moving image information separation unit 101. And the variable length coded quantized transform coefficient information 43, and then decoded by the variable length decoder 102 to become encoded motion vector information 44 and quantized transform coefficient 45, respectively.

  The encoded motion vector information 44 is converted into motion vector information 46 for each block variably divided by the motion vector information interpretation unit 103. The quantized transform coefficient 45 is inversely quantized by the inverse quantization unit 105 and converted into a transform coefficient 47. The conversion coefficient 47 is converted into a pixel coefficient 48 by the inverse coefficient converter 106. The pixel coefficient 48 is prediction error information obtained by motion compensation processing at the time of moving image encoding.

  On the other hand, the image memory unit 108 stores a reference image 49 used for motion compensation, and the motion vector compensation unit 104 supplies the reference image 49 read from the image memory unit 108 and the motion vector information interpretation unit 103. A motion compensation image 50 is generated based on the motion vector information 46 for each block. The motion compensated image 50 and the prediction error information 48 are added by the adder 107 to generate a decoded image 51.

  The decoded image 51 is stored as a reference image 49 in the image memory unit 108, and is supplied to the image resolution conversion unit 201 where the decoded image 51 is input in accordance with a predetermined image resolution conversion coefficient 52 input from the resolution conversion coefficient input terminal 20. The image information 53 is obtained by converting the image resolution. Further, the image resolution conversion unit 201 passes the image resolution information 54 of the image information 53 to the motion vector estimation unit 202. The motion vector estimation unit 202 receives the motion vector information 46 decoded for each variably divided block and the image resolution information 54 and outputs motion vector calculation determination information 55 after moving image information conversion by a method described later. To do.

  Based on the motion vector calculation determination information 55, the motion vector search unit 301 integrates the motion vectors of the image information 53 or recalculates the motion vectors, and outputs a motion vector 56 after moving image information conversion. The motion compensation unit 302 receives the reference image 57 from the image memory unit 303 and the motion vector 56 after the motion image information conversion generated by the motion vector search unit 301 as input, and generates a motion compensation image 58. The motion compensated image 58 is supplied to the subtractor 304 together with the decoded image 53 whose resolution has been converted. Here, the difference between pixel coefficients is taken and prediction error information 59 is generated.

  The prediction error information 59 is converted into the conversion coefficient 60 by the coefficient conversion unit 305 and then supplied to the quantization unit 306 to be converted into the quantization conversion coefficient 61. The quantized transform coefficient 61 is converted into a conversion error coefficient 62 by an inverse quantization unit 307, and further supplied to an inverse coefficient conversion unit 308 to be decoded prediction error information 63, which is then supplied to an adder 309. The reference image 57 for motion compensation is added to the compensation image 58. The motion compensation reference image 57 is stored in the image memory unit 303.

  Also, the quantized transform coefficient 61 and the motion vector information 56 are respectively supplied to the variable length coding unit 310 and subjected to variable length coding, and become an encoded quantized transform coefficient 64 and encoded motion vector information 65, respectively. These are encoded moving image information 66 having a fixed macroblock unit motion vector after conversion of MPEG-1, MPEG-2, MPEG-4, and the like by the moving image information synthesis unit 311, and an encoded information output terminal 30 is output.

  Next, the resolution conversion coefficient 52 is set to “1”, that is, the image resolution is not converted, and the motion vector decoded in units of the variably divided macroblocks is fixed to the conventional macroblock. A motion vector estimation method in the case of conversion to a unit motion vector will be described with reference to the flowchart of FIG. 2, steps S1 to S6 are executed by the motion vector estimation unit 202 in FIG. 1, and steps S7 to S11 are executed by the motion vector search unit 301.

  In brief, first, motion vector information 46 for each block variably divided by the motion vector information interpretation unit 103 in FIG. 1 is input (step S1). Subsequently, the size of the sub-macroblock size variably divided before conversion is compared with the size of the macroblock size after conversion (step S2). If the sub-macroblock size variably divided before conversion is smaller than the macroblock size after conversion, the variance is calculated for the horizontal and vertical components of the image of this input motion vector, and these are integrated and normalized. The converted value is calculated as the correlation value ρ of the motion vector (step S3). A method for calculating the correlation value ρ will be described later.

Subsequently, it is compared whether or not the correlation value ρ of the motion vector is equal to or larger than the smaller threshold ρ2 of two preset thresholds ρ1 and ρ2 (where 0 <ρ2 <ρ1 <1) (step S4). , The following inequality ρ2 ≦ ρ <1
Is satisfied, after calculating the average Vm of motion vectors of each sub-macroblock before conversion (step S5), this time, the following inequality ρ ≧ ρ1
Is satisfied (step S6).

  When the correlation value ρ of the motion vector is larger than the first threshold value ρ1 (that is, ρ ≧ ρ1), as described later, the motion vector is in the same direction before conversion and after conversion, and is reused. It is determined that it is possible, and the average Vm of the motion vectors is reused as a motion vector for a macroblock of a coding unit (step S7). On the other hand, if the inequality is not satisfied in step S6 and it is determined that ρ <ρ1, that is, if the correlation value ρ of the motion vector is ρ2 ≦ ρ <ρ1, the direction of the motion vector after conversion is described later. Therefore, it is determined that recalculation is necessary, and the motion vector for the macro block of the coding unit whose search range is changed around the average Vm of the motion vector is recalculated (step) S8).

  On the other hand, when the correlation value ρ does not satisfy the inequality in step S4, that is, when the correlation value ρ of the motion vector is smaller than the second threshold ρ2 (0 <ρ <ρ2), as described later, Since the direction of the motion vector is significantly different from that before the conversion, it is determined that recalculation is necessary, and the motion vector is recalculated in the normal range (step S9).

  As described above, when it is determined that the size of the sub-macroblock size variably divided before conversion in step S2 is smaller than the size of the macroblock size after conversion, the average of the motion vectors before conversion is determined. Any one of reuse by Vm (steps S6 and S7), motion vector recalculation with the search range changed (steps S6 and S8), and motion vector recalculation in the normal range (steps S4 and S9) is selected.

  However, if the sub-macroblock size variably divided before conversion is the same as or larger than the macroblock size after conversion in step S2, the above selection is not performed and the motion vector before conversion is used as it is (step S10). . Finally, the calculated motion vector (56 in FIG. 1) after moving image information conversion is output (step S11).

  According to the present embodiment, it is possible to replace the cost of motion vector recalculation such as block matching with the calculation of the correlation value of the motion vector for each macroblock, thereby reducing one calculation cost. In addition, even when recalculation is performed in the determination, since a motion vector with high accuracy is calculated by a moving image encoding method using a variably divided sub-macroblock, deterioration due to an error can be suppressed as compared with the conventional method. Is possible.

  Here, as an example, the resolution conversion coefficient 52 is “1”, that is, the image resolution is not converted, the size of the macroblock is 16 × 16 pixels, and the size of the divided variable blocks is 16 respectively. Motion vectors calculated in sub-macroblock units that are variably divided before conversion when x8 pixels, 8x16 pixels, 8x8 pixels, 8x4 pixels, 4x8 pixels, and 4x4 pixels Will be described for the case of reusing in units of macroblocks after conversion. In this case, there are cases where the following three block sizes are converted when considered in units of macroblocks.

(1) Before conversion: 16 × 16 blocks → After conversion: 16 × 16 blocks (2) Before conversion: Two 16 × 8 or 8 × 16 blocks → After conversion: 16 × 16 blocks (3) Before conversion: A plurality of variably divided blocks of 8 × 8 or less are included → After conversion: 16 × 16 blocks Hereinafter, a motion vector integration method will be described for these cases. As shown in FIG. 3, when the motion vector calculation macroblock 1010 before conversion has a 16 × 16 pixel size, the area is flat, and the direction of the motion vector 1110 matches the entire area of the macroblock 1010. May be considered. Therefore, the motion vector 2110 of the macroblock 2010 after conversion is the same as the motion vector 1110 before conversion as is apparent from FIG. Therefore, in the case of FIG. 3, the motion vector 1110 before conversion is used as it is to obtain the motion vector 2110 after conversion (step S10 in FIG. 2).

  Subsequently, as shown in FIG. 4, the 16 × 16 pixel size macroblock before conversion is divided into two in the vertical direction, 16 × 8 pixel size sub-macroblocks 1020 and 1021, or in the horizontal direction into two. When two sub-macroblocks 1030 and 1031 having a size of 8 × 16 pixels are provided, two motion vectors 1120 and 1121 or 1130 and 1131 exist in one sub-macroblock.

  In encoding before moving image information conversion, the division of the sub macroblocks 1020 and 1021 or the sub macroblocks 1030 and 1031 can be divided when the directions of the motion vectors are different in the regions. The directions of 1120 and 1121 or 1130 and 1131 are not exactly the same. Therefore, when the direction of the motion vector is remarkably different, if the average of the two motion vectors is simply taken as the converted motion vector 2120 or 2130, the prediction error is considered to increase.

  Therefore, in the present embodiment, as described above, the correlation value ρ (0 ≦ ρ ≦ 1) is obtained from the variance of the horizontal and vertical components of the two motion vectors 1120 and 1121, or 1130 and 1131, and the magnitude of ρ Thus, the difference between the direction and the magnitude of the motion vector is determined by a threshold value.

  Here, a method of calculating the correlation value ρ in step S3 in FIG. 2 will be described. Let the correlation value in i two-dimensional vectors Vi = (xi, yi) (i = 1, 2,..., N) be ρ.

  The correlation value ρ defined here is an index for investigating the correlation of motion vectors because it varies depending on the difference in direction and magnitude by using the variance value of the vector component. It is possible to determine the difference in direction and size of the motion vector based on the reference.

  If the correlation value ρ calculated in this way is greater than 0 and less than 1 and greater than or equal to the first threshold ρ1, the correlation is large, and the converted macroblock 2020 or 2030 shown in FIG. 4 has a 16 × 16 pixel size. Assuming that the motion directions are substantially the same, the average Vm of the two motion vectors calculated before the conversion is reused as the motion vector after the conversion (steps S4 to S7 in FIG. 2).

  Further, when a second threshold ρ2 (0 <ρ2 <ρ1 <1) that is smaller than ρ1 is defined and the correlation value ρ takes a value between these two thresholds (ρ2 ≦ ρ <ρ1), the correlation However, the direction is not significantly different, but it is determined that the error increases when the average Vm is used as a motion vector. Therefore, the motion vector after conversion is recalculated around the point of average Vm of the two motion vectors before conversion (steps S5, S6, and S8 in FIG. 2).

  Furthermore, when the correlation value ρ is smaller than the second threshold ρ2 (ρ <ρ2), it is determined that the directions of the motion vectors are significantly different because the correlation is extremely small. Therefore, the motion vector after conversion is recalculated in the normal search range without using the motion vector before conversion (steps S4 and S9 in FIG. 2). These are summarized as follows.

i) ρ ≧ ρ1
The motion vectors are in almost the same direction and are determined to be reusable. In this case, the average Vm of the two motion vectors 1120 and 1121 or 1130 and 1131 before the conversion is calculated and used as the motion vector 2120 or 2130 after the conversion.

ii) ρ2 ≦ ρ <ρ1
Since there is a slight difference in the direction of the motion vector, it is determined that recalculation is necessary. In this case, the average Vm of the two motion vectors 1120 and 1121 or 1130 and 1131 before conversion is calculated, and the converted motion vector 2120 or 2130 is recalculated around that point.

iii) 0 <ρ <ρ2
Since the direction of the motion vector is significantly different, it is determined that recalculation is necessary. In this case, the average Vm of the two motion vectors 1120 and 1121 or 1130 and 1131 before the conversion is not used, and the converted motion vector 2120 or 2130 is recalculated in the normal search range.

  Next, as shown in FIG. 5A, in a motion vector before conversion in an encoding method (of H.264) in which variably divided blocks are adopted, the block size is 8 × 8 pixel size or less. A motion vector estimation method after conversion in an encoding scheme in which a fixed block is employed will be described in the case of including sub-macroblocks 1040 and 1041 divided into two.

  In this case, similarly to the case where the resolution conversion coefficient 52 is “1” and has two motion vectors before conversion, the motion for each sub-macroblock variably divided before conversion is described above. The correlation value of the vector is calculated, the correlation value is compared with two threshold values, and the converted motion vector 1140, 1141 or the like is reused after conversion according to the comparison result, or the converted motion vector is Recalculate based on the average of motion vectors 1140 and 1141 before conversion, or recalculate the motion vector 2140 after conversion in the normal search range without using the motion vectors 1140 and 1141 before conversion, One of the above is determined.

  However, in this case, since the occupation ratio of the macroblock differs depending on the motion vector, the motion vectors of other submacroblocks are weighted in accordance with the smallest variably divided submacroblock as shown in FIG. Then, the correlation value is calculated. In the case of FIG. 5A, the smallest variably divided sub-macroblock is a block 1040 having a block size of 4 × 4 pixels. Therefore, as shown in FIG. 5B, the block 1041 having a block size of 4 × 8 pixels is used. , It is assumed that there are two apparent motion vectors (dotted line vectors), and there are apparently four blocks in the block 1042 having a block size of 8 × 8 pixels.

  Therefore, in the case of FIG. 5A, as shown in FIG. 5B, in the entire macroblock having a block size of 16 × 16 pixels, there are a total of 16 motion vectors of the vectors indicated by the solid line and the dotted line. I will do it. Therefore, in the case of the converted 16 × 16 pixel macroblock 2040 shown in FIG. 5C, when calculating the converted motion vector 2140, the correlation value ρ and the threshold values ρ1 and ρ2 described above are used. Whether or not the motion vector before conversion can be reused can be determined according to the comparison result.

  The case classification of motion vector reuse determination for the correlation value ρ and the threshold values ρ1 and ρ2 is the same as in the case of having two motion vectors before conversion as described above, and is as follows.

i) ρ ≧ ρ1
The motion vectors are in almost the same direction and are determined to be reusable. In this case, the average of the motion vectors 1140, 1141, etc. before conversion is calculated and set as the motion vector 2140 after conversion.

ii) ρ2 ≦ ρ <ρ1
Since there is a slight difference in the direction of the motion vector, it is determined that recalculation is necessary. In this case, the average of the motion vectors 1140 and 1141 before conversion is calculated, and the converted motion vector 2140 is recalculated around that point.

iii) ρ <ρ2
Since the direction of the motion vector is significantly different, it is determined that recalculation is necessary. In this case, the motion vectors 1140 and 1141 before conversion are not used, and the converted motion vector 2140 is recalculated in the normal search range.

  As described above, when converting only the encoding method without performing the resolution conversion, the motion vector correlation value calculated for each variably divided block before the conversion is used. Therefore, it is possible to reduce the motion vector recalculation process, and at the same time, it is possible to use a highly accurate motion vector as a converted motion vector.

  By the way, the above explanation is an operation explanation when the resolution conversion coefficient 52 is “1”, but the present invention can be applied even when the resolution conversion coefficient 52 is smaller than “1” or larger than “1”. In this case, before the processing in steps S7, S8, S9, and S10 in the flowchart of FIG. 2, the input motion vector average value Vm or the input motion vector is multiplied by the value of the resolution conversion coefficient 52, and the multiplication result is obtained. In step S7, S8, S9, and S10, the processing is performed.

  Then, next, the case where resolution conversion is performed is demonstrated. First, a case where the image resolution conversion coefficient 52 is less than “1”, that is, a case where the image resolution is reduced will be described. FIG. 6 shows a motion vector of a variable block before conversion when the resolution conversion coefficient 52 is “1/2”, a motion vector of a macroblock after conversion, and the like.

  In the case of FIG. 6, 2 × 2 macroblocks 1050 before conversion are used in the horizontal and vertical directions, and weighted by the blocks variably divided in the 2 × 2 macroblocks 1050 before conversion described above. Calculating a correlation value of the motion vector 1150, reusing the motion vector before conversion as the motion vector 2150 of the converted macroblock 2050, re-searching the motion vector after conversion based on the motion vector before conversion; One of motion vector re-searches by a normal search range is selected.

  However, in the case of reuse, the size of the motion vector 1150 before conversion is halved and scaled according to the resolution conversion coefficient 52. The same can be considered when the resolution conversion coefficient 52 is 1 / n (n is a natural number).

  That is, the case classification of the motion vector reuse determination for the correlation value ρ and the threshold values ρ1 and ρ2 is the same as that described above, and is as follows.

i) ρ ≧ ρ1
The motion vectors are in almost the same direction and are determined to be reusable. In this case, the average of the motion vector 1150 before conversion is calculated, and the calculation result is multiplied by the value of the resolution conversion coefficient 52 to reduce it to 1 / n times, and this is used as the motion vector 2150 after conversion.

ii) ρ2 ≦ ρ <ρ1
Since there is a slight difference in the direction of the motion vector, it is determined that recalculation is necessary. In this case, the average of the motion vector 1150 before conversion is calculated, and the calculation result is multiplied by the value of the resolution conversion coefficient 52 to reduce it to 1 / n times, and the converted motion vector centering on the point of the multiplication result Recalculate 2150.

iii) ρ <ρ2
Since the direction of the motion vector is significantly different, it is determined that recalculation is necessary. In this case, the motion vector 1150 before conversion is not used, and the converted motion vector 2150 is recalculated in the normal search range. As described above, even when the encoding method and image resolution are reduced and converted, the converted motion vector can be calculated at high speed by using the correlation value of the motion vector before conversion.

  Next, a case where the resolution conversion coefficient 52 is larger than “1” (natural number m times) will be described. FIG. 7 shows the motion vector of a variable block before conversion when the resolution conversion coefficient 52 is “3” (m = 3), the motion vector of a macroblock after conversion, and the like. In this case, first, while maintaining the motion vector information 1160 for each of the sub-macroblocks 1060 variably divided before conversion shown in FIG. 7A, the area is tripled as shown in FIG. The enlarged macro block 1061 having a block size of 3 × 3 macro blocks each having 16 × 16 pixels is created, and the correlation value of the motion vector 1160 is obtained in each sub macro block region.

  Similarly to the above, as shown in FIG. 7C, the motion vector 2160 is estimated for each 16 × 16 pixel macroblock region of the 3 × 3 macroblocks 2060 after conversion. When the motion vector 1160 before conversion is reused, it is multiplied by the resolution conversion coefficient 52 and used. Therefore, the case classification of the motion vector reuse determination for the correlation value ρ and the threshold values ρ1 and ρ2 is the same as described above, and is as follows.

i) ρ ≧ ρ1
The motion vectors are in almost the same direction and are determined to be reusable. In this case, the average of the motion vector 1160 before conversion is calculated, the calculation result is multiplied by the value m of the resolution conversion coefficient 52, and the result is expanded M times, and the multiplication result is used as the converted motion vector 2160.

ii) ρ2 ≦ ρ <ρ1
Since there is a slight difference in the direction of the motion vector, it is determined that recalculation is necessary. In this case, the average of the motion vector 1160 before conversion is calculated, the calculation result is multiplied by the value m of the resolution conversion coefficient 52, and the result is expanded to M times. Is recalculated.

iii) ρ <ρ2
Since the direction of the motion vector is significantly different, it is determined that recalculation is necessary. In this case, the motion vector 1160 before conversion is not used, and the converted motion vector 2160 is recalculated in the normal search range.

  As described above, according to the present embodiment, the motion vector correlation value before encoding moving image conversion is used and compared with two preset thresholds ρ1 and ρ2, thereby estimating the motion vector after conversion. To re-use the motion vector before conversion, re-calculate the motion vector after conversion based on the motion vector before conversion, or re-calculate the motion vector in the normal motion vector search range Therefore, the encoding efficiency in the re-encoding process can be improved.

  Note that the present invention is not limited to the above-described embodiment. For example, in the above-described embodiment, the motion vector obtained by the variably divided sub-macroblock is expressed in units of normal macroblocks. However, the converted macroblock size may include a block having a size of 8 × 8 pixels, such as an MPEG-4 sub-block. In this case, when the sub-macroblock size variably divided before conversion includes macroblocks of 4 × 8 pixels, 8 × 4 pixels, and 4 × 4 pixels, the motion vector after conversion based on the correlation value is estimated. Can be done.

  In the above embodiment, the macroblock size is set to 16 × 16 pixels, but other sizes may be used. Similarly, sub macroblocks that are variably divided can be combined in accordance with the macroblock size.

  Further, the present invention includes a computer program for realizing the moving image information conversion coding apparatus of the above-described embodiment by a computer. In this case, the computer program may be taken into the computer via a recording medium or may be taken into the computer via a communication network.

[BRIEF DESCRIPTION OF THE DRAWINGS] It is a block diagram of one Embodiment of the moving image information conversion encoding apparatus of this invention. It is a flowchart for description of the motion vector reuse procedure in the embodiment of FIG. It is a figure which shows an example of motion vector conversion when the block size before the conversion in this invention apparatus is 16x16 pixel, and the block size after conversion is 16x16 pixel. It is a figure which shows an example of motion vector conversion when the block size before the conversion in this invention apparatus has 16x8 pixels or two 8x16 pixels, and the block size after conversion is 16x16 pixels. It is a figure which shows an example of motion vector conversion when the block size before the conversion in this invention apparatus has a block of a magnitude | size of 8x8 pixels or less, and the block size after conversion is 16x16 pixels. It is a figure which shows an example of the motion vector conversion when the resolution conversion coefficient in this invention apparatus is less than 1 (n = 1/2). It is a figure which shows an example of the motion vector conversion when the resolution conversion coefficient in this invention apparatus is set to larger than 1 (m = 3). It is a figure which shows the block divided | segmented variably in a moving image encoding.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Encoding information input terminal 20 Resolution conversion coefficient input terminal 30 Encoding information output terminal 100 Moving picture decoding part 101 Moving picture information separation part 102 Variable length decoding part 103 Motion vector information interpretation part 104 Motion vector compensation part 105,307 Inverse quantum Conversion unit 106, 308 inverse coefficient conversion unit 108, 303 image memory unit 200 moving image information conversion unit 201 image resolution conversion unit 202 motion vector estimation unit 300 moving image encoding unit 301 motion vector search unit 302 motion compensation unit 305 coefficient conversion unit 306 Quantization unit 310 Variable length coding unit 311 Moving picture information synthesis unit

Claims (2)

For each variable first encoding unit, the first moving image encoded signal encoded by the first encoding method is decoded based on the first motion vector, and the decoded moving image information is fixed to the fixed first unit. A moving picture information transform coding apparatus that generates a second moving picture coded signal by re-encoding by a second coding scheme based on a second motion vector for every two coding units,
Motion vector information interpreting means for obtaining the first motion vector from a signal obtained by decoding the first moving image encoded signal;
Image resolution conversion means for performing image resolution conversion based on an image resolution conversion coefficient for moving image information obtained by decoding the first moving image encoded signal;
Correlation value calculating means for calculating a correlation value of the first motion vector obtained by the motion vector information interpreting means when the first coding unit is smaller than the second coding unit;
A motion vector estimating means for comparing the calculated correlation value with a preset threshold value and estimating the second motion vector according to the comparison result;
In accordance with the second motion vector estimation result obtained by the motion vector estimation means, the first motion vector based on the multiplication result of the average of the first motion vector and the image resolution conversion coefficient Select either re-calculation by averaging or re-calculation of motion vector with changed search range, or re-calculation of motion vector in the normal range based on the result of multiplication of the first motion vector and the image resolution conversion coefficient Motion vector search means for generating the second motion vector;
Based on the moving image information whose image resolution has been converted by the image resolution conversion means and the second motion vector generated by the motion vector search means, re-encoding is performed by the second encoding method, and the second A moving picture information conversion coding apparatus comprising: coding means for generating two moving picture coded signals. The encoding means replaces the second motion vector generated by the motion vector search means when the first encoding unit is larger than the second encoding unit, and the motion vector information Using the first motion vector obtained by the interpreting means and the image resolution conversion coefficient, the moving image information whose image resolution is converted by the image resolution converting means is re-encoded by the second encoding method. The moving image information conversion encoding apparatus according to claim 1, wherein:

Images