JP5255512B2 - Encoding device and program thereof - Google Patents

Description

  The present invention relates to an encoding apparatus that compresses and encodes a moving image by performing motion compensation of the moving image including a plurality of frame images, and a program thereof.

  Conventionally, AVC (Advanced Video Coding) / H. H.264, MPEG (Moving Picture Experts Group) -2, and the like are widely used (see, for example, Patent Document 1). In the conventional compression coding method, motion compensation / prediction is performed in which a motion from a reference frame image that has already been encoded to a current frame image is predicted to perform motion compensation, and a predicted frame image of the current frame image is generated. Yes. In the conventional compression coding method, the DCT transform is performed on the difference image between the current frame image and the predicted frame image, and the DCT transform coefficient is quantized with the quantization parameter.

JP 2005-244503 A

  However, AVC / H. In the H.264 compression coding system, when compression coding with a fixed quantization parameter is performed, the conspicuous visual deterioration of each macroblock is not considered. Specifically, AVC / H. In the H.264 compression coding scheme, the same quantization parameter is used for a macroblock with a simple motion pattern and a macroblock with a complex motion pattern. For this reason, AVC / H. In the H.264 compression coding system, a macroblock with a smooth motion is conspicuous in a macroblock with a strong motion, and there is a strong demand to reduce the bit rate while suppressing the degradation of the image quality of a moving image. ing.

  Therefore, an object of the present invention is to provide an encoding apparatus and a program for the same that can reduce the bit rate while suppressing the deterioration of the image quality of moving images.

  In order to solve the above-described problem, an encoding apparatus according to the first aspect of the present invention provides a motion prediction in an encoding apparatus that performs motion compensation of a moving image composed of a plurality of frame images and compresses and encodes the moving image. A motion compensation unit, a difference image generation unit, a motion vector dispersion value calculation unit, a quantization parameter correction unit, and a quantization unit.

  According to this configuration, the encoding device predicts a motion vector for each block obtained by dividing the moving image from the reference frame image of the current frame image in the moving image by the motion prediction unit. This block includes, for example, a macro block having a predetermined size and a sub macro block smaller than the macro block. Also, the encoding device performs motion compensation on the basis of the reference frame image and the motion vector by the motion compensation unit, and generates a predicted frame image in which the current frame image is predicted. Then, the encoding device generates a difference image that is a difference between the current frame image and the predicted frame image by the difference image generation unit.

In addition, the encoding device calculates an average vector obtained by averaging the motion vectors of all the blocks based on the motion vector predicted by the motion prediction unit, and calculates a difference between the motion vector and the average vector. Is calculated, and the variance is calculated by summing the values. This variance value indicates the complexity of the movement. That is, the encoding apparatus calculates the motion complexity of each macroblock by the motion vector variance value calculation unit. The encoding device by the quantization parameter correction unit, a motion is made coarser quantization for complex parts, movement finer quantization is preset to be performed in a smooth portion, variance and Quantum Based on the correspondence relationship with the quantization parameter correction value, a quantization parameter correction value is calculated from the variance value, and a preset quantization parameter is corrected with the quantization parameter correction value. That is, the encoding apparatus reflects the complexity of motion in the quantization parameter by the quantization parameter correction unit. Furthermore, the encoding apparatus quantizes the orthogonal transform coefficient of the difference image based on the quantization parameter corrected by the quantization parameter correction unit by the quantization unit. As a result, the encoding device performs coarse quantization on the part where the movement that is difficult to visually degrade is complex (the part where the sensitivity of the human eye is low), and the part where the movement that is visually noticeable is smooth is smooth. Fine quantization is performed at the part where the sensitivity of the human eye is high.

  In the encoding device according to the second invention of this application, the motion vector variance value calculation unit divides the magnitude of the motion vector by the number of frame intervals between the current frame image and the reference frame image, and based on the divided motion vector, A variance value is calculated.

  Here, for example, AVC / H. In a compression encoding method such as H.264, not only a frame image adjacent to the current frame image but also a frame image separated by two or more frame images before and after can be used as a reference frame image (bi-prediction). In this case, the amount of motion of the motion vector increases as the reference frame image moves away from the current frame image. Therefore, when this motion vector is used as it is, the calculated variance value does not accurately indicate the complexity of the motion. Therefore, the encoding apparatus accurately calculates the variance value by correcting the motion amount of the motion vector according to the number of frame intervals between the current frame image and the reference frame image.

  In the first and second inventions of the present application, the quantization parameter correction unit calculates the quantization parameter correction value within a range of −2 to 2, and adds the calculated quantization parameter correction value to the quantization parameter or You may subtract. In this case, the encoding apparatus can prevent overcorrection of the quantization parameter.

In order to solve the above-described problem, an encoding program according to the third invention of the present application performs motion compensation on a moving image composed of a plurality of frame images, and moves a computer to compress and encode the moving image. prediction unit, a motion compensation unit, the difference image generation unit, a motion vector variance value calculation unit, a quantization parameter correction unit, Ru to function as a quantization unit.

  According to this configuration, the encoding program predicts a motion vector for each block obtained by dividing the moving image from the reference frame image of the current frame image in the moving image by the motion prediction unit. In addition, the encoding program performs motion compensation based on the reference frame image and the motion vector by the motion compensation unit, and generates a predicted frame image in which the current frame image is predicted. Then, the encoding program generates a difference image as a difference between the current frame image and the predicted frame image by the difference image generation unit.

In addition, the encoding program calculates an average vector obtained by averaging the motion vectors of all blocks based on the motion vector predicted by the motion prediction unit, and calculates a difference between the motion vector and the average vector. Is calculated, and the variance is calculated by summing the values. That is, the encoding program calculates the motion complexity of each macroblock by the motion vector variance value calculation unit. The encoding program by the quantization parameter correction unit, motion coarser quantization is done in the complex part, movement finer quantization is preset to be performed in a smooth portion, variance and Quantum Based on the correspondence with the quantization parameter correction value, the quantization parameter correction value is calculated from the variance value, and a preset quantization parameter is added and subtracted with the quantization parameter correction value to be corrected. That is, in the encoding program, the quantization parameter is reflected on the quantization parameter by the quantization parameter correction unit. Further, the encoding program quantizes the orthogonal transform coefficient of the difference image based on the quantization parameter corrected by the quantization parameter correction unit by the quantization unit. As a result, the encoding program performs rough quantization on the parts where movements that are difficult to visually degrade are complex (parts where the sensitivity of the human eye is poor), and the parts where movements that are visually noticeable are smooth are smooth. Fine quantization is performed in the area where the sensitivity of the human eye is improved.

According to the present invention, the following excellent effects can be obtained.
According to the first and third inventions of the present application, coarse quantization is performed in a portion where movement that is difficult to visually deteriorate is complex, and fine quantization is performed in a portion where movement that is visually conspicuous is smooth, The bit rate can be reduced while suppressing the deterioration of the image quality of the moving image.

  According to the second aspect of the present invention, since the variance value is accurately calculated by correcting the motion amount of the motion vector according to the number of frame intervals between the current frame image and the reference frame image, the image quality degradation of the moving image is minimized. Can be suppressed.

It is a block diagram which shows the structure of the encoding apparatus which concerns on embodiment of this invention. It is a figure explaining the block type in embodiment of this invention, (a) shows a macroblock type, (b) shows a submacroblock type. FIG. 2 is a diagram for explaining correction of a motion vector by a motion vector dispersion value calculation unit in FIG. 1, (a) is a diagram in which an adjacent frame image is a reference frame image, and (b) is a frame image separated by m frames. It is the figure made into the reference frame image. It is a figure explaining the assignment of the vector by the motion vector dispersion | distribution value calculation part of FIG. 1, (a) is a figure before assigning a vector, (b) is a figure after assigning a vector. It is a flowchart which shows operation | movement of the encoding apparatus of FIG.

[Configuration of Encoding Device]
With reference to FIG. 1, the structure of the encoding apparatus based on embodiment of this invention is demonstrated. The encoding device 1 is an AVC / H. In accordance with the H.264 compression encoding method, motion compensation of a moving image composed of a plurality of frame images is performed, and the moving image is compressed and encoded. As illustrated in FIG. 1, the encoding device 1 includes a difference image generation unit 11, an orthogonal transformation unit 12, a quantization unit 13, an inverse quantization unit 14, an inverse orthogonal transformation unit 15, and a motion prediction unit 16. A motion compensation unit 17, a motion vector dispersion value calculation unit 18, a quantization parameter correction unit 19, and a multiplexing unit 20.

  The difference image generation unit 11 generates a difference image that is a difference between a current frame image to be encoded and a predicted frame image generated by the motion compensation unit 17 to be described later in a frame image constituting the moving image. It is. That is, the difference image generation unit 11 performs the AVC / H. In accordance with the H.264 compression encoding method, a motion compensated prediction error image of the current frame image is output. At this time, the difference image generation unit 11 receives a frame image as a video signal in which Y (luminance signal) Cb (color difference signal) Cr (color difference signal) has a ratio of “4: 2: 2”. Note that the difference image generation unit 11 may output the current frame image to the orthogonal transform unit 12 as it is when performing intra coding (intra-screen coding).

The orthogonal transform unit 12 performs orthogonal transform on the difference image from the difference image generation unit 11 and outputs an orthogonal transform coefficient (DCT transform coefficient) as a result. Here, regarding the luminance signal of the difference image, the orthogonal transform unit 12 divides each macro block into 4 × 4 pixel sub macro blocks, and performs integer precision DCT transform (integer precision discrete cosine transform). Note that, when performing intra coding, the orthogonal transform unit 12 divides each macro block into 4 × 4 pixel sub-macro blocks and performs integer precision DCT transform for the luminance of the difference image. Then, the orthogonal transformation unit 12, collectively the DC component of the 16 sub-macroblocks integer precision DCT transform, generates a DC component blocks of 4 × 4. Then, the orthogonal transform unit 12 performs a discrete Hadamard transform on this DC component block.

Further, the orthogonal transform unit 12 divides each macroblock into 4 × 4 pixel sub-macroblocks and performs integer precision DCT transform on the color difference signal of the difference image. Then, the orthogonal transformation unit 12, collectively the DC component of the four sub-macroblocks integer precision DCT transform, to produce a 2 × 2 DC components blocks. Then, the orthogonal transform unit 12 performs a discrete Hadamard transform on this DC component block.

  When the encoding device 1 is applied to other compression encoding methods such as MPEG2 and MPEG4, the orthogonal transformation unit 12 may perform orthogonal transformation such as DCT transformation (discrete cosine transformation).

  The quantization unit 13 quantizes the orthogonal transformation coefficient from the orthogonal transformation unit 12 based on the quantization parameter corrected by the quantization parameter correction unit 19 described later. Here, the quantization unit 13 performs AVC / H. In accordance with the H.264 compression encoding method, the orthogonal transform coefficient is quantized so that the quantization step is doubled when the quantization parameter is increased by “6”. Then, the quantization unit 13 outputs the quantized data obtained by quantizing the orthogonal transform coefficient to the inverse quantization unit 14 and the multiplexing unit 20.

  The inverse quantization unit 14 performs inverse quantization on the quantized data from the quantization unit 13. Here, the inverse quantization unit 14 performs AVC / H. In accordance with the H.264 compression encoding method, the quantized data is inversely quantized and returned to the orthogonal transform coefficient. Then, the inverse quantization unit 14 outputs the inversely quantized orthogonal transform coefficient to the inverse orthogonal transform unit 15.

  The inverse orthogonal transform unit 15 performs inverse orthogonal transform on the orthogonal transform coefficient from the inverse quantization unit 14. Here, the inverse orthogonal transform unit 15 performs inverse orthogonal transform on the orthogonal transform coefficient to return to the video signal (reference frame image) in the reverse procedure of the orthogonal transform unit 12. Then, the inverse orthogonal transform unit 15 temporarily stores the reference frame image in a frame memory (not shown) and outputs the reference frame image stored in the frame memory to the motion prediction unit 16 and the motion compensation unit 17. .

  The motion prediction unit 16 predicts a motion vector for each block from the reference frame image output by the inverse orthogonal transform unit 15. Here, the motion prediction unit 16 performs AVC / H. In accordance with the H.264 compression encoding method, a motion vector is predicted using a current frame image and a reference frame image which is a frame image before and after the current frame image. Specifically, the motion prediction unit 16 selects any one of the seven types of block types as shown in FIG. 2, and divides the current frame image into the selected block types. Then, the motion prediction unit 16 predicts a motion vector with reference to one or more reference frame images for each divided macro block and sub macro block. At this time, the motion prediction unit 16 selects the median value (median value) of the motion vectors of the left block, the upper block, and the upper right block adjacent to the encoding target macro block and sub macro block for each vertical component and horizontal component. Then, the motion vector of the encoding target block is predicted.

  Further, when the macroblock to be encoded is 8 × 16 pixels or 16 × 8 pixels, the motion prediction unit 16 uses this code among the motion vectors of the left block, the upper block, and the upper right block instead of the above-described median value. A motion vector is predicted using the one indicating the macroblock to be converted. Then, the motion prediction unit 16 generates a motion vector signal including the predicted motion vector, the reference frame image number, and the like, and outputs the motion vector signal to the motion compensation unit 17, the motion vector variance value calculation unit 18, and the multiplexing unit 20.

  The motion compensation unit 17 performs motion compensation on the basis of the reference frame image from the inverse orthogonal transform unit 15 and the motion vector (motion vector signal) from the motion prediction unit 16, and obtains a predicted frame image obtained by predicting the current frame image. Is to be generated. Here, the motion compensation unit 17 performs AVC / H. Motion compensation is performed in accordance with the H.264 compression coding system. Specifically, the motion compensation unit 17 uses a 6-tap FIR (Finite Impulse Response) filter with each coefficient “1, −5, 20, 20, −5, 1/32” for the luminance signal, A prediction signal with 1/2 pixel accuracy is generated. Then, the motion compensation unit 17 applies a 2-tap average value filter to the prediction signal to generate a 1/4 pixel accuracy prediction signal.

In addition, for the color difference signal, the motion compensation unit 17 linearly interpolates four pixels (color difference signal) with 1/8 pixel accuracy to generate a prediction signal. Specifically, the motion compensation unit 17 multiplies each of the four pixels (color difference signals) by a coefficient corresponding to the distance from each pixel to the pixel to be interpolated to generate a prediction signal with 1/8 pixel accuracy. To do. Then, the motion compensation unit 17 outputs these prediction signals to the difference image generation unit 11 as prediction frame images.

Based on the motion vector (motion vector signal) from the motion prediction unit 16, the motion vector variance value calculation unit 18 calculates an average vector obtained by averaging the motion vectors of all the blocks, and calculates the difference between the motion vector and the average vector. In addition to calculating the squared value, the values are summed to calculate the variance value. Specifically, the motion vector variance value calculation unit 18 calculates an average vector obtained by averaging the motion vectors of all blocks, as represented by the following formula (1). In Equation (1), MV indicates a motion vector, MV _ave indicates an average vector, and N indicates the total number of blocks.

In addition, the motion vector variance value calculation unit 18 calculates a value obtained by squaring the difference between the motion vector and the average vector for each block, as shown by the following formula (2), and sums this value to obtain a variance value. Is calculated. Thereafter, the motion vector variance value calculation unit 18 outputs the calculated variance value to the quantization parameter correction unit 19. In Equation (2), MV represents a motion vector, MV _ave represents an average vector, N represents the total number of blocks, and X represents a variance value.

<Correction of Motion Vector by Motion Vector Variance Calculation Unit>
Hereinafter, the correction of the motion vector by the motion vector variance value calculation unit 18 will be described with reference to FIG. 3 (see FIG. 1 as appropriate). In FIG. 3A and FIG. 3B, the motion vector is indicated by an arrow.

  For example, AVC / H. In the H.264 compression encoding method, not only the frame image adjacent to the current frame image is set as the reference frame image as shown in FIG. 3A, but also m frames are separated as shown in FIG. 3B. The frame image can be a reference frame image. At this time, in FIG. 3A and FIG. 3B, the motion amount of the motion vector of the current frame image is the same. However, in FIG. 3B, since the reference frame image and the current frame image are separated from each other by m, it goes without saying that the movement per unit time is smaller than that in FIG. In other words, the greater the distance from the current frame image to the reference frame image, the greater the movement. Therefore, if the motion amount of the motion vector is used as it is, the calculated variance value does not accurately indicate the motion complexity as compared to when referring to an adjacent frame.

  For this reason, the motion vector variance value calculation unit 18 represents the size (motion amount) of the motion vector from the motion prediction unit 16 as the number of frame intervals between the current frame image and the reference frame image, as represented by Expression (3). It is preferable to divide by. Thereafter, the motion vector variance value calculation unit 18 calculates a variance value using the above-described equations (1) and (2) based on the corrected motion vector. As described above, the motion vector variance value calculation unit 18 corrects the motion vector motion amount according to the number of frame intervals between the current frame image and the reference frame image, and accurately calculates the variance value. Degradation can be minimized. In Equation (3), MV represents a motion vector, m represents the number of frame intervals between the current frame image and the reference frame image, and MV ′ represents a motion vector after correction.

<Vector allocation by motion vector variance calculation unit>
Hereinafter, with reference to FIG. 4, vector assignment by the motion vector variance value calculation unit 18 will be described (see FIG. 1 as appropriate). 4A and 4B, the blocks divided by the motion prediction unit 16 are illustrated by solid lines, and the motion vectors are illustrated by solid arrows. Further, in FIG. 4B, the blocks divided by the motion vector variance value calculation unit 18 are indicated by broken lines, and the assigned vectors are indicated by broken line arrows.

  For example, AVC / H. In the H.264 compression encoding method, as shown in FIG. 4A, motion vectors can be generated in units of sub-macroblocks smaller than macroblocks, and motion prediction / compensation can be performed. In this case, in a certain block, the sub macroblock region has more motion vectors than the other regions. For example, in FIG. 4A, a 16 × 16 pixel macroblock is divided into four sub macroblocks. Here, four motion vectors are generated for each 4 × 4 pixels in the upper left sub-macroblock, and one motion vector is generated in the lower left sub-macroblock. In the upper right sub-macroblock, two motion vectors are generated every 8 × 4 pixels, and in the lower right sub-macroblock, two motion vectors are generated every 4 × 8 pixels. That is, as shown in FIG. 4A, the motion vector is biased toward the upper left sub-macroblock that is the smallest of the macroblocks. In this case, the calculated variance value does not accurately indicate the motion complexity of the entire block.

  For this reason, it is preferable that the motion vector dispersion value calculation unit 18 assigns the same vector as the motion vector for every four vertical pixels and four horizontal pixels so that the motion vectors are uniform in the block. Specifically, as shown in FIG. 4B, the motion vector variance value calculation unit 18 divides the upper right sub-macroblock of 8 × 4 pixels into two in the horizontal direction so as to be 4 × 4 pixel units. To do. Then, the motion vector variance value calculation unit 18 refers to the motion vector in the vertical direction so that the motion vector is arranged for every 4 × 4 pixels in the upper right sub-macroblock, and the same vector is calculated as 2 Allocate.

  In addition, the motion vector variance value calculation unit 18 divides the lower right submacroblock of 4 × 8 pixels into two in the vertical direction so as to be 4 × 4 pixel units. Then, the motion vector variance value calculation unit 18 refers to the motion vector in the horizontal direction so that the motion vector is arranged for every 4 × 4 pixels in the lower right sub-macroblock, and uses the same vector as this. Assign two.

  In addition, the motion vector variance value calculation unit 18 divides the 8 × 8 pixel lower left sub-macroblock into four so as to form a 4 × 4 pixel unit. Then, the motion vector variance value calculation unit 18 assigns four vectors identical to the motion vector so that a motion vector is arranged for every 4 × 4 pixels in the lower left sub-macroblock. Thereafter, the motion vector variance value calculation unit 18 uses all the motion vectors and the assigned vectors to calculate the variance values using the above-described formulas (1), (2), and (3).

  Thus, the motion vector variance value calculation unit 18 allocates vectors so that the motion vectors are equal within the macroblock, thereby eliminating the motion vector bias between the sub-macroblocks and accurately calculating the variance value. Since it can be calculated, it is possible to minimize degradation of the image quality of the moving image. It should be noted that the motion vector variance value calculation unit 18 does not have to assign a vector to the sub-macroblock divided into 4 × 4 pixels (the sub-macroblock at the upper left in FIG. 4).

Hereinafter, returning to FIG. 1, the description of the configuration of the encoding device 1 will be continued.
The quantization parameter correction unit 19 calculates a quantization parameter correction value from the dispersion value calculated by the motion vector dispersion value calculation unit 18 based on the correspondence between the preset dispersion value and the quantization parameter correction value. It is. Then, the quantization parameter correction unit 19 corrects the preset quantization parameter with the quantization parameter correction value, and outputs this quantization parameter (shown as QP in FIG. 1) to the quantization unit 13.

  The quantization parameter is, for example, AVC / H. In the H.264 compression encoding method, the value is from “0” to “51”. In order to use the fixed quantization parameter, in the encoding device 1, the quantization parameter is manually set by an operator or the like as a fixed value of the quantization parameter from “0” to “51” for each sequence, for example. The

<Calculation method of dispersion value: first example>
Hereinafter, a specific example of a method for calculating the dispersion value by the quantization parameter correction unit 19 will be described.
The quantization parameter correction unit 19 calculates a quantization parameter correction value using the correspondence relationship represented by the following equation (4). In Equation (4), T1 to T4 represent threshold values, X represents a dispersion value, and ΔA to ΔE represent quantization parameter correction values.

  Here, the quantization parameter correction unit 19 preferably calculates the quantization parameter correction value within a range of −2 or more and 2 or less. For example, in the correspondence relationship expressed by the above-described formula (4), ΔA is “2”, ΔB is “1”, ΔC is “0”, ΔD is “−1”, and ΔE is “−2”. It is set in advance. At this time, in Expression (4), for example, T1 is set to “4000”, T2 is set to “3000”, T3 is set to “2000”, and T4 is set to “1000” in advance. By using such a correspondence relationship, the quantization parameter correction unit 19 can calculate the quantization parameter within a range of −2 to 2. As a result, the quantization parameter correction unit 19 can prevent over-correction of the quantization parameter, so that the quantization parameter can be maintained at an appropriate value, and adverse effects such as an increase in bit rate can be suppressed.

<Calculation method of dispersion value: second example>
The quantization parameter correction unit 19 sets the quantization parameter correction value to “+2” when the variance value is equal to or greater than a preset threshold value (eg, 2500), and when the variance value is less than the threshold value, the quantization parameter correction value. The quantization parameter correction value is calculated using a correspondence relationship in which “0” is set to “0”.

  The multiplexing unit 20 performs multiplexing based on the quantized data from the quantization unit 13 and the motion vector (motion vector information) from the motion prediction unit 16. Here, the multiplexing unit 20 is connected to the AVC / H. Entropy coding (variable length coding) is performed in accordance with the H.264 compression coding system. Specifically, the multiplexing unit 20 generates multiplexed data (encoded data) obtained by multiplexing the quantized data using CAVLC (context-applied variable length encoding method). Further, the multiplexing unit 20 may multiplex the quantized data and the motion vector information using CABAC (context application type binary arithmetic coding method). Then, the multiplexing unit 20 outputs this multiplexed data to a decoding device (not shown).

  Note that the encoding device 1 may include an A / D conversion unit that performs analog-digital conversion of a moving image and a frame image rearrangement unit that rearranges frame images before the difference image generation unit 11. Further, the encoding device 1 may include a deblocking filter that reduces block distortion between the inverse orthogonal transform unit 15 and the motion prediction unit 16.

[Operation of encoding apparatus]
Hereinafter, the operation of the encoding device of FIG. 1 will be described with reference to FIG. 5 (see FIG. 1 as appropriate). First, the encoding apparatus 1 performs inverse orthogonal transform on the orthogonal transform coefficient by the inverse orthogonal transform unit 15 to generate a reference frame image (step S11).

  Following the processing of step S11, the encoding apparatus 1 selects any of the seven types of block types by the motion prediction unit 16, and divides the current frame image by the selected block type (step S12). Also, the encoding apparatus 1 predicts a motion vector for each divided block by the motion prediction unit 16 (step S13).

  Following the processing of step S13, the encoding apparatus 1 performs motion compensation by the motion compensation unit 17 based on the reference frame image from the inverse orthogonal transform unit 15 and the motion vector (motion vector signal) predicted in step S13. A predicted frame image obtained by predicting the current frame image is generated (step S14).

  Following the processing in step S14, the encoding apparatus 1 causes the difference image generation unit 11 to generate a difference image that is a difference between the current frame image to be encoded and the predicted frame image generated in step S14. (Step S15).

  Following the processing of step S15, the encoding apparatus 1 performs orthogonal transformation such as integer precision DCT transformation and discrete Hadamard transformation on the difference image generated in step S15 by the orthogonal transformation unit 12, and obtains an orthogonal result. A conversion coefficient (DCT conversion coefficient) is output (step S16).

  Further, following the processing of step S13, the encoding apparatus 1 uses the motion vector variance value calculation unit 18 to average the motion vectors of all blocks based on the motion vector (motion vector signal) from the motion prediction unit 16. The calculated average vector is calculated, a value obtained by squaring the difference between the motion vector and the average vector is calculated, and the values are summed to calculate a variance value (step S17). Here, the encoding apparatus 1 preferably performs the above-described motion vector correction and vector assignment by the motion vector variance value calculation unit 18.

  Following the processing in step S17, the encoding apparatus 1 uses the quantization parameter correction unit 19 to calculate a quantization parameter correction value corresponding to the variance value calculated in step S17 using a preset correspondence relationship. (Step S18). Here, the quantization parameter correction unit 19 calculates the quantization parameter correction value in the first example or the second example of the dispersion value calculation method described above.

  Following the processing in step S18, the encoding apparatus 1 corrects the preset quantization parameter by the quantization parameter correction unit 19 with the quantization parameter correction value calculated in step S18 (step S19). Here, the encoding device 1 uses the quantization parameter correction unit 19 to calculate a quantization parameter correction value within a range of −2 to 2 and adds the calculated quantization parameter correction value to the quantization parameter or Subtraction is preferred.

  That is, the encoding device 1 executes the processes in steps S14 to S16 and the processes in steps S17 to S19 in parallel.

  Following the processing of step S19, the encoding apparatus 1 quantizes the orthogonal transform coefficient orthogonally transformed in step S16 by the quantization unit 13 based on the quantization parameter corrected in step S19 (step S20).

  Following the processing in step S20, the encoding device 1 performs multiplexing such as entropy encoding based on the quantized data quantized in step S20 and the motion vector (motion vector information) predicted in step S13. This is performed (step S21).

  As described above, the encoding device 1 according to the present embodiment of the present invention performs rough quantization in a portion where the motion that is difficult to visually degrade is complex, and the motion that is easily noticeable in visual degradation is complex. Since fine quantization is performed in the portion, the bit rate can be reduced while suppressing the deterioration of the image quality of the moving image. In addition, in the encoding device 1 according to the present embodiment of the present invention, each unit other than the motion vector variance value calculation unit 18 and the quantization parameter correction unit 19 includes an AVC / H. Since it complies with the H.264 compression encoding system, it is easy to implement and excellent in cost. Furthermore, the encoding apparatus 1 according to the present embodiment of the present invention converts the generated multiplexed data into AVC / H. Since decoding can be carried out as it is with a decoding device compliant with the H.264 compression encoding method, a dedicated decoding device is not required and introduction is easy.

  In the present embodiment, the encoding apparatus 1 is connected to the AVC / H. Although an example of application to the H.264 compression coding system has been described, it goes without saying that any coding system that performs motion compensation can be applied to other coding systems. In this case, the encoding device 1 is easy to implement and excellent in cost because each unit other than the motion vector variance calculation unit 18 and the quantization parameter correction unit 19 conforms to the encoding scheme to be applied. Note that the encoding apparatus 1 can use the invention described in Japanese Patent Laid-Open No. 2005-12527 as the motion prediction unit 16 and the motion compensation unit 17.

  Although the frame image has been described as a “4: 2: 2” format, the present invention is not limited to this. For example, the present invention can also handle frame images in the “4: 2: 0” format and monochrome format.

  In the present embodiment, the encoding device 1 has been described as an independent device, but in the present invention, hardware resources such as a general computer storage means and an arithmetic device are operated in a cooperative manner as the above-described units. It can also be realized. This program may be distributed via a communication line, or may be distributed by writing in a recording medium such as a CD-ROM or a flash memory.

  Hereinafter, an example of the present invention will be briefly described (see FIG. 1 as appropriate). The encoding apparatus 1 according to the present embodiment employs the above-described second example as a variance value calculation method. Specifically, in the quantization parameter correction unit 19, when the threshold value is “2500” or more, the correlation is set to “+1” as the quantization parameter correction value. The quantization parameter value was set to “46” as a fixed value for each sequence.

  The encoding apparatus 1 that performs this setting and the AVC / H. In addition to the conventional encoding apparatus conforming to the H.264 compression encoding method, the ITE standard moving image No. 46 (sprinkling) was input. At this time, the bit rate of the encoded data output from both encoding apparatuses was measured. In addition, AVC / H. In a decoding device compliant with the H.264 compression encoding method, encoded data output from both encoding devices was decoded, and these moving images were compared visually.

  As a result, it was found that the encoding apparatus 1 according to the present embodiment can reduce the bit rate of encoded data by about 3.5% as compared with the conventional encoding apparatus. In addition, in the moving image obtained by decoding the encoded data of both the encoding devices, the difference in image quality cannot be recognized. That is, it has been found that the encoding apparatus 1 according to the present embodiment can suppress the deterioration of the image quality of the moving image to the same extent as the conventional encoding apparatus.

DESCRIPTION OF SYMBOLS 1 Encoding apparatus 11 Difference image generation part 12 Orthogonal transformation part 13 Quantization part 14 Inverse quantization part 15 Inverse orthogonal transformation part 16 Motion prediction part 17 Motion compensation part 18 Motion vector dispersion value calculation part 19 Quantization parameter correction part 20 Multiplexing Chemical department

Claims (3)

In an encoding device that performs motion compensation of a moving image composed of a plurality of frame images and compresses and encodes the moving image,
A motion prediction unit that predicts a motion vector for each block obtained by dividing the moving image from a reference frame image of the current frame image in the moving image;
A motion compensation unit that performs motion compensation based on the reference frame image and the motion vector, and generates a predicted frame image obtained by predicting the current frame image;
A difference image generation unit that generates a difference image that is a difference between the current frame image and the predicted frame image;
Based on the motion vector predicted by the motion prediction unit, an average vector obtained by averaging the motion vectors of all the blocks is calculated, and a value obtained by squaring the difference between the motion vector and the average vector is calculated. a motion vector variance value calculation unit for calculating a total of partial Chichinyuyoku the value,
Based on the correspondence between the dispersion value and the quantization parameter correction value , which is set in advance so that coarse quantization is performed in a portion where the motion is complex, and fine quantization is performed in a portion where the motion is smooth , A quantization parameter correction unit that calculates the quantization parameter correction value from a variance value and corrects a preset quantization parameter with the quantization parameter correction value;
Based on the quantization parameter corrected by the quantization parameter correction unit, a quantization unit that quantizes orthogonal transform coefficients of the difference image;
An encoding device comprising:   The motion vector variance value calculation unit divides the magnitude of the motion vector by the number of frame intervals between the current frame image and the reference frame image, and calculates the variance value based on the divided motion vector. The encoding apparatus according to claim 1. In order to perform motion compensation of a moving image composed of a plurality of frame images and to compress and code the moving image,
A motion prediction unit that predicts a motion vector for each block obtained by dividing the moving image from a reference frame image of the current frame image in the moving image;
A motion compensation unit that performs motion compensation based on the reference frame image and the motion vector, and generates a predicted frame image obtained by predicting the current frame image;
A difference image generation unit that generates a difference image that is a difference between the current frame image and the predicted frame image;
Based on the motion vector predicted by the motion prediction unit, an average vector obtained by averaging the motion vectors of all the blocks is calculated, and a value obtained by squaring the difference between the motion vector and the average vector is calculated. motion vector variance value calculation unit for calculating a total of partial Chichinyuyoku the value,
Based on the correspondence between the dispersion value and the quantization parameter correction value , which is set in advance so that coarse quantization is performed in a portion where the motion is complex, and fine quantization is performed in a portion where the motion is smooth , A quantization parameter correction unit that calculates the quantization parameter correction value from a variance value and corrects a preset quantization parameter with the quantization parameter correction value;
A quantization unit that quantizes orthogonal transform coefficients of the difference image based on the quantization parameter corrected by the quantization parameter correction unit;
Encoding program to function as.

Images