JP2024022374A - Encoding device, decoding device, and program - Google Patents

Abstract

An object of the present invention is to improve image quality and encoding efficiency while reducing the weight of calculation processing when encoding video at a required bit rate. An encoding device 1 calculates a first Lagrangian multiplier that minimizes a cost function and an allocated bit rate for each block using a bit rate allocated to a video frame and parameters of a first RD curve. a bit allocation unit 101; a quantization parameter calculation unit 102 that calculates a quantization parameter for each block using a second Lagrange multiplier calculated from the allocated bit rate for each block and the parameters of the second RD curve; and an RD optimization unit 103 that determines an encoding mode that minimizes the cost function based on the Lagrangian multiplier. The first RD curve is approximated using a logarithmic function or an exponential function. [Selection diagram] Figure 2

Description

Application for application of Article 30, Paragraph 2 of the Patent Act 1. Published by Yuichi Kondo, Yasuko Sugito, Atsuro Ichigaya, and others in the FIT2021 20th Information Science and Technology Forum lecture proceedings on August 12, 2021. 2. Published by Yuichi Kondo, Yasuko Sugito, Atsuro Ichigaya, and others at the 20th Information Science and Technology Forum of FIT2021 on August 27, 2021. 3. https://www. nhk. or. jp/strl/publica/giken_dayori/205/4. html Published by Yuichi Kondo in the Giken Newsletter April 2022 issue on April 1, 2022.

The present invention relates to an encoding device, a decoding device, and a program.

Encoding control is a control technology for realizing video encoding that suits the application and purpose. Encoding control includes rate control that performs processing to maximize encoding quality under bit rate constraint conditions and delay control that minimizes delay. In particular, rate control plays an important role in providing stable broadcasting and communication services with high image quality.

Rate control includes a bit allocation process _OBA (Optimal Bit Allocation) that allocates a bit rate in units of GOP (Group Of Picture), in units of frames, and in units of blocks so that encoding can be performed at the target bit rate Rt allocated to the frame; This is achieved by two processes: RD optimization processing RDO (Rate Distortion Optimization), which selects the encoding mode that maximizes the image quality (that is, minimizes encoding distortion) within the allocated bit rate range. Ru.

Here, an example will be described in which bit allocation for each frame has been completed and bit allocation is performed for each block within the frame. In the bit allocation process, bits are allocated so as not to exceed the target bit rate _Rt allocated to the frame while monitoring the consumed bit rate _Rc . FIG. 4 shows that seven blocks marked with diagonal lines have been encoded, and the remaining five blocks are in a state before encoding. One of the simplest bit allocation processing techniques is to derive the available bit rate R _a (R _a = R _t − R _c ) from the consumed bit rate R _c and spread R _a to the remaining blocks at a constant rate. This is the method of allocating. In the example shown in FIG. 4, the bits allocated per block for the remaining five blocks are R _a /5.

The purpose of the RD optimization process is to select an encoding mode that minimizes encoding distortion when encoding at a bit rate allocated by the bit allocation process. In recent years, reference software HM (HEVC Test Model) and VTM (VVC Test Model) for video coding standards such as HEVC (High Efficiency Video Coding) and VVC (Versatile Video Coding) have This is achieved by selecting the encoding mode so as to minimize the cost function J=D+λ _RDO R calculated from the Lagrangian multiplier λ _RDO (see, for example, Non-Patent Document 1). However, the Lagrangian multiplier λ _RDO is a value calculated using the RD curve of the encoded video and the allocated bit rate. The RD curve is a curve that approximates the relationship between the bit rate R and the encoding distortion D during video encoding, and in HM and VTM, the hyperbolic function shown in equation (1) is used. .

Here, c and k are parameters that differ depending on the video, and can be determined using the consumed bit rate R and encoding distortion D of the encoded block during the encoding process. When minimizing the cost function J, ∂J/∂R obtained by partially differentiating J with respect to R becomes 0, so λ _RDO can be determined by equation (2).

Furthermore, Non-Patent Document 2 discloses that bit allocation processing is realized by performing approximation using Taylor expansion, processing to reduce errors generated in approximation, and the like.

Sakae Okubo, "H265/HEVC Textbook", first edition, Impress Japan, published October 21, 2013 Li et al., “Optimal Bit Allocation for CTU Level Rate Control in HEVC,” IEEE Trans. Circuits Syst. Video Technol., 2017

In the bit allocation process described above, bits are allocated equally to each block, but by controlling the bits allocated to each block, it is possible to improve the image quality of the entire screen. The method is described below.

The purpose of the bit allocation process is to minimize the coding distortion D of the entire screen. However, the sum of the bit rates of all blocks must not exceed the bit rate _Rt assigned to that frame. If this is expressed numerically, it becomes equation (3). Here, the subscript i represents the block number, and d _i and r _i represent the encoding distortion and bit rate of the i-th block, respectively. Furthermore, M represents the number of blocks on the entire screen.

We would like to find r _i that satisfies equation (3). Equation (3) can be rewritten into Equation (4) using the Lagrangian multiplier λ _OBA .

The RD curve of the hyperbolic model is expressed by equation (5). c _i and k _i (1≦i≦M) are estimated values calculated from the blocks of the encoded frame using equations (1) and (2).

When minimizing the cost function J, the partial differential with respect to r _i is 0, so Equation (6) holds true.

であり、ｒ_ｉの総和がフレームに割り当てられたビットレートＲ_ｔとなることから、式（７）が成り立たなければならない。

Since the sum of r _i is the bit rate R _t assigned to the frame, equation (7) must hold.

When λ _OBA is found by solving Equation (7), the bit rate of the i-th block can be determined by r _i found from Equation (6). However, since c _i and k _i have different values for each block, equation (7) cannot be solved analytically.

In addition, in Non-Patent Document 2, bit allocation processing is realized by approximation using Taylor expansion to solve equation (7) and processing to reduce errors generated in the approximation, but the processing is complicated. There was a problem that it became

In view of the above circumstances, an object of the present invention is to provide an encoding device and a decoding device that can improve image quality and encoding efficiency while reducing calculation processing when encoding video at a required bit rate. , and to provide programs.

In order to solve the above problem, an encoding device according to an embodiment divides a video frame into encoding target regions, performs encoding for each block, and generates an RD curve that approximates the relationship between bit rate and encoding distortion. The encoding device performs RD optimization processing using a first Lagrangian multiplier that minimizes a cost function using a bit rate assigned to the video frame and a parameter of a first RD curve, and a bit allocation unit that calculates an allocated bit rate; and a quantization parameter that calculates a quantization parameter for each block using a second Lagrangian multiplier calculated from the allocated bit rate for each block and a parameter of a second RD curve. and an RD optimization unit that determines an encoding mode that minimizes the cost function based on the second Lagrangian multiplier, and the first RD curve is approximated using a logarithmic function or an exponential function. Ru.

Furthermore, in one embodiment, the bit allocation unit uses the bit rate allocated to the entire unencoded block in the video frame and the parameters of the first RD curve every time encoding of each block is completed. Then, the first Lagrangian multiplier and the allocated bit rate for each block may be calculated.

Furthermore, in one embodiment, the bit allocation unit allocates all unencoded blocks in the video frame when a difference between the bit rate of the encoded block and the expected bit rate exceeds a threshold. The first Lagrangian multiplier and the allocated bit rate for each block may be calculated using the determined bit rate and the parameters of the first RD curve.

Further, in order to solve the above problem, a decoding device according to an embodiment obtains the quantization parameter calculated by the quantization parameter calculation unit of the encoding device, and uses the quantization parameter to The data encoded by the encoding device is decoded.

Further, a program according to an embodiment causes a computer to function as the encoding device.

Further, a program according to an embodiment causes a computer to function as the decoding device.

According to the present invention, when encoding a video at a required bit rate, it is possible to improve image quality and encoding efficiency while reducing computational processing.

FIG. 1 is a block diagram illustrating a configuration example of an encoding device according to an embodiment. FIG. 2 is a block diagram illustrating a configuration example of a rate control unit in an encoding device according to an embodiment. FIG. 1 is a block diagram illustrating a configuration example of a decoding device according to an embodiment. FIG. 3 is a diagram illustrating an example of conventional bit allocation for each block.

Hereinafter, one embodiment will be described in detail with reference to the drawings.

(encoding device)
FIG. 1 shows a configuration example of an encoding device according to an embodiment of the present invention. The encoding device 1 shown in FIG. , an addition section 17, a storage section 18, a prediction section 19, an entropy encoding section 20, and an encoded data storage/output section 21.

The encoding device 1 divides a video frame into encoding target regions, performs encoding for each encoding block (hereinafter simply referred to as "block"), and approximates the relationship between bit rate R and encoding distortion D. RD optimization processing is performed using the RD curve. At this time, quantization parameters are determined for each block and encoding is performed so that the bit rate is below a certain level.

The block dividing unit 11 generates a block image by dividing the video frame into encoding target areas in units of blocks, and outputs the block image to the subtracting unit 12 and the predicting unit 19.

The subtraction unit 12 subtracts each pixel value of the predicted block image input from the prediction unit 19 (described later) from each pixel value of the block image input from the block division unit 11, and calculates the difference between the block image and the predicted block image. The residual block image shown in FIG.

The transformation unit 13 performs transformation processing such as orthogonal transformation on the residual block image input from the subtraction unit 12 to calculate transformation coefficients, and outputs them to the quantization unit 14 .

The rate control unit 10 performs rate control by appropriately determining a quantization parameter when quantizing a residual block image. Details of the processing by the rate control unit 10 will be described later. The rate control unit 10 outputs the determined quantization parameter to the quantization unit 14 and the entropy encoding unit 20.

The quantization unit 14 converts the transformation coefficients input from the conversion unit 13 into quantization steps corresponding to the quantization parameters input from the rate control unit 10 (for example, the quantization parameters are arranged so that the logarithms of the quantization steps are proportional to each other). ) and quantization to generate a quantization coefficient, which is output to the inverse quantization section 15 and the entropy encoding section 20. The quantization unit 14 reduces the amount of data.

The inverse quantization section 15 multiplies the quantization coefficients input from the quantization section 14 by the quantization step to restore transform coefficients, and outputs the restored transform coefficients to the inverse transform section 16 .

The inverse transformer 16 performs an inverse transform process (processing to undo the transformation performed by the transformer 13) on the transform coefficients input from the inverse quantizer 15 to restore a residual block image, and the adder 16 restores the residual block image. Output to 17. For example, when the transform unit 13 performs a discrete cosine transform, the inverse transform unit 16 performs an inverse discrete cosine transform.

The addition unit 17 adds the residual block image input from the inverse transformation unit 16 and the predicted image input from the prediction unit 19, and outputs the result to the storage unit 18 as an encoded image.

The inverse quantization section 15, the inverse transformation section 16, and the addition section 17 constitute a coded image generation section (local decoding section). That is, the encoded image generation unit restores the transform coefficient by multiplying the quantization coefficient by the quantization step, performs inverse transform processing on the transform coefficient to restore the residual block image, and restores the residual block image by multiplying the quantization coefficient by the quantization step. A coded image is generated by adding the block image and the intra-predicted image or the motion-compensated predicted image.

The encoding device 1 may perform post-processing such as filter processing using a deblocking filter on the encoded image output by the addition unit 17 and then output the encoded image to the storage unit 18 .

The storage unit 18 stores the encoded image input from the addition unit 17.

The prediction unit 19 performs intra prediction (intra-screen prediction) or inter prediction (inter-screen prediction, motion compensation prediction). In intra prediction, an intra prediction image is generated by performing intra prediction on the encoded image stored in the storage unit 18 according to an intra prediction mode. In inter prediction, a motion-compensated predicted image is generated by performing motion-compensated prediction on the encoded image stored in the storage unit 18 according to a motion vector. The prediction unit 19 switches between the intra-predicted image and the motion-compensated predicted image to form a predicted block image, and outputs it to the subtraction unit 12 and addition unit 17 . The prediction unit 19 outputs prediction parameters (intra prediction mode and motion vector information) used in the prediction process to the entropy encoding unit 20.

The entropy encoding unit 20 performs entropy encoding on the quantization coefficient input from the quantization unit 14, the quantization parameter input from the rate control unit 10, and the prediction parameter input from the prediction unit 19, and performs data compression. to generate encoded data and output it to the encoded data storage/output section 21. For entropy encoding, any entropy encoding method such as zero-order exponential Golomb code or context-based adaptive binary arithmetic coding (CABAC) can be used.

The encoded data storage/output unit 21 outputs encoded data encoded using the optimal encoding mode determined by the rate control unit 10 to the outside of the encoding device 1.

(rate control section)
Next, the processing of the rate control unit 10 will be explained. FIG. 2 shows an example of the configuration of the rate control section 10. The rate control section 10 shown in FIG. 2 includes a bit allocation section 101, a quantization parameter calculation section 102, and an RD optimization section 103. Hereinafter, the Lagrange multiplier λ _OBA will be referred to as a first Lagrange multiplier, and the Lagrange multiplier λ _RDO will be referred to as a second Lagrange multiplier. Further, the RD curve used by the bit allocation section 101 is referred to as a first RD curve, and the RD curve used by the quantization parameter calculation section 102 is referred to as a second RD curve.

The bit allocation unit 101 performs a process of allocating bits to blocks constituting a frame so as not to exceed the bits _Rt allocated to the frame. Let r i be the bit rate assigned to the i-th block (hereinafter referred to as "block rate"), and let _{d i be} the encoding distortion of the i-th block (hereinafter referred to as "block distortion"), and Let M be the number of blocks. To maximize the image quality of the entire frame (minimize coding distortion), use the second Lagrangian multiplier λ _RDO to select a prediction mode that minimizes the cost function J so as to satisfy the above equation (4). All you have to do is choose.

As described above, when the conventional RD curve of the hyperbolic model expressed by equation (5) is used, the block rate r _i cannot be solved analytically, or the processing becomes complicated. Therefore, in the present invention, a logarithmic function or an exponential function is used. By approximating the first RD curve using a logarithmic function or an exponential function of the block rate r _i , equation (4) can be easily solved, and the block rate r _i can be solved analytically.

When the first RD curve is approximated using a logarithmic function, the relationship between block distortion d _i and block rate r _i is expressed by equation (8).

When approximating the first RD curve using an exponential function, the relationship between block distortion d _i and block rate r _i is expressed by equation (9).

Below, a case will be described in which the first RD curve is approximated using a logarithmic function. c′ _i and k′ _i (1≦i≦M) are calculated from the blocks of the encoded frame using equations (10) and (11).

When minimizing the cost function J, the partial differential with respect to the block rate r _i becomes 0, so Equation (12) holds true.

なので、式（１３）が成立する。すると、第１ラグランジュ乗数λ_ＯＢＡは式(１４)により簡単に求めることができて、ブロックレートｒ_ｉは式(１５)により求まる。

Therefore, equation (13) holds true. Then, the first Lagrangian multiplier λ _OBA can be easily obtained using equation (14), and the block rate r _i can be obtained using equation (15).

In this way, the bit allocation unit 101 uses the bit rate _Rt allocated to the video frame and the parameters of the first RD curve calculated from the encoded block to calculate the first Lagrangian that minimizes the cost function J. The multiplier λ _OBA (see equation (14)) and the block rate r _i (see equation (15)) are calculated. Then, the calculated block rate r _i is output to the quantization parameter calculation unit 102. By using a logarithmic model or an exponential model instead of a conventional hyperbolic model, the bit allocation unit 101 easily calculates the block rate _r i from equation (15) using c′ _i obtained from the encoded result. can be determined and solved analytically.

When using the conventional RD curve as the second RD curve, the quantization parameter calculation unit 102 calculates the second Lagrangian multiplier λ _RDO by substituting the block rate r _i into equation (6). Furthermore, when using the same RD curve as the first RD curve as the second RD curve, the second Lagrangian multiplier λ _RDO is calculated by substituting the block rate r _i into equation (12).

Then, the quantization parameter calculation unit 102 calculates the quantization parameter QP _i when quantizing the residual signal for each block, based on the second Lagrangian multiplier λ _RDO , as shown in equation (16). . Then, the quantization parameter calculation unit 102 outputs the calculated quantization parameter QP _i to the quantization unit 14 of the encoding device 1. Further, the quantization parameter calculation unit 102 outputs the second Lagrangian multiplier λ _RDO and the quantization parameter QP _i to the RD optimization unit 103.

The RD optimization unit 103 uses the second Lagrangian multiplier λ _RDO and the quantization parameter QP _i input from the quantization parameter calculation unit 102 to determine the optimal encoding mode that minimizes the cost function J. The optimal encoding mode is a combination of encoding tools and parameters (such as DC prediction mode for intra prediction). Then, the RD optimization section 103 outputs the determined optimal encoding mode to the encoded data storage/output section 21 of the encoding device 1. Note that r _i used in RD optimization is different from the block rate r _i calculated by bit allocation section 101. In RD optimization, encoding is performed in all modes, and the combination of d _i and r _i obtained for each mode is assigned to the cost function J=d _i +λ _RDO・r _i to find the mode in which J is the minimum. select.

(First modification of bit allocation processing)
Although the block rate r _i and the quantization parameter QP _i are uniquely determined by the bit allocation section 101 and the quantization parameter calculation section 102, the expected bit rate is not necessarily obtained when actually encoding. Therefore, the amount of allocated bits may be modified on a block-by-block basis so that the actual bit rate _Rt allocated to the frame does not deviate significantly. In other words, in the first modification, the bit allocation unit 101 calculates the bit rate R _j allocated to all unencoded blocks in a frame and the parameters of the first RD curve every time encoding of each block is completed. is used to calculate the first Lagrangian multiplier λ _OBA and the block rate r _i .

Specifically, if the bits consumed in the i-th block are _bj , then when the encoding of blocks up to the j-th block is completed, the bit rate _Rj assigned to all unencoded blocks in the frame is given by the formula It is expressed as (17). Therefore, the bit allocation unit 101 corrects the block rate r _i (j<i≦M) using equation (18).

(Second modification example of bit allocation processing)
If the block rate r _i is controlled on a block-by-block basis according to the first modification, the image quality of the entire frame may deteriorate. Therefore, the image quality of the entire frame may be taken into consideration, and corrections may be made according to the first modified example as necessary. In other words, in the second modification, when the difference between the bit rate R _actual of the encoded block and the assumed bit rate R _ideal exceeds the threshold, the bit allocation unit 101 assigns A first Lagrangian multiplier λ _OBA and a block rate r _i are calculated using the bit rate R _j assigned to the entire block and the parameters of the first RD curve.

Specifically, control is performed based on how far the bit rate R _actual actually consumed up to the j-th time deviates from the expected bit rate R _ideal . The bit allocation unit 101 calculates the expected value R _ideal of the bit rate consumed up to the jth bit rate from Equation (19), and calculates the degree of deviation S by (R _actual - R _ideal ) as shown in Equation (20). Calculated by the absolute value of /R _ideal . The bit allocation unit 101 compares the deviation degree S and the threshold value s for each block encoding, and corrects the block rate r _i shown in equation (18) only for blocks where S>s, thereby eliminating unnecessary control. You can prevent it from happening. The threshold value s can be set arbitrarily, for example, s=0.1.

(Decoding device)
Next, a decoding device according to an embodiment of the present invention will be described. FIG. 3 shows a configuration example of a decoding device according to an embodiment of the present invention. The decoding device 2 shown in FIG. 3 includes an entropy decoding section 31, an inverse quantization section 32, an inverse transformation section 33, an addition section 34, a storage section 35, and a prediction section 36.

The decoding device 2 acquires the quantization parameter QP _i calculated by the quantization parameter calculation unit 102 from the encoding device 1, and uses the quantization parameter QP _i to generate the code encoded by the encoding device 1. decrypt the encoded data.

The entropy decoding unit 31 decodes the encoded data output by the encoding device 1 and obtains the quantization parameter QP _i , the quantization coefficient, and the prediction parameter (intra prediction mode and motion vector information). Then, the entropy decoding unit 31 outputs the quantization parameter QP _i and the quantization coefficient to the inverse quantization unit 32, and outputs the prediction parameter to the prediction unit 36.

The inverse quantization unit 32 inputs the quantization coefficient and the quantization parameter QP _i from the entropy decoding unit 31, multiplies the quantization coefficient by the quantization step derived from the quantization parameter QP _i , and performs orthogonal transformation for each block. The coefficients are restored and output to the inverse transform section 33.

The inverse transformer 33 performs inverse transform on the orthogonal transform coefficients input from the inverse quantizer 32 to generate a residual image, and outputs it to the adder 34 .

The addition unit 34 adds each pixel value of the residual image input from the inverse transformation unit 33 and the predicted image input from the prediction unit 36 to generate a decoded image, and outputs the decoded image to the storage unit 35 and the outside of the decoding device 2. do.

Similar to the encoding device 1, the decoding device 2 may perform post-processing such as filter processing using a deblocking filter on the decoded image output by the addition section 34, and then output the image to the storage section 35.

The storage unit 35 stores the decoded image input from the addition unit 34.

The prediction unit 36 performs intra prediction (intra-screen prediction) or inter prediction (inter-screen prediction, motion compensation prediction). In the intra prediction, an intra prediction image is generated by performing intra prediction on the decoded image stored in the storage unit 35 according to the intra prediction mode input from the entropy decoding unit 31. In inter prediction, a motion-compensated predicted image is generated by performing motion-compensated prediction on the decoded image stored in the storage unit 35 according to motion vector information input from the entropy decoding unit 31. The prediction unit 36 switches between the intra predicted image and the motion compensated predicted image to form a predicted block image, and outputs the predicted block image to the addition unit 34 .

As described above, the present invention can encode video at a required bit rate, reducing computational processing and improving image quality and encoding efficiency, without changing the encoding/decoding method from conventional methods such as HEVC and VVC. This makes it possible to improve image quality and encoding efficiency when converting images.

(program)
In order to function as the above-mentioned encoding device 1 and decoding device 2, it is also possible to use a computer that can execute program instructions, respectively. Here, the computer may be a general-purpose computer, a dedicated computer, a workstation, a PC (Personal Computer), an electronic notepad, or the like. Program instructions may be program code, code segments, etc. to perform necessary tasks.

The computer includes a processor, a storage section, an input section, an output section, and a communication interface. Processors include CPUs (Central Processing Units), MPUs (Micro Processing Units), GPUs (Graphics Processing Units), DSPs (Digital Signal Processors), and SoCs (System on a Chip). may be configured. The processor controls each of the above components and performs various calculation processes by reading and executing programs from the storage unit. Note that at least a part of these processing contents may be realized by hardware. The input unit is an input interface that receives a user's input operation and obtains information based on the user's operation, and is a pointing device, keyboard, mouse, or the like. The output unit is an output interface that outputs information, such as a display or a speaker. The communication interface is an interface for communicating with an external device, and is, for example, a LAN (Local Area Network) interface.

The program may be recorded on a computer readable recording medium. Using such a recording medium, it is possible to install a program on a computer. Here, the recording medium on which the program is recorded may be a non-transitory recording medium. The non-transitory recording medium is not particularly limited, and may be, for example, a CD-ROM, a DVD-ROM, a USB (Universal Serial Bus) memory, or the like. Further, this program may be downloaded from an external device via a network.

For example, a program for making a computer function as the above-mentioned encoding device 1 includes a first Lagrangian multiplier and a block rate that minimize the cost function using the bit rate assigned to the video frame and the parameters of the first RD curve. calculating r _i ; determining a coding mode that minimizes the cost function using a second Lagrange multiplier calculated from the block rate r _i and the parameters of the second RD curve; and determining a second Lagrange multiplier. The first RD curve is approximated using a _logarithmic function or an exponential function.

Further, the program for causing a computer to function as the decoding device 2 described above includes the steps of acquiring the quantization parameter QP i calculated by the quantization parameter calculation unit 102 from the encoding device 1, and the step of acquiring the quantization parameter QP _i calculated by the quantization parameter calculation unit ₁₀₂ The computer is caused to perform the step of decoding the data encoded by the encoding device 1 using the encoder.

Although the embodiments described above have been described as representative examples, it will be apparent to those skilled in the art that many modifications and substitutions can be made within the spirit and scope of the invention. Therefore, the present invention should not be construed as being limited to the above-described embodiments, and various modifications and changes can be made without departing from the scope of the claims. For example, it is possible to integrate a plurality of configuration blocks described in the configuration diagram of the embodiment, or to divide one configuration block.

1 Encoding device 2 Decoding device 10 Rate control unit 11 Block division unit 12 Subtraction unit 13 Transformation unit 14 Quantization unit 15 Inverse quantization unit 16 Inverse transformation unit 17 Addition unit 18 Storage unit 19 Prediction unit 20 Entropy encoding unit 21 Code data storage/output unit 31 entropy decoding unit 32 inverse quantization unit 33 inverse transformation unit 34 addition unit 35 storage unit 36 prediction unit 101 bit allocation unit 102 quantization parameter calculation unit 103 RD optimization unit

Claims (6)

An encoding device that divides a video frame into encoding target regions, encodes each block, and performs RD optimization processing using an RD curve that approximates the relationship between bit rate and encoding distortion,
a bit allocation unit that calculates a first Lagrangian multiplier that minimizes a cost function and an allocated bit rate for each block using the bit rate allocated to the video frame and the parameters of the first RD curve;
a quantization parameter calculation unit that calculates a quantization parameter for each block using a second Lagrangian multiplier calculated from the allocated bit rate for each block and a parameter of a second RD curve;
an RD optimization unit that determines an encoding mode that minimizes the cost function based on the second Lagrangian multiplier,
The encoding device, wherein the first RD curve is approximated using a logarithmic function or an exponential function. The bit allocation unit calculates the first Lagrangian multiplier using the bit rate allocated to the entire uncoded block in the video frame and the parameters of the first RD curve each time encoding of each block is completed. The encoding device according to claim 1, wherein the encoding device calculates an allocated bit rate for each block. When the difference between the bit rate of the encoded block and the expected bit rate exceeds a threshold, the bit allocation unit divides the bit rate allocated to the entire unencoded blocks in the video frame and the first RD. The encoding device according to claim 1, wherein the first Lagrangian multiplier and the allocated bit rate for each block are calculated using curve parameters. Obtaining the quantization parameter calculated by the quantization parameter calculation unit of the encoding device according to any one of claims 1 to 3,
A decoding device that decodes data encoded by the encoding device using the quantization parameter. A program for causing a computer to function as the encoding device according to claim 1. A program for causing a computer to function as the decoding device according to claim 4.

Images