JP6557589B2 - Encoding device, decoding device, and program - Google Patents

Description

  The present invention relates to an encoding device that encodes an image, a decoding device that decodes encoded data, and a program thereof.

  Conventionally, in order to compress a data amount when transmitting or storing an image (a still image or a moving image) or sound, a coding method has been studied. In recent years, in video coding technology, super high-resolution video represented by 8K-SHV has been widely used. As a technique for transmitting a large amount of data over a broadcast wave or IP network, ITU- TH. H.264 / MPEG-4 AVC and ITU-T H.264. Coding schemes such as H.265 / MPEG-H HEVC are known. In this high compression coding method, high efficiency is realized by combining elemental technologies such as block division, orthogonal transform, quantization, entropy coding, intra prediction, and motion compensation (see, for example, Non-Patent Document 1). .

  In the conventional coding method described above, weighted quantization is performed on the transform coefficient after orthogonal transform processing, and coarse quantization is applied to high-frequency components that are not visually noticeable, and low-frequency signals that are important on the signal components. By applying fine quantization to the components, compression processing considering visual characteristics is performed. In general, many coefficients on the high frequency component side show small values and become almost zero after quantization, so that the amount of data after quantization can be reduced.

Supervised by Satoshi Okubo, “Impress Standard Textbook Series H.265 / HEVC Textbook”, Impress Japan Co., Ltd., October 21, 2013

  In the conventional encoding method, data compression can be performed by quantization, but non-zero numerical data on the low frequency component side cannot be compressed and there is a limit to efficiency. In addition, if the continuous non-zero numerical data is a large count value, the amount of data increases accordingly. Furthermore, if a rough quantization is performed such that zero is continued on the high frequency component side in order to achieve a high compression rate, there is a problem that subjective image quality is significantly impaired.

  An object of the present invention made in view of such circumstances is to provide an encoding device, a decoding device, and a program capable of improving encoding efficiency as compared with a conventional encoding method.

  In order to solve the above-described problem, an encoding apparatus according to the present invention is an encoding apparatus that performs encoding using a fitting function, and a block dividing unit that generates a block image by dividing an input image into a plurality of blocks; A prediction unit that generates a prediction block image obtained by predicting the block image, and a conversion process is performed on a residual block image indicating a difference between the block image and the prediction block image, thereby calculating a conversion coefficient for each block. A curve fitting unit that generates a fitting function information including a coefficient of a fitting function for approximating a one-dimensional array of transform coefficients for each block and defining an approximated curve, and the fitting function information; Generates bitstream by entropy coding of prediction information necessary for prediction of prediction block image An inverse curve fitting unit that obtains the approximate curve based on the fitting function information, arranges data defined by the approximate curve in two dimensions, and restores the transform coefficient for each block; An inverse transform unit that performs inverse transform on the restored transform coefficient and restores the residual block image, and the prediction unit adds the restored residual block image and the predicted block image The prediction block image is generated with reference to the decoded image.

  Further, in the encoding device according to the present invention, the curve fitting unit scans the transform coefficient in a predetermined direction for each block to generate one-dimensional array data that is one-dimensionally arranged, and And a fitting unit that approximates a curve using one or more fitting functions for the one-dimensional array data.

  Furthermore, in the encoding device according to the present invention, the curve fitting unit scans the transform coefficient in a plurality of directions for each block and generates a plurality of one-dimensional array data that is one-dimensionally arrayed. A generating unit; and a fitting unit that approximates a curve using one or more fitting functions for each of the plurality of one-dimensional array data, and the entropy encoding unit includes the fitting function for each scan direction. Information is entropy-encoded, an optimum scan direction based on a cost value is determined, a bit stream generated in the scan direction, and identification information for identifying the scan direction are output.

  Furthermore, in the encoding apparatus according to the present invention, a quantization unit that divides the transform coefficient for each block by a quantization step to generate a quantization coefficient, and the block obtained by multiplying the quantization coefficient by the quantization step An inverse quantization unit that restores transform coefficients for each of the transform coefficients, wherein the inverse transform unit performs inverse transform on the transform coefficients restored by the inverse quantization unit or the inverse curve fitting unit, and The difference block image is restored, and the entropy coding unit performs entropy coding of the fitting function information and the quantization coefficient, and outputs a bitstream having better coding efficiency based on a cost value. Features.

  Furthermore, the encoding apparatus according to the present invention further includes a quantization unit that divides the transform coefficient for each block by a quantization step, and the curve fitting unit is a one-dimensional transform coefficient divided by the quantization unit. The array is approximated with a curve using one or more fitting functions, and fitting function information including coefficients of the fitting function is generated. The inverse curve fitting unit obtains the approximate curve based on the fitting function information, and the approximation The data defined by the curve is two-dimensionally arranged to restore the transform coefficient divided by the quantization unit, and by multiplying the transform coefficient generated by the inverse curve fitting unit by a quantization step, for each block An inverse quantization unit that restores a transform coefficient is further provided, and the inverse transform unit is provided by the inverse quantization unit. Perform inverse transform on the restored transform coefficients, and wherein the restoring the residual block image.

  In order to solve the above problems, a decoding device according to the present invention is a decoding device that decodes a bitstream encoded using a fitting function, and decodes the bitstream to define an approximate curve. Data obtained by obtaining an approximation curve based on the fitting function information, an entropy decoding unit that obtains fitting function information including a coefficient of the fitting function and prediction information necessary for block image prediction, and the fitting function information An inverse curve fitting unit that restores a transform coefficient for each block, an inverse transform unit that performs inverse transform on the transform coefficient to restore a residual block image, and the block image according to the prediction information A prediction unit that generates a prediction block image predicting the image, the residual block image, and the prediction block Tsu includes an adder which generates a decoded block image by adding click image, and the prediction unit, and generates the prediction block image with reference to the decoded the decoded block image.

  Further, in the decoding device according to the present invention, the entropy decoding unit further includes an inverse quantization unit that further acquires a quantization coefficient and restores a transform coefficient for each block by multiplying the quantization coefficient by a quantization step. The inverse transform unit performs an inverse transform on the transform coefficient restored by the inverse quantization unit or the inverse curve fitting unit to restore the residual block image.

  Furthermore, the decoding apparatus according to the present invention further includes an inverse quantization unit that restores the second transform coefficient for each block by multiplying the first transform coefficient restored by the inverse curve fitting unit by a quantization step. The inverse transform unit restores the residual block image by performing inverse transform on the second transform coefficient.

  In order to solve the above problems, a program according to the present invention causes a computer to function as the encoding device.

  In order to solve the above problem, a program according to the present invention causes a computer to function as the decoding device.

  According to the present invention, it is possible to improve the encoding efficiency as compared with the conventional encoding method.

It is a block diagram which shows the structural example of the encoding apparatus which concerns on the 1st Embodiment of this invention. It is a block diagram which shows the 1st structural example of the curve fitting part in the encoding apparatus which concerns on the 1st Embodiment of this invention. It is a figure which shows the example of the scanning direction by the curve fitting part in the encoding apparatus which concerns on the 1st Embodiment of this invention. It is a figure which shows an example of the curve fitting by the curve fitting part in the encoding apparatus which concerns on the 1st Embodiment of this invention. It is a block diagram which shows the 2nd structural example of the curve fitting part in the encoding apparatus which concerns on the 1st Embodiment of this invention. It is a block diagram which shows the structural example of the encoding apparatus which concerns on the 2nd Embodiment of this invention. It is a block diagram which shows the structural example of the encoding apparatus which concerns on the 3rd Embodiment of this invention. It is a block diagram which shows the structural example of the decoding apparatus which concerns on the 4th Embodiment of this invention. It is a block diagram which shows the structural example of the reverse curve fitting part in the decoding apparatus which concerns on the 4th Embodiment of this invention. It is a block diagram which shows the structural example of the decoding apparatus which concerns on the 5th Embodiment of this invention. It is a block diagram which shows the structural example of the decoding apparatus which concerns on the 6th Embodiment of this invention.

  Hereinafter, after describing an encoding device, an embodiment of the present invention will be described.

(First embodiment)
An encoding apparatus according to the first embodiment of the present invention will be described below. FIG. 1 shows a configuration example of an encoding apparatus according to the first embodiment of the present invention. 1 includes a block dividing unit 11, a subtracting unit 12, a converting unit 13, a curve fitting unit 14, an inverse curve fitting unit 15, an inverse converting unit 16, an adding unit 17, A storage unit 18, a prediction unit 19, and an entropy encoding unit 20 are provided.

  The block division unit 11 divides the input image into a plurality of blocks, and outputs the block image to the subtraction unit 12 and the prediction unit 19. The block size may be variable, for example, 32 × 32 pixels, 16 × 16 pixels, 8 × 8 pixels, or 4 × 4 pixels.

  The subtraction unit 12 generates a residual block image obtained by subtracting each pixel value of the prediction 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 a conversion unit 13 is output.

  The transform unit 13 performs transform processing such as orthogonal transform on the residual block image input from the subtraction unit 12 to calculate transform coefficients, and outputs the transform coefficients for each block to the curve fitting unit 14. Note that the conversion unit 13 may perform conversion other than orthogonal conversion.

  The curve fitting unit 14 approximates a one-dimensional array of transform coefficients for each block input from the transform unit 13 using one or more fitting functions. Then, fitting function information which is information including one or more fitting function coefficients for defining the approximate curve is output to the inverse curve fitting unit 15 and the entropy coding unit 20.

  FIG. 2 shows a first configuration example of the curve fitting unit 14. The curve fitting unit 14 illustrated in FIG. 2 includes a one-dimensional array data generation unit 141 and a fitting unit 142.

  The one-dimensional array data generation unit 141 generates one-dimensional array data obtained by scanning the conversion coefficient input from the conversion unit 13 in a predetermined direction for each block to form a one-dimensional array, and outputs the generated one-dimensional array data to the fitting unit 142. If the scan reading order is, for example, one-dimensional array data that is arranged in order from visually important components, the fitting can be concentrated in that range. It becomes possible. When the size of the block image is 8 × 8 pixels, 64 transform coefficients are input from the one-dimensional array data generation unit 141 to the fitting unit 142 in the scan order.

  When the transform unit 13 performs an orthogonal transform such as a discrete cosine transform, a natural image generally has a larger value for a low frequency component and a smaller value for a high frequency component. Therefore, the number of fitting functions can be reduced if scanning is performed in such an order that the transform coefficient gradually changes in the forward direction or the reverse direction from the low frequency component to the high frequency component. Such a scanning method is described in ITU-T H.264. Zigzag scanning used in H.264 / MPEG-4 AVC and ITU-T H.264. Any scanning method can be used, such as the diagonal scan used in H.265 / MPEG-H HEVC.

  FIG. 3 shows an example of the scanning direction by the one-dimensional array data generation unit 141 when the size of the block image is 8 × 8 pixels. 3A and 3B show the zigzag scan direction, FIGS. 3C and 3D show the oblique scan direction, and FIGS. 3E and 3F are 4 × 4. The scan direction of the oblique scan for each pixel sub-block is shown.

  The fitting unit 142 generates an approximate curve that approximates the conversion coefficient by performing curve fitting on the one-dimensional array data input from the one-dimensional array data generating unit 141 using one or more fitting functions. Then, the fitting function information is output to the entropy encoding unit 20. The curve fitting can be performed by, for example, the least square method.

  When a plurality of fitting functions are used for curve fitting, the type of fitting function is also included in the fitting function information. Further, the number of fitting functions may be included in the fitting function information. In addition, information on which fitting function is used for curve fitting (function type, number of functions) is shared between the encoding device and the decoding device in advance, and the fitting function information is used only as coefficients of the fitting function. The compression efficiency may be increased. The smaller the number of fitting functions, the more severe the coding deterioration, but the higher the compression, and the larger the number of fitting functions, the more the amount of data increases.

  The types of fitting functions are polynomial, exponential function, logarithmic function, fractional function, trigonometric function, sinc function, conic curve (ellipse, parabola, hyperbola), sigmoid function, Bessel function, Poisson distribution, Gaussian distribution, Laplace distribution, The probability density (mass) function of probability distributions such as Lorentz distribution and Landau distribution. The number of fitting functions used for curve fitting may be one or plural.

  For example, when the fitting function is a probability density function having a normal distribution, the fitting function is expressed by Expression (1), and the coefficients of the fitting function are a size A, an average value μ, and a standard deviation σ. By specifying the coefficient of the fitting function, the shape of the fitting function is determined.

  FIG. 4 shows an example of curve fitting. Here, curve fitting is performed using an 8 × 8 pixel image block using one Lorentz distribution probability density function (Lorentz function) and two Gaussian distribution probability density functions (Gauss function). The horizontal axis represents the scan order of 64 transform coefficients by the one-dimensional array data generation unit 141, and the vertical axis represents the transform coefficients. Although the conversion coefficient is a negative value, only a positive value is used here for easy understanding.

  In curve fitting, fitting according to the purpose is possible. For example, when considering the case where discrete cosine transform is performed in the previous stage, it is important in terms of visual characteristics when performing a scan in which the scan element can be separated from the low frequency component side to the high frequency component side like zigzag scanning. It is possible to perform fitting so as not to change the data value of the DC component, to concentrate the fitting on the low frequency component side, or to distribute the fitting uniformly from the low frequency component to the high frequency component.

  The inverse curve fitting unit 15 obtains an approximate curve based on the fitting function information input from the curve fitting unit 14. Then, the data defined by the approximate curve is scanned in the opposite direction to that of the one-dimensional array data generation unit 141 and arranged two-dimensionally to restore the transform coefficients for each block and output to the inverse transform unit 16.

  The inverse transform unit 16 performs the inverse transform of the transform performed by the transform unit 13 on the transform coefficient input from the inverse curve fitting unit 15 to restore the residual block image, and outputs the residual block image to the adder unit 17. For example, when the transform unit 13 performs discrete cosine transform, the inverse transform unit 16 performs inverse discrete cosine transform.

  The addition unit 17 adds the residual block image input from the inverse conversion unit 16 and each pixel value of the prediction block image input from the prediction unit 19 and outputs the result to the storage unit 18 as a decoded block image. To do.

  The storage unit 18 is a memory that stores the decoded block image input from the addition unit 17.

  The prediction unit 19 refers to the decoded decoded image stored in the storage unit 18, predicts the pixel value in the target block, generates a predicted block image, and outputs the prediction block image to the subtraction unit 12 and the addition unit 17. In addition, prediction information (motion vector or intra prediction mode) that is information necessary for generating a prediction block image is output to the entropy encoding unit 20.

  The entropy encoding unit 20 performs entropy encoding on the fitting function information input from the curve fitting unit 14 and the prediction information input from the prediction unit 19, performs data compression, and generates a bitstream. Output to the outside of the encoding device 1. For entropy coding, any entropy coding scheme such as 0th-order exponent Golomb code or CABC (Context-Adaptive Binary Arithmetic Coding) can be used.

(Modification of curve fitting)
Next, a modified example of curve fitting will be described. FIG. 5 shows a second configuration example of the curve fitting unit 14. The curve fitting unit 14 illustrated in FIG. 5 includes a one-dimensional array data generation unit 141 and a fitting unit 142.

  In the modification, the one-dimensional array data generation unit 141 scans the conversion coefficient input from the conversion unit 13 in a plurality of directions for each block, and generates a plurality of one-dimensional array data in a one-dimensional array. The data is output to the fitting unit 142. In addition, the one-dimensional array data generation unit 141 outputs scan direction identification information, which is information for identifying the scanned direction, to the entropy encoding unit 20.

  The fitting unit 142 performs a curve fitting on each of the plurality of one-dimensional array data input from the one-dimensional array data generation unit 141 using one or more fitting functions, thereby approximating a plurality of conversion coefficients. Generate an approximate curve. Then, the fitting function information is output to the inverse curve fitting unit 15 and the entropy coding unit 20.

  In this case, the entropy encoding unit 20 inputs fitting function information and prediction information for each direction of scanning by the one-dimensional array data generation unit 141. Then, entropy coding is performed for each fitting direction on the fitting function information and the prediction information to generate a bit stream, and the respective cost values are calculated. Here, the cost value is a value determined by a cost function including the encoded data information amount and the distortion amount, and the encoding efficiency can be evaluated based on the cost value. If a cost function that can be considered to have better coding efficiency as the cost value is smaller is applied, optimization processing based on rate and distortion can be realized by minimizing the cost value. The entropy encoding unit 20 determines an optimal scan direction based on the cost value, and outputs a bitstream generated in the scan direction and identification information for identifying the scan direction to the outside of the encoding device 1. .

  As described above, the encoding apparatus 1 uses the curve fitting unit 14 to approximate a one-dimensional array of transform coefficients for each block using one or more fitting functions, and outputs fitting function information. Therefore, the encoding device 1 can reduce the amount of data to be encoded as compared with the conventional case, and can improve the encoding efficiency. In addition, it is possible to reduce the influence of the size and number of non-zero numerical data on the low frequency component side, and the coefficient value data on the high frequency component side can also be compressed by using a fitting function as a baseline. Can be left continuously. As a result, it is possible to improve the encoding deterioration due to the conventional quantization process in which zero values are continued.

  Also, the encoding device 1 generates a plurality of fitting function information by scanning the transform coefficient in a plurality of directions by the curve fitting unit 14, and the entropy of the fitting function information for each scanning direction by the entropy encoding unit 20. Encoding may be performed, and optimization processing based on the cost value may be performed. By performing the optimization process for each block in this way, it is possible to further improve the encoding deterioration.

(Second Embodiment)
Next, an encoding apparatus according to the second embodiment of the present invention will be described. FIG. 6 shows a configuration example of the encoding apparatus according to the second embodiment. The encoding device 2 shown in FIG. 6 includes a block dividing unit 11, a subtracting unit 12, a converting unit 13, a curve fitting unit 14, an inverse curve fitting unit 15, an inverse converting unit 16, an adding unit 17, A storage unit 18, a prediction unit 19, an entropy encoding unit 20, a quantization unit 21, an inverse quantization unit 22, and a selection unit 23 are provided. The encoding device 2 of the second embodiment is different from the encoding device 1 of the first embodiment in that it further includes a quantization unit 21, an inverse quantization unit 22, and a selection unit 23. Since other configurations are the same as those in the first embodiment, description thereof will be omitted as appropriate.

  The selection unit 23 outputs the transform coefficient for each block input from the transform unit 13 to the curve fitting unit 14 or the quantization unit 21.

  The quantizing unit 21 divides the transform coefficient for each block input from the transform unit 13 by the quantization step and quantizes the quantized coefficient to generate a quantized coefficient, which is transmitted to the inverse quantizing unit 22 and the entropy coding unit 20. Output. The quantization step may be changed for each transform coefficient.

  The inverse quantization unit 22 restores the transform coefficient for each block by multiplying the quantization coefficient input from the quantization unit 21 by a quantization step, and outputs the transform coefficient to the inverse transform unit 16.

  The inverse transformation unit 16 restores the residual block image by performing inverse transformation of the transformation performed by the transformation unit 13 on the transformation coefficient restored by the inverse quantization unit 22 and the inverse curve fitting unit 15.

  The entropy encoding unit 20 performs arbitrary entropy encoding on the fitting function information input from the curve fitting unit 14, the quantization coefficient input from the quantization unit 21, and the prediction information input from the prediction unit 19. To compress data. Then, entropy coding of the fitting function information and the quantized coefficient is performed, and the bit stream having the better coding efficiency based on the cost value is output to the outside of the coding device 2. Also, identification information for identifying which one of the fitting function information and the quantization coefficient has been transmitted (whether the process by the curve fitting unit 14 or the process by the quantization unit 21 has been selected) is output.

  As described above, the encoding device 2 performs the fitting process by the curve fitting unit 14 and the quantization process by the quantization unit 21 on the same input image. For this reason, it is possible to select the one having the better coding efficiency by comparing the conventional quantization process and the fitting process applied to the present invention, and further improve the coding deterioration. .

(Third embodiment)
Next, an encoding apparatus according to the third embodiment of the present invention will be described. FIG. 7 shows a configuration example of an encoding apparatus according to the third embodiment. 7 includes a block division unit 11, a subtraction unit 12, a conversion unit 13, a curve fitting unit 14, an inverse curve fitting unit 15, an inverse conversion unit 16, an addition unit 17, A storage unit 18, a prediction unit 19, an entropy encoding unit 20, a quantization unit 21, and an inverse quantization unit 22 are provided. The encoding device 3 of the third embodiment is different from the encoding device 1 of the first embodiment in that it includes a quantization unit 21 and an inverse quantization unit 22. Since other configurations are the same as those in the first embodiment, description thereof will be omitted as appropriate.

  The quantization unit 21 generates a quantized coefficient by dividing the transform coefficient for each block input from the transform unit 13 by a quantization step using a quantization matrix, and generates a quantized coefficient. Output.

  The curve fitting unit 14 applies curve fitting to data obtained by one-dimensionally arraying the quantization coefficients input from the quantization unit 21 using one or more fitting functions. Then, fitting function information including the coefficient of the fitting function used for curve fitting is output to the inverse curve fitting unit 15 and the entropy coding unit 20.

  The inverse curve fitting unit 15 obtains an approximate curve based on the fitting function information input from the curve fitting unit 14. Then, the data defined by the approximate curve is scanned in the opposite direction to that of the one-dimensional array data generation unit 141 and arranged two-dimensionally to restore the transform coefficients for each block and output to the inverse quantization unit 22.

  The inverse quantization unit 22 restores the transform coefficient for each block by multiplying the transform coefficient input from the inverse curve fitting unit 15 by a quantization step and outputs the transform coefficient to the inverse transform unit 16.

  The inverse transform unit 16 restores the residual block image by performing inverse transform of the transform performed by the transform unit 13 on the transform coefficient restored by the inverse quantization unit 22.

  The entropy encoding unit 20 performs arbitrary entropy encoding on the fitting function information input from the curve fitting unit 14, the quantization information input from the quantization unit 21, and the prediction information input from the prediction unit 19. To compress the data and output the bitstream to the outside.

  As described above, the encoding device 3 performs the fitting process by the curve fitting unit 14 after the conventional quantization process by the quantization unit 21. In general, one-dimensional array data after weighted quantization processing using a quantization matrix has a smoother shape on the low-frequency component side than high-frequency components compared to one-dimensional array data before quantization processing. Since the peak on the side is reduced, it is considered that the shape is more suitable for curve fitting, and it is expected that fitting can be performed with a small number of fitting functions. Therefore, the encoding device 3 is effective when it is desired to give priority to the reduction of the data amount even if the encoding deterioration in the conventional quantization process is allowed.

  It should be noted that a computer can be suitably used to function as the above-described encoding devices 1, 2, and 3, and such a computer describes processing contents for realizing each function of the encoding devices 1, 2, and 3. This program can be realized by storing the program in a storage unit of the computer, and reading and executing the program by the CPU of the computer. This program can be recorded on a computer-readable recording medium.

(Fourth embodiment)
Next, a decoding apparatus according to the fourth embodiment of the present invention will be described. FIG. 8 shows a configuration example of a decoding device according to the fourth embodiment of the present invention. The decoding device 4 illustrated in FIG. 8 includes an entropy decoding unit 31, an inverse curve fitting unit 32, an inverse transformation unit 33, an addition unit 34, a storage unit 35, and a prediction unit 36. The encoded data (the bit stream obtained by encoding the image conversion coefficient using the fitting function) is decoded.

  The entropy decoding unit 31 is paired with the entropy encoding unit 20 of the encoding device 1, decodes the bitstream encoded by the entropy encoding unit 20, and includes one or more fitting functions for defining an approximate curve. Fitting function information, which is information including coefficients, and prediction information, which is information necessary for block image prediction processing, are acquired. Then, the fitting function information is output to the inverse curve fitting unit 32, and the prediction information is output to the prediction unit 36.

  The inverse curve fitting unit 32 obtains an approximate curve based on the fitting function information input from the entropy decoding unit 31. Then, the data defined by the approximate curve is two-dimensionally arranged to restore the transform coefficients for each block and output to the inverse transform unit 33.

  FIG. 9 shows a configuration example of the inverse curve fitting unit 32. In the example illustrated in FIG. 9, the inverse curve fitting unit 32 includes an inverse fitting unit 321 and a two-dimensional array data generation unit 322.

  The inverse fitting unit 321 obtains an approximate curve based on the fitting function information input from the entropy decoding unit 31 and outputs the approximate curve to the two-dimensional array data generation unit 322.

  The two-dimensional array data generation unit 322 calculates conversion coefficients for the number of pixels of the block image from the approximate curve input from the inverse fitting unit 321. Assuming that the approximate curve is y = f (x), for example, when the size of the block image is 8 × 8 pixels, f (0), f (1),..., F (63) are conversion coefficients. Then, the two-dimensional array data generation unit 322 restores the transform coefficient for each block by arranging in two dimensions so as to restore the array transform performed by the curve fitting unit 14 of the encoding device 1, and sends the transform coefficient to the inverse transform unit 33. Output.

  The two-dimensional array data generation unit 322 is identified by the scan direction identification information when acquiring the scan direction identification information that is information for identifying the scan direction of the array conversion by the curve fitting unit 14 of the encoding device 1. Scan in the opposite direction to the scan direction to be arranged in two dimensions.

  The inverse transform unit 33 performs inverse transform on the transform coefficient input from the inverse curve fitting unit 32 to restore a residual block image, and outputs the residual block image to the adder 34. Inverse transform is the same processing as the inverse transform unit 16 of the encoding device 1.

  The addition unit 34 adds the residual block image input from the inverse conversion unit 33 and each pixel value of the prediction block image input from the prediction unit 36, and stores the result as a decoded block image in the storage unit 35, and Output to the outside of the decoding device 4.

  The storage unit 35 is a memory that stores the decoded block image input from the addition unit 34.

  The prediction unit 36 refers to the decoded decoded block image stored in the storage unit 35, performs prediction processing according to the prediction information input from the entropy decoding unit 31, generates a predicted block image, and outputs the prediction block image to the addition unit 34. To do.

  As described above, the decoding device 4 uses the inverse curve fitting unit 32 to obtain an approximate curve based on the fitting function information, and two-dimensionally arranges data defined by the approximate curve to restore transform coefficients for each block. To do. By performing the fitting process using the fitting function, it is possible to improve the coding deterioration due to quantization.

(Fifth embodiment)
Next, a decoding device according to the fifth embodiment of the present invention will be described. FIG. 10 shows a configuration example of a decoding apparatus according to the fifth embodiment of the present invention. 10 includes an entropy decoding unit 31, an inverse curve fitting unit 32, an inverse transform unit 33, an addition unit 34, a storage unit 35, a prediction unit 36, an inverse quantization unit 37, And a selector 38 for decoding the data encoded by the encoding device 2. The decoding device 5 according to the fifth embodiment is different from the decoding device 4 according to the fourth embodiment in that an inverse quantization unit 37 is further provided in parallel with the inverse curve fitting unit 32. Since other configurations are the same as those in the fourth embodiment, description thereof will be omitted as appropriate.

  The entropy decoding unit 31 is paired with the entropy encoding unit 20 of the encoding device 2, decodes the data compressed by the entropy encoding unit 20, and obtains fitting function information, quantization coefficients, and prediction information. Then, the fitting function information is output to the inverse curve fitting unit 32, the quantization coefficient is output to the inverse quantization unit 37, and the prediction information is output to the prediction unit 36.

  The inverse curve fitting unit 32 obtains an approximate curve based on the fitting function information input from the entropy decoding unit 31. Then, the data defined by the approximate curve is two-dimensionally arranged to restore the transform coefficients for each block and output to the inverse transform unit 33.

  The inverse quantization unit 37 restores the transform coefficient for each block by multiplying the quantization coefficient input from the entropy decoding unit 31 by a quantization step, and outputs the restored transform coefficient to the inverse transform unit 33.

  The inverse transform unit 33 performs inverse transform on the transform coefficient input from the inverse curve fitting unit 32 or the inverse quantization unit 37 to restore the residual block image, and outputs the residual block image to the adder 34.

  As described above, since the decoding device 5 includes the inverse quantization unit 37 in parallel with the inverse curve fitting unit 32, the data encoded by the encoding device 2 can be decoded.

(Sixth embodiment)
Next, a decoding apparatus according to the sixth embodiment of the present invention will be described. FIG. 11 shows a configuration example of a decoding apparatus according to the sixth embodiment of the present invention. The decoding device 6 illustrated in FIG. 11 includes an entropy decoding unit 31, an inverse curve fitting unit 32, an inverse transform unit 33, an addition unit 34, a storage unit 35, a prediction unit 36, and an inverse quantization unit 37. The data encoded by the encoding device 3 is decoded. The decoding device 6 according to the sixth embodiment is different from the decoding device 4 according to the fourth embodiment in that an inverse quantization unit 37 is further provided in series with the inverse curve fitting unit 32. Since other configurations are the same as those in the fourth embodiment, description thereof will be omitted as appropriate.

  The entropy decoding unit 31 is paired with the entropy encoding unit 20 of the encoding device 3, decodes the data compressed by the entropy encoding unit 20, and acquires fitting function information and prediction information. Then, the fitting function information is output to the inverse curve fitting unit 32, and the prediction information is output to the prediction unit 36.

  The inverse curve fitting unit 32 obtains an approximate curve based on the fitting function information input from the entropy decoding unit 31. Then, the data defined by the approximate curve is two-dimensionally arranged to restore the transform coefficients for each block and output to the inverse quantization unit 37.

  The inverse quantization unit 37 restores the transform coefficient for each block by multiplying the transform coefficient input from the inverse curve fitting unit 32 by a quantization step, and outputs it to the inverse transform unit 33. The quantization step may be acquired from the encoding device 3 or may have a quantization table common to the encoding device 3 in advance.

  The inverse transform unit 33 performs inverse transform on the transform coefficient input from the inverse quantization unit 37 to restore the residual block image, and outputs the residual block image to the addition unit 34.

  As described above, since the decoding device 6 includes the inverse quantization unit 37 in series with the inverse curve fitting unit 32, the data encoded by the encoding device 3 can be decoded.

  It should be noted that a computer can be suitably used to function as the above-described decoding devices 4, 5, and 6, and such a computer is a program describing processing contents for realizing the functions of the decoding devices 4, 5, and 6. Is stored in a storage unit of the computer, and the program is read and executed by the CPU of the computer. This program can be recorded on a computer-readable recording medium.

  Although the above embodiment has been described as a representative example, it will be apparent to those skilled in the art that many changes and substitutions can be made within the spirit and scope of the invention. Therefore, the present invention should not be construed as being limited by the above-described embodiments, and various modifications and changes can be made without departing from the scope of the claims. For example, a plurality of constituent blocks described in the embodiments can be combined into one, or one constituent block can be divided.

1, 2, 3 Encoding device 4, 5, 6 Decoding device 11 Block division unit 12 Subtraction unit 13 Conversion unit 14 Curve fitting unit 15 Inverse curve fitting unit 16 Inverse conversion unit 17 Addition unit 18 Storage unit 19 Prediction unit 20 Entropy code Conversion unit 21 Quantization unit 22 Inverse quantization unit 23 Selection unit 31 Entropy decoding unit 32 Inverse curve fitting unit 33 Inverse transformation unit 34 Addition unit 35 Storage unit 36 Prediction unit 37 Inverse quantization unit 141 One-dimensional array data generation unit 142 Fitting Part

Claims (10)

An encoding device that encodes using a fitting function,
A block dividing unit that generates a block image by dividing an input image into a plurality of blocks;
A prediction unit that generates a predicted block image obtained by predicting the block image;
A transform unit that performs a transform process on a residual block image indicating a difference between the block image and the predicted block image, and calculates a transform coefficient for each block;
A curve fitting unit that approximates a one-dimensional array of transform coefficients for each block and generates fitting function information including a coefficient of a fitting function for defining an approximate curve;
An entropy encoding unit that performs entropy encoding of the fitting information and prediction information necessary for prediction of the prediction block image to generate a bitstream;
An inverse curve fitting unit that obtains the approximate curve based on the fitting function information, arranges data defined by the approximate curve in two dimensions, and restores the conversion coefficient for each block;
An inverse transform unit that performs inverse transform on the restored transform coefficient and restores the residual block image,
The encoding device, wherein the prediction unit generates the prediction block image with reference to the decoded residual block image and a decoded image obtained by adding the prediction block image. The curve fitting portion is
A one-dimensional array data generation unit that generates one-dimensional array data in which the transform coefficient is scanned in a predetermined direction for each block to form a one-dimensional array;
A fitting unit that approximates the curve using one or more fitting functions for the one-dimensional array data;
The encoding device according to claim 1, comprising: The curve fitting portion is
A one-dimensional array data generation unit that scans the transform coefficient in a plurality of directions for each block and generates a plurality of one-dimensional array data that is one-dimensionally arranged;
A fitting unit that approximates a curve using one or more fitting functions for each of the plurality of one-dimensional array data,
The entropy encoding unit performs entropy encoding of the fitting function information for each scan direction, determines an optimal scan direction based on a cost value, a bitstream generated in the scan direction, and the scan direction The encoding apparatus according to claim 1, wherein identification information for identifying is output. A quantization unit that generates a quantization coefficient by dividing the transform coefficient for each block by a quantization step;
An inverse quantization unit that restores the transform coefficient for each block by multiplying the quantization coefficient by a quantization step;
The inverse transform unit performs an inverse transform on the transform coefficients restored by the inverse quantization unit and the inverse curve fitting unit, and restores the residual block image,
The entropy coding unit performs entropy coding of the fitting function information and the quantized coefficient, and outputs a bitstream having better coding efficiency based on a cost value. To 4. The encoding device according to any one of claims 1 to 3. A quantization unit that divides the transform coefficient for each block by a quantization step;
The curve fitting unit approximates a one-dimensional array of transform coefficients divided by the quantization unit using one or more fitting functions, and generates fitting function information including the coefficients of the fitting function;
The inverse curve fitting unit obtains the approximate curve based on the fitting function information, arranges data defined by the approximate curve in two dimensions, and restores the transform coefficient divided by the quantization unit,
Further comprising an inverse quantization unit that restores the transform coefficient for each block by multiplying the transform coefficient generated by the inverse curve fitting unit by a quantization step;
4. The method according to claim 1, wherein the inverse transform unit performs inverse transform on the transform coefficient restored by the inverse quantization unit to restore the residual block image. 5. The encoding device described. A decoding device for decoding a bitstream encoded using a fitting function,
An entropy decoding unit that decodes the bitstream to obtain fitting function information including coefficients of a fitting function for defining an approximate curve, and prediction information necessary for prediction of a block image;
An approximate curve is obtained based on the fitting function information, and an inverse curve fitting unit that arranges data defined by the approximate curve in two dimensions and restores a conversion coefficient for each block;
An inverse transform unit that performs an inverse transform on the transform coefficient to restore a residual block image;
A prediction unit that generates a predicted block image obtained by predicting the block image according to the prediction information;
An addition unit that adds the residual block image and the prediction block image to generate a decoded block image, and
The prediction device generates the prediction block image with reference to the decoded block image that has been decoded. The entropy decoding unit further acquires a quantization coefficient,
A dequantization unit that restores a transform coefficient for each block by multiplying the quantization coefficient by a quantization step;
The inverse transform unit performs the inverse transform on the transform coefficient restored by the inverse quantization unit or the inverse curve fitting unit, and restores the residual block image. Decoding device. An inverse quantization unit that restores the second transform coefficient for each block by multiplying the first transform coefficient restored by the inverse curve fitting unit by a quantization step;
The decoding device according to claim 6, wherein the inverse transform unit performs the inverse transform on the second transform coefficient to restore the residual block image.   The program for functioning a computer as an encoding apparatus as described in any one of Claim 1 to 5.   The program for functioning a computer as a decoding apparatus as described in any one of Claims 6-8.

Images