JP2009021673A - Coded parameter determining method, coded parameter determining device, coded parameter determining program and computer-readable recording medium recording the program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To establish a technique for preventing deterioration in image quality in a skip mode and reducing code amount of the skip mode in determining a coded parameter by minimizing a cost function using coded distortion weighted on the basis of time spatial visual characteristics. <P>SOLUTION: The coded parameter determining method includes: coding target block, when a prediction mode of minimum cost selected by calculating the cost by using a distortion quantity weighted using a sensitivity coefficient indicating time spatial visual sensitivity is a skip mode, the skip mode is not determined as it is as the optimum prediction mode of the block, but the displacement quantity to be used in coding is compared to the estimated displacement quantity calculated prior to coding; and when the difference between the two displacement quantities is large and the estimated displacement quantity is large, the prediction mode of the minimum cost selected by calculating the cost by using the unweighted distortion quantity is determined as a prediction mode. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an encoding parameter determination method and apparatus used in moving image encoding, an encoding parameter determination program used to realize the encoding parameter determination method, and a computer-readable recording medium on which the program is recorded. In particular, when encoding a moving image in consideration of spatio-temporal visual sensitivity, a code that realizes determination of an encoding parameter that can efficiently reduce the amount of code while maintaining the subjective quality of the decoded image The present invention relates to an encoding parameter determination method and apparatus, an encoding parameter determination program used for realizing the encoding parameter determination method, and a computer-readable recording medium on which the program is recorded.

[Encoding method using cost function of square error criterion]
H. In H.264, with the introduction of intra prediction and variable shape motion compensation, the types of prediction modes are increasing compared to the conventional standardized method. For this reason, in order to reduce the amount of codes while maintaining a constant subjective image quality, it is necessary to select an appropriate prediction mode. H. In the H.264 reference software JM (see Non-Patent Document 1), the following prediction mode that minimizes the RD cost is selected. In the following notation, the symbol ^ in “^ X” (where X is a letter) indicates a symbol attached to “X”.

Here, S is an original signal, q is a quantization parameter, m is a number representing a prediction mode, ^ S _{m, q} is predicted for the original signal S using the prediction mode m, and the quantization parameter q is It is a decoded signal when quantized by using. Further, λ is a Lagrange's undetermined multiplier used for mode selection. Further, D (S, ^ S _{m, q} ) is a sum of square errors shown in the following equation.

Here, S ^Y , S ^U , and S ^V are Y, U, and V components of the original signal, respectively, and ＾ S ^Y _{m, q} , ＳS ^U _{m, q} , and ＳS ^V _{m, q} are respectively the decoded signals. Y, U and V components.

  H. The calculation method of the decoded signal in H.264 is shown below. The symbols used for the explanation are summarized in the following table.

H. In the H.264 encoding process, the orthogonal transformation using the transformation matrix Φ is applied to the prediction error signal R _m (= S−P _m ) when the prediction of the mode number m is used as follows: .

Here, Φ ^t represents a transposed matrix with respect to the transformation matrix Φ. Note that the transformation matrix Φ is an orthogonal matrix of integer elements expressed by the following equation.

  Next, since the matrix Φ is a non-normal matrix, processing equivalent to matrix normalization is performed as shown in the following equation.

C _n = N (C)
Further, quantization using the quantization parameter q is performed on C _n as follows. H. In the H.264 reference software JM, normalization is built into quantization.

V = Q (C _n )
On the other hand, H. In the H.264 decoding process, V is inversely quantized as in the following equation to obtain a decoded value of a transform coefficient.

Next, inverse transformation is applied to { _circumflex over (C) _{} q} as shown in the following equation to obtain a prediction error decoded signal.

  Finally, a decoded signal of the encoding target image is obtained by the following equation.

[Weighting distortion amount considering subjective image quality]
As described above, H.P. The measure of subjective image quality used in the H.264 reference software JM is a square error. However, this square error is not necessarily a distortion amount reflecting subjective image quality degradation. For example, a change in a high frequency component is less visually detected than a change in a low frequency component. However, an encoder (for example, JM) that does not use such visual characteristics still has room for improvement in terms of efficient code amount reduction.

  Therefore, studies have been made to use the difference in visual sensitivity with respect to spatio-temporal frequency components. A distortion amount corresponding to the subjective image quality is defined by weighting the distortion amount for each spatial frequency component in accordance with the visual sensitivity with respect to the orthogonal transform coefficient. Further, in consideration of the visual sensitivity in the time direction, the above-described weighted distortion amount is further weighted in accordance with the shift amount. Thus, the weighting amount weighted based on the visual sensitivity of the space time is used in the cost function for selecting the encoding parameter.

  The weighting based on the visual sensitivity for the quantization error signal will be described below. Here, it is assumed that the RD cost of the following equation is used.

Here, C _m is a transform coefficient for the prediction residual signal R _m when the mode number m is used, and ^ C _{m, q} is a coefficient obtained by quantizing and dequantizing C _m with the quantization parameter q. Is the decoded value. The following weighted distortion amount is used as the distortion amount used for calculating the RD cost.

Here, the symbols surrounding 16 / N and 8 / N mean truncation of the decimal part. C ^{Y (i)} _m [k, l], C ^{U (i)} _m [k, l], and C ^{V (i)} _m [k, l] are elements of C _m , and are macroblocks (Y component) Of the sub-block (N × N [pixel]) in 16 × 16 [pixel] in the case of U, V component and 8 × 8 [pixel] in the case of U and V components) is scanned i-th in the raster scan. This is a conversion coefficient included in the sub-block. ^ C ^{Y (i)} _{m, q} [k, l], ^ C ^{U (i)} _{m, q} [k, l] and ^ C ^{V (i)} _{m, q} [k, l] are ^ C _{m , q} elements and sub-blocks (N × N [pixels]) in a macroblock (16 × 16 [pixels] for Y components and 8 × 8 [pixels] for U and V components) Among them, the decoding transform coefficient included in the i-th scanned sub-block in raster scanning.

Further, W ^Y _{k, l} , W ^U _{k, l} and W ^V _{k, l} are weighting factors set to 1 or less, and take a smaller value as the spatial frequency and time frequency are higher.

In the above equation, setting W ^Y _{k, l} , W ^U _{k, l} , W ^V _{k, l} to a small value is equivalent to estimating the quantization distortion D (C _m , ^ C _{m, q} ) to be small. To do. From the normality of orthogonal transformation, if W ^Y _{k, l} = 1, W ^U _{k, l} = 1, W ^V _{k, l} = 1 for all _{k, 1} , the above-mentioned weighted distortion amount Is equivalent to the sum of squared errors.
KPLim and G. Sullivan and T. Wiegand, Text Description of Joint Model Reference Encoding Methods and Decoding Concealment Methods. Joint Video Team (JVT) of ISO / IEC MPEG and ITU-T VCEG, JVT-R95, Jan., 2006.

  H. In H.264, a prediction vector for a motion vector of a macroblock to be encoded is generated using the motion vectors of the macroblock immediately above, immediately left, and immediately above left.

  In the case of a prediction mode other than the skip mode, a motion vector is expressed using difference information from this prediction vector.

  On the other hand, in the skip mode, this prediction vector is used as a motion vector. Further, the inter-frame prediction residual is set to zero and a decoded signal is generated.

  For this reason, if reference is made to a region of a picture with a different prediction vector, a large coding distortion is superimposed on the skip mode macroblock. On the other hand, the amount of codes can be kept extremely small.

  Since such coding distortion and code amount are in a trade-off relationship, H.J. In the H.264 encoder, a Lagrangian cost function is introduced as a weighted sum of encoding distortion and code amount, and a prediction mode is selected in consideration of this trade-off balance.

However, when weighting is applied to the distortion amount based on the above-described contrast sensitivity function (weighting by W ^Y _{k, l} , W ^U _{k, l} , W ^V _{k, l} ), the skip mode includes large image quality degradation. Nevertheless, the skip mode may be selected as the optimal prediction mode.

  The occurrence of such image quality degradation suggests that it is necessary to consider the peculiarity of this mode with respect to the skip mode when weighting the coding distortion. However, such a study has not been made in the prior art.

  The present invention has been made in view of such circumstances, and in determining the encoding parameter by minimizing the cost function using the encoding distortion weighted based on the spatio-temporal visual characteristics, the skip mode is used. Established a new coding parameter determination technology that can avoid degradation of image quality in skip mode and enjoy the maximum benefit of code amount reduction in skip mode by introducing a measure of coding distortion that appropriately evaluates subjective image quality The purpose is to do.

[1] First Configuration To achieve the above object, the coding parameter determination apparatus of the present invention performs transform coding and quantization on a prediction error signal obtained by intraframe prediction and interframe prediction. In order to determine the encoding parameters used for image encoding that performs information compression by (i), the estimated displacement amount indicating the temporal movement of the image signal is calculated prior to the encoding process for the block to be encoded. And (b) a block in which the prediction mode with the lowest cost selected by calculating the cost using the weighted distortion amount using the sensitivity coefficient indicating the spatiotemporal visual sensitivity is the skip mode. As a comparison between the amount of displacement used when encoding the block and the estimated amount of displacement calculated by the calculation means, the difference between the two amounts of displacement is greater than a prescribed threshold, In addition, by determining whether or not the size of the estimated displacement amount is larger than a prescribed threshold, it is determined whether or not the block is likely to fall in image quality deterioration when encoded in the skip mode. (C) For a block to be determined that the determination means determines that it is not the corresponding block, the skip mode is determined as the optimum prediction mode of the block, and the determination target that the determination means determines to be the corresponding block is determined. For a block, a determination means for determining a prediction mode with the least cost selected by calculating a cost using an unweighted distortion amount as an optimal prediction mode of the block; and (d) a weighted distortion amount. Prediction of the block with the least cost prediction mode selected by calculating the cost using the skip mode A second determination unit that determines the mode as the optimum prediction mode of the block, and (e) a block selected by the determination unit is selected by calculating a cost using an unweighted distortion amount When the prediction mode with the lowest cost is also the skip mode, a third determination unit that determines the skip mode as the optimum prediction mode of the block without performing the determination process by the determination unit is provided. .

  The encoding parameter determination method of the present invention realized by the operation of each of the above processing means can also be realized by a computer program, and this computer program is provided by being recorded on an appropriate computer-readable recording medium. Alternatively, the present invention is realized by being provided via a network, installed when the present invention is carried out, and operating on a control means such as a CPU.

  In the coding parameter determination apparatus of the present invention configured as described above, a block is selected by calculating a cost using a distortion amount weighted using a sensitivity coefficient indicating spatiotemporal visual sensitivity for a block to be coded. If the prediction mode with the lowest cost is a skip mode (a prediction mode instructing to use a prediction vector as a motion vector), the optimal prediction mode for the block is not determined and the skip mode is not determined as it is. The displacement amount used for encoding is compared with the estimated displacement amount calculated prior to encoding, and if the difference between the two displacement amounts is large and the estimated displacement amount is large, no weighting is performed. The prediction mode with the lowest cost selected by calculating the cost using the amount of distortion is determined as the optimal prediction mode for the block. That.

[2] Second Configuration Further, in order to achieve the above object, the coding parameter determination apparatus of the present invention performs transform coding and prediction on a prediction error signal obtained by intraframe prediction and interframe prediction. In order to determine encoding parameters used for image encoding for performing information compression by quantization, (a) an estimated displacement amount indicating temporal movement of an image signal prior to encoding processing for a block to be encoded And (b) a block in which the prediction mode with the lowest cost selected by calculating the cost using the distortion amount weighted using the sensitivity coefficient indicating the spatiotemporal visual sensitivity is the skip mode. As a judgment target, the displacement amount used when coding the block is compared with the estimated displacement amount calculated by the calculation means, and the difference between the two displacement amounts is larger than a prescribed threshold value. And whether or not the estimated displacement amount is larger than a predetermined threshold value, it is determined whether or not the block corresponds to a block that is likely to have a large image quality deterioration when encoded in the skip mode. And (c) for the block to be determined that the determining unit determines not to be the corresponding block, the skip mode is determined as the optimum prediction mode of the block, and the determination unit determines that the block is the corresponding block With respect to the target block, using a determination unit that determines the prediction mode selected as the next lowest cost after the skip mode with the lowest cost as the optimum prediction mode of the block, and (d) using the weighted distortion amount The prediction mode selected for the block with the lowest cost selected by calculating the cost is not the skip mode. And (2) a minimum cost selected by calculating the cost using the unweighted distortion amount for the block to be determined by (2) the determination unit. If the prediction mode is also the skip mode, the third determination means for determining the skip mode as the optimum prediction mode for the block without performing the determination process by the determination means is provided.

  The encoding parameter determination method of the present invention realized by the operation of each of the above processing means can also be realized by a computer program, and this computer program is provided by being recorded on an appropriate computer-readable recording medium. Alternatively, the present invention is realized by being provided via a network, installed when the present invention is carried out, and operating on a control means such as a CPU.

  In the coding parameter determination apparatus of the present invention configured as described above, a block is selected by calculating a cost using a distortion amount weighted using a sensitivity coefficient indicating spatiotemporal visual sensitivity for a block to be coded. If the prediction mode with the lowest cost is a skip mode (a prediction mode instructing to use a prediction vector as a motion vector), the optimal prediction mode for the block is not determined and the skip mode is not determined as it is. The amount of displacement used for encoding is compared with the estimated amount of displacement calculated prior to encoding. When the difference between the two amounts of displacement is large and the estimated amount of displacement is large, the cost is minimized. The prediction mode selected as the next lowest cost after the skip mode is determined as the optimum prediction mode of the block.

[3] About the present invention As described above, in the present invention, when the prediction mode with the minimum cost selected by calculating the cost using the weighted distortion amount for a certain block is the skip mode, Rather than determining the skip mode as the optimum prediction mode of the block, the displacement amount used for encoding is compared with the estimated displacement amount calculated prior to encoding, and the two displacement amounts are compared. If the discrepancy between the two is large and the estimated displacement is large, it is selected with the lowest cost, but it is highly possible that coding with the skip mode is likely to include a large image quality degradation. A configuration is adopted in which a prediction mode other than the mode is determined as the optimum prediction mode of the block.

  The reason why the skip mode is selected as the optimum prediction mode despite the large image quality degradation will be considered below.

  In the conventional method, the true shift amount of each macroblock is estimated prior to the encoding process. The amount of change obtained here is called an estimated amount of change. The estimated amount of displacement is used to weight the amount of distortion. The shift amount used in the encoding process is separately calculated by an encoder. This shift amount is called an encoder shift amount.

  In the conventional method, it is assumed that the estimated shift amount and the encoder shift amount do not greatly deviate. On the other hand, in the skip mode, since the motion search is not performed, there is a possibility that the prediction vector that is the encoder shift amount greatly deviates from the estimated shift amount. As a result, if a different picture area is referred to, a large coding distortion occurs.

  However, when the estimated shift amount is large, even if such a large coding distortion occurs, the coding distortion is estimated to be small by weighting. As a result, the cost is reduced and the optimum mode may be selected.

  Therefore, in the present invention, the distortion amount is weighted in consideration of the degree of deviation between the estimated shift amount and the encoder shift amount. In other words, if the estimated shift amount and the encoder shift amount deviate greatly, and the estimated shift amount is large, the possibility of the above-mentioned image quality degradation increases, so the cost is calculated using the unweighted distortion amount. The cost next to the prediction mode with the lowest cost selected (skip mode) is determined by calculating the cost using the weighted distortion amount. It is decided to use a prediction mode selected as a small one.

  Next, an evaluation function for evaluating the degree of deviation between the estimated shift amount and the encoder shift amount, and an evaluation function for evaluating the magnitude of the estimated shift amount will be described.

[When the estimated displacement reference frame is the same as the encoder displacement reference frame]
The following represents a macro block for the estimated displacement amount _{"^ d = (^ d x} , ^ d y)" and the encoder displacement amount "d = (dx, dy) " represents a.

  When the following two conditions (1) and (2) are satisfied, no weighting is applied to the distortion amount.

  Here, d () is a function representing the degree of divergence between two vectors. For example, the inner product of two vectors represented by the following expression (1) or two expressions represented by the following expression (2): A vector distance, a distance between two vectors represented by the following equation (3), or the like is used.

  The left side of the condition (2) is the absolute value norm of the vector represented by the following formula (4). Here, θ and ψ are thresholds given from the outside.

[When the estimated transition amount reference frame is different from the encoder transition amount reference frame]
Macroblock to be processed is present in the t-th frame, the estimated displacement amount refers to the t-th-r _e frame, the encoder displacement amount is in the case of referring to the t-r _c frame, the next 2 When one condition is satisfied, weighting is not performed on the distortion amount.

That is, when the estimated shift amount reference frame and the encoder shift amount reference frame are different, a value obtained by normalizing the two shift amounts to the shift amount between adjacent frames is used for evaluation. Further, r _c, r _e can take both positive and negative values. A positive value corresponds to forward prediction, and a negative value corresponds to backward prediction. Thus, for dividing each displacement amount in r _c, r _e, together with the length normalization of the vector, when the direction of the vector are different, also it means there align the direction of both vectors.

[Skip mode in B frame]
In the case of the skip mode in the B frame, two types of methods (spatial direct and temporal direct) are defined as methods for calculating the encoder shift amount.

(I) In the case of temporal direct The encoder shift amount is an anchor which is a block located at the same position as the encoding target block in the anchor picture (usually, the nearest reference frame behind the encoding target frame in the display order). This is set using the block shift amount (referred to as anchor shift amount). Bi-predicted motion vector, where d _Col is the anchor displacement, t _d is the time interval between the anchor picture and its reference frame, and t _b is the time interval between the reference frame and the encoding target frame Is obtained as follows.

  It can be seen from the above equation that the bi-directional motion vector can be obtained from the anchor shift amount. Therefore, when the following two conditions are satisfied, the distortion amount is not weighted.

  Apparently, even if there are two bi-directional motion vectors, the directions of the two vectors are dependent on the anchor shift amount. For this reason, if the estimated shift amount and the anchor shift amount are evaluated, the respective divergence degrees between the two vectors of bidirectional prediction and the estimated shift amount can be obtained. For this reason, in the above equation, the degree of deviation between the estimated shift amount and the anchor shift amount is used as an evaluation scale.

(Ii) Spatial direct One of the following is determined according to the value of the anchor displacement.

- a motion vector of the bidirectional prediction to be derived from the motion vectors of-neighboring macroblocks to set the motion vector of the bidirectional prediction to zero vector and d _L0 and d _L1, each reference frame the t-r _c0 frame and the t Assuming that _{−rc 1} frame is used, weighting is not applied to the distortion amount when the following condition is satisfied.

  In the present invention, using such an evaluation function, the degree of deviation between the displacement amount (encoder displacement amount) used for encoding and the estimated displacement amount calculated prior to encoding, and the estimated displacement amount When the size is evaluated and it is determined that coding in the skip mode is likely to include large image quality degradation based on the size, a prediction mode other than the skip mode is selected as the optimal prediction of the block. The mode is determined.

  In the present invention, when weighting is applied to the distortion amount based on the spatio-temporal contrast sensitivity function, the skip mode is selected so as not to be selected as the optimum prediction mode even though the skip mode includes a large image quality degradation. By introducing a measure of coding distortion that appropriately evaluates the subjective image quality of the mode, image quality degradation is avoided.

  As a result, according to the present invention, it is possible to avoid image quality deterioration in the skip mode and enjoy the maximum benefit of code amount reduction in the skip mode. In other words, even in the skip mode that instructs to use a prediction vector as a motion vector, the code amount is reduced for a region that is difficult to detect visually, so that it is efficient while maintaining the subjective image quality of the decoded image. Therefore, the amount of codes can be reduced.

  Hereinafter, the present invention will be described in detail according to embodiments.

  FIG. 1 shows a device configuration of a video encoding device 1 having the present invention.

  The video encoding apparatus 1 provided with the present invention is an H.264 standard. H.264 is used to perform processing for encoding a moving image. As shown in this figure, an encoding parameter determination unit 10 that determines an encoding parameter used for encoding an encoding target macroblock, and an encoding parameter determination And an encoding unit 20 that encodes the encoding target macroblock using the encoding parameter determined by the unit 10.

  The encoding parameter determination unit 10 receives the encoding target frame signal, the reference frame signal, and the quantization parameter as inputs to determine a prediction mode that is one of the encoding parameters, and predicts the encoding target macroblock. A prediction mode determination unit 11 that performs a process of determining a mode, and the prediction mode determination unit 11 calculates an estimated displacement of an encoding target macroblock in order to perform the determination. An amount calculation unit 12 is provided.

  FIG. 2 illustrates an example of a flowchart executed by the encoding parameter determination unit 10. Here, in this flowchart, it is assumed that the encoding parameter determination unit 10 determines the quantization parameter and the prediction mode.

  When determining the quantization parameter and the prediction mode to be used for encoding the encoding target macroblock, the encoding parameter determination unit 10 firstly, in step S10, in the register C, as shown in the flowchart of FIG. Register C and register M are initialized by storing an initial cost indicating a large value for, and storing a meaningless value for register M.

  Subsequently, in step S11, the minimum value QPmin of the quantization parameter is set to the variable QP that stores the value of the quantization parameter.

  Subsequently, in step S12, the quantization mode set in the variable QP is designated and the prediction mode determination unit 11 is activated to determine the optimum prediction mode for the quantization parameter set in the variable QP. The prediction mode determination process executed at this time will be described later according to the flowcharts of FIGS. 3 and 5.

  Subsequently, in step S13, the cost for encoding using the quantization parameter set in the variable QP and the prediction mode determined in step S12 is calculated.

  Subsequently, in step S14, it is determined whether or not the cost calculated in step S13 is smaller than the cost stored in the register C.

  When it is determined that the cost calculated in step S13 is smaller according to this determination process, the process proceeds to step S15, the cost calculated in step S13 is stored in the register C, and the variable QP is set in the subsequent step S16. The set information of the quantization parameter and the identification information of the prediction mode determined in step S12 is stored in the register M. On the other hand, when it is determined that the cost calculated in step S13 is larger, the processes in steps S15 and S16 are omitted.

  Subsequently, in step S17, it is determined whether or not the value of the quantization parameter set in the variable QP exceeds the maximum value QPmax. When determining that the value does not exceed the maximum value QPmax, the process proceeds to step S18. Thus, the value of the quantization parameter set in the variable QP is incremented by the specified amount ΔQP, and the process returns to step S12.

  When it is determined in this way that the value of the quantization parameter set in the variable QP exceeds the maximum value QPmax in step S17 by repeating the processing in steps S12 to S18, the process proceeds to step S19. Thus, it is determined that the encoding is performed using the quantization parameter and the prediction mode stored in the register M, and the determination process of the quantization parameter and the prediction mode is ended.

[1] First Embodiment FIG. 3 illustrates an embodiment of a flowchart executed by the prediction mode determination unit 11.

  Next, according to this flowchart, the prediction mode determination process executed by the prediction mode determination unit 11 in the present embodiment will be described in detail.

  When a prediction mode determination request is issued by designating a quantization parameter for the encoding target macroblock, the prediction mode determination unit 11 first, as shown in the flowchart of FIG. Is set to the initial value of the prediction mode (prediction mode identification information as the initial value), and the initial cost indicating a large value for the two registers C0 and C1 that store the RD cost. And the registers X, C0, C1, M0, and M1 are initialized by storing meaningless values in the two registers M0 and M1 that store the prediction mode identification information.

  As will be described below, in addition to these registers X, C0, C1, M0, and M1, a register α that stores a code amount, a register β that stores an undetermined multiplier, and a weighted distortion amount are stored. Four registers are used: a register γ0 and a register γ1 that stores an unweighted distortion amount.

  Subsequently, in step S101, the shift amount of the encoding target macroblock is estimated. This estimation method is assumed to be given from the outside. For example, H.M. It is also possible to use a motion vector calculated by the H.264 reference software JM as an estimated value of the shift amount used below. Alternatively, it is also possible to calculate the estimated displacement amount according to a standard for minimizing the sum of absolute value errors between the encoding target macroblock and the reference macroblock.

  Subsequently, in step S102, the prediction mode of the register X, the prediction vector, the quantization parameter, the encoding target frame signal, and the reference frame signal are input, and the code amount when encoding using the prediction mode is calculated. The calculated value is written to the register α. The specific calculation method is as follows. It follows the method of H.264 reference software JM.

  Subsequently, in step S103, the prediction mode and the quantization parameter of the register X are input, an undetermined multiplier for encoding using the prediction mode is calculated, and the calculated value is written to the register β. The specific calculation method is as follows. It follows the method of H.264 reference software JM.

  Subsequently, in step S104, a weight is first determined based on the estimated displacement calculated in step S101, and then the prediction mode, prediction vector, quantization parameter, encoding target frame signal, reference frame of the register X are determined. Based on the input signal and the determined weight, the weighted distortion amount when encoding using the prediction mode is calculated, and the calculated value is written to the register γ0. A specific calculation method follows the above-described formula (D) (the formula shown in [Equation 9]).

  Subsequently, in step S105, the code amount stored in the register α, the undetermined multiplier stored in the register β, and the weighted distortion amount stored in the register γ0 are read to calculate the RD cost. A specific calculation method follows the above-described formula (C) (the formula shown in [Equation 8]).

  Subsequently, in step S106, when the calculated RD cost is compared with the value of the register C0 and it is determined that the calculated RD cost is smaller than the value of the register C0, step S107 is performed. Then, the calculated RD cost is stored in the register C0, and in step S108, the prediction mode identification information stored in the register X is stored in the register M0. On the other hand, when it is determined in step S106 that the calculated RD cost is larger than the value of the register C0, the processes in steps S107 and 108 are omitted.

  The processes in steps S104x to S108x are executed in parallel with the processes in steps S104 to S108.

That is, in step S104x, the prediction mode, the prediction vector, the quantization parameter, the encoding target frame signal, and the reference frame signal of the register X are input, and the unweighted distortion amount when encoding using the prediction mode is calculated. The calculated value is written in the register γ1. As a specific calculation method, there is no weighting (W _{k, l} ^Y = W _{k, l} ^U = W _{k, l} ^V = 1) in the above-described formula (D) (the formula shown in [Equation 9]). Or according to the above-described formula (B) (the formula shown in [Equation 2]).

  Subsequently, in step S105x, the code amount stored in the register α, the undetermined multiplier stored in the register β, and the unweighted distortion amount stored in the register γ1 are read to calculate the RD cost. The specific calculation method is based on the above-described equation (C) (the equation shown in [Equation 8]) using a distortion-free value, or the above-described equation (A) ([Equation 1]). Follow the formula shown in

  Subsequently, in step S106x, when the calculated RD cost is compared with the value of the register C1, and it is determined that the calculated RD cost is smaller than the value of the register C1, step S107x Then, the calculated RD cost is stored in the register C1, and in step S108x, the prediction mode identification information stored in the register X is stored in the register M1. On the other hand, when it is determined in step S106x that the calculated RD cost is larger than the value of the register C1, the processes in steps S107x and 108x are omitted.

  When the processes of step S108 and step S108x are finished, subsequently, in step S109, it is determined whether or not all prediction modes have been processed, and when it is determined that not all prediction modes have been processed, step S110 is performed. Then, one prediction mode is selected from the unprocessed prediction modes in accordance with a predetermined order, the identification information of the selected prediction mode is stored in the register X, and the process returns to step S102.

  In this manner, by repeating the processing of step S102 to step S110, when it is determined in step S109 that all prediction modes have been processed, the process proceeds to step S111, and the prediction mode stored in the register M0 is determined. It is determined whether or not the identification information indicates a skip mode.

  That is, it is determined whether or not the prediction mode with the minimum RD cost evaluated using the weighted distortion amount is the skip mode.

  When it is determined that the prediction mode identification information stored in the register M0 is not the skip mode according to the determination process in step S111, the image quality deterioration due to the skip mode does not matter, so the process proceeds to step S115 and is stored in the register M0. The stored prediction mode is output as the optimum prediction mode for the encoding target block, and the process is terminated.

  On the other hand, when it is determined that the prediction mode identification information stored in the register M0 is the skip mode according to the determination process in step S111, the process proceeds to step S112, and the prediction mode identification information stored in the register M1 is skipped. It is determined whether or not the mode is indicated.

  That is, it is determined whether or not the prediction mode with the minimum RD cost evaluated using the unweighted distortion amount is the skip mode.

  When it is determined that the prediction mode identification information stored in the register M1 is the skip mode according to the determination process in step S112, the skip mode is set regardless of whether the weighted distortion amount or the unweighted distortion amount is used. Since it has been selected, the process proceeds to step S115, the skip mode, which is the prediction mode stored in the register M0, is output as the optimum prediction mode for the block to be encoded, and the process is terminated.

  On the other hand, when it is determined that the prediction mode identification information stored in the register M1 is not the skip mode according to the determination process in step S112, the process proceeds to step S113, and the estimated displacement amount and the encoding target estimated in step S101 are processed. It is determined whether the degree of deviation from the predicted vector of the block is greater than a threshold and whether the estimated displacement is greater than the threshold.

  That is, since it is determined in step S111 that the prediction mode with the minimum RD cost is the skip mode, the prediction vector obtained from the motion vectors of the surrounding macroblocks is used as the motion vector when actually encoding. Therefore, whether or not the difference between the two displacement amounts is larger than the threshold value and the estimated displacement amount is larger than the threshold value with the estimated displacement amount estimated in step S101 and the predicted vector as a determination target. This determination is made.

  When it is determined that the condition that the degree of deviation between the two displacement amounts is greater than the threshold value and the estimated displacement amount is greater than the threshold value according to the determination process in step S113, the process proceeds to step S114 and the register is registered. The prediction mode stored in M1 is output as the optimal prediction mode for the encoding target block, and the process is terminated.

  On the other hand, when it is determined that the condition that the degree of deviation between the two displacement amounts is greater than the threshold value and the estimated displacement amount is greater than the threshold value is not satisfied according to the determination process in step S113, the process proceeds to step S115. Then, the skip mode, which is the prediction mode stored in the register M0, is output as the optimum prediction mode for the encoding target block, and the process is terminated.

  Thus, in this embodiment, the prediction mode determination unit 11 determines that the prediction mode with the minimum RD cost obtained using the weighted distortion amount for the encoding target macroblock is the skip mode. However, when it is determined that coding in the skip mode is likely to include large image quality deterioration, the prediction mode with the minimum RD cost obtained using the unweighted distortion amount is optimal. It is processed so as to be determined as a proper prediction mode.

  FIG. 4 illustrates an example of a device configuration of the prediction mode determination unit 11 that realizes the present embodiment.

  Here, 100 is a displacement amount storage unit, 101 is a prediction vector calculation unit, 102 is a prediction vector storage unit, 103 is an initial mode setting unit, 104 is a mode storage unit, 105 is a code amount calculation unit, and 106 is a code amount storage unit 107 is a weighted distortion amount calculation unit, 108 is a weighted distortion amount storage unit, 109 is an unweighted distortion amount calculation unit, 110 is an unweighted distortion amount storage unit, 111 is an undetermined multiplier calculation unit, and 112 is an undetermined multiplier storage unit. 113 is a cost calculation unit, 114 is a cost storage unit, 115 is an unweighted cost calculation unit, 116 is an unweighted cost storage unit, 117 is a minimum cost determination unit, 118 is a minimum cost storage unit, and 119 is an unweighted minimum cost storage. , 120 is a mode update unit, 121 is an optimal mode storage unit, 122 is an unweighted optimal mode storage unit, 123 is a final mode determination unit, 124 is a mode setting unit, 25 is the optimal mode output unit.

  Next, each of these processing units will be described.

[Displacement storage unit 100]
The displacement amount storage unit 100 stores the estimated displacement amount for the encoding target macroblock calculated by the estimated displacement amount calculation unit (not shown). As an estimation method of the estimated displacement amount, for example, H.264. It is also possible to use a motion vector calculated by the H.264 reference software JM as an estimated value of the shift amount. Alternatively, it is also possible to calculate the estimated displacement amount according to a standard for minimizing the sum of absolute value errors between the encoding target macroblock and the reference macroblock.

[Predicted vector calculation unit 101]
Prediction vector calculation section 101 receives the motion vector of the macroblock adjacent to the encoding target macroblock and the macroblock division information as input, calculates a prediction vector for the encoding target macroblock, and writes it to prediction vector storage section 102. The specific calculation method is as follows. H.264 is followed.

[Initial mode setting unit 103]
The initial mode setting unit 103 writes the initial value of the prediction mode in the mode storage unit 104.

[Code amount calculation unit 105]
The code amount calculation unit 105 receives the prediction mode, the prediction vector, the quantization parameter, the encoding target frame signal, and the reference frame signal stored in the mode storage unit 104, and encodes using the prediction mode. And the calculated value is written in the code amount storage unit 106. The specific calculation method is as follows. It follows the method of H.264 reference software JM.

[Weighted distortion amount calculation unit 107]
The weighted distortion amount calculation unit 107 determines the weight based on the estimated displacement amount stored in the displacement amount storage unit 100, and stores the prediction mode, prediction vector, quantization parameter, and encoding target stored in the mode storage unit 104. Calculates the weighted distortion amount when encoding using the prediction mode based on the input signal and the determined weight, with the frame signal and the reference frame signal as input, and the calculated value is weighted. Write to the distortion amount storage unit 108. A specific calculation method follows the above-described formula (D) (the formula shown in [Equation 9]).

[Unweighted distortion amount calculation unit 109]
The unweighted distortion amount calculation unit 109 receives the prediction mode, the prediction vector, the quantization parameter, the encoding target frame signal, and the reference frame signal stored in the mode storage unit 104 and performs encoding using the prediction mode. The unweighted distortion amount is calculated, and the calculated value is written in the unweighted distortion amount storage unit 110. The specific calculation method is based on the above formula (D) (the formula shown in [Equation 9]) with no weighting, or the above formula (B) (shown in [Equation 2]). Follow the formula.

[Undetermined multiplier calculation unit 111]
The undetermined multiplier calculation unit 111 receives the prediction mode and quantization parameter stored in the mode storage unit 104, calculates an undetermined multiplier when encoding using the prediction mode, and calculates the calculated value to the undetermined multiplier storage unit. Write to 112. The specific calculation method is as follows. It follows the method of H.264 reference software JM.

[Cost calculation unit 113]
The cost calculation unit 113 reads the code amount stored in the code amount storage unit 106, the weighted distortion amount stored in the weighted distortion amount storage unit 108, and the undetermined multiplier stored in the undetermined multiplier storage unit 112. Then, the RD cost is calculated, and the calculated value is written in the cost storage unit 114. A specific calculation method follows the above-described formula (C) (the formula shown in [Equation 8]).

[Unweighted cost calculation unit 115]
The weightless cost calculation unit 115 includes a code amount stored in the code amount storage unit 106, an unweighted distortion amount stored in the unweighted distortion amount storage unit 110, and an undetermined multiplier stored in the undetermined multiplier storage unit 112. , The RD cost is calculated, and the calculated value is written in the unweighted cost storage unit 116. The specific calculation method is based on the above-described equation (C) (the equation shown in [Equation 8]) using a distortion-free value, or the above-described equation (A) ([Equation 1]). Follow the formula shown in

[Minimum cost determination unit 117]
The minimum cost determination unit 117 reads out the RD cost stored in the cost storage unit 114 and the minimum cost stored in the minimum cost storage unit 118, and whether the RD cost is smaller than the minimum cost. If the RD cost is smaller than the minimum cost, the RD cost is written to the minimum cost storage unit 118 and control is passed to the mode update unit 120. On the other hand, if the RD cost is greater than the minimum cost, control is passed to the final mode determination unit 123.

  Then, the RD cost stored in the non-weighted cost storage unit 116 and the minimum cost stored in the non-weighted minimum cost storage unit 119 are read, and whether or not the RD cost is smaller than the minimum cost. If the RD cost is smaller than the minimum cost, the RD cost is written to the weightless minimum cost storage unit 119 and control is passed to the mode update unit 120. On the other hand, if the RD cost is greater than the minimum cost, control is passed to the final mode determination unit 123.

[Mode update unit 120]
When the minimum cost determination unit 117 writes the RD cost to the minimum cost storage unit 118, the mode update unit 120 predicts the prediction mode that is the calculation source of the RD cost (stored in the mode storage unit 104 at that time). Is written to the optimum mode storage unit 121, and control is passed to the final mode determination unit 123.

  Then, when the minimum cost determination unit 117 writes the RD cost to the weightless minimum cost storage unit 119, the prediction mode that is the calculation source of the RD cost (stored in the mode storage unit 104 at that time) Is written to the optimum mode storage unit 122 without weight, and then the control is passed to the final mode determination unit 123.

[Final mode determination unit 123]
When control is passed from the minimum cost determination unit 117 or the mode update unit 120, the final mode determination unit 123 determines whether or not the mode setting unit 124 has finished setting all prediction modes, and determines all prediction modes. When determining that the setting of the prediction mode has been completed, the optimal mode output unit 125 is instructed to output the optimum prediction mode. When determining that all the prediction modes have not been set, the mode setting unit 124 is selected. Is instructed to set the next prediction mode.

[Mode setting unit 124]
When there is a prediction mode setting instruction from the final mode determination unit 123, the mode setting unit 124 sets the next prediction mode in the mode storage unit 104.

[Optimum mode output unit 125]
The optimal mode output unit 125 determines whether or not the prediction mode stored in the optimal mode storage unit 121 is the skip mode when there is an instruction to output the optimal prediction mode from the final mode determination unit 123. Otherwise, the prediction mode stored in the optimum mode storage unit 121 is output as the optimum prediction mode of the encoding target block.

  On the other hand, when the prediction mode stored in the optimal mode storage unit 121 is the skip mode, it is determined whether or not the prediction mode stored in the unweighted optimal mode storage unit 122 is the skip mode. If it is, the skip mode stored in the optimum mode storage unit 121 is output as the optimum prediction mode of the encoding target block.

  On the other hand, when the prediction mode stored in the optimal mode storage unit 121 is the skip mode and the prediction mode stored in the unweighted optimal mode storage unit 122 is not the skip mode, the estimated displacement stored in the displacement amount storage unit 100 The amount and the prediction vector stored in the prediction vector storage unit 102 are read out, and the degree of deviation between the estimated displacement amount and the prediction vector is calculated, and the magnitude of the displacement estimation amount is calculated, and the degree of deviation is calculated. Is larger than the threshold value, and it is determined whether or not the magnitude of the transition estimation amount is larger than the threshold value. Then, based on the determination result, when the condition that the degree of deviation between the estimated displacement amount and the prediction vector is large and the displacement estimation amount is large is satisfied, the prediction stored in the unweighted optimum mode storage unit 122 The mode is output as the optimal prediction mode of the encoding target block, and when the condition is not satisfied, the skip mode stored in the optimal mode storage unit 121 is output as the optimal prediction mode of the encoding target block.

  According to the configuration shown in FIG. 4, the prediction mode determination unit 11 performs the processing of the flowchart of FIG. 3 to minimize the RD cost obtained using the weighted distortion amount for the encoding target macroblock. Even if the prediction mode is the skip mode, when it is determined that there is a high possibility that the image is encoded in the skip mode, the image quality is greatly deteriorated. -D The processing is performed so as to determine the prediction mode with the lowest cost as the optimum prediction mode of the encoding target block.

[2] Second Embodiment FIG. 5 illustrates another embodiment of the flowchart executed by the prediction mode determination unit 11.

  Next, according to this flowchart, the prediction mode determination process executed by the prediction mode determination unit 11 in the present embodiment will be described in detail.

  When a prediction mode determination request is issued by designating a quantization parameter for a macroblock to be encoded and a prediction mode determination request is issued, first, as shown in the flowchart of FIG. Is set to the initial value of the prediction mode (prediction mode identification information to be an initial value), and the three registers C0, C1, and C2 that store the RD cost have large values. By storing an insignificant value for the three registers M0, M1, and M2 that store the initial cost and the identification information of the prediction mode, the registers X, C0 to C2, and M0 to M2 are stored. initialize.

  As will be described below, in addition to these registers X, C0 to C2, and M0 to M2, a register α that stores a code amount, a register β that stores an undetermined multiplier, and a weighted distortion amount are stored. Four registers are used: a register γ0 and a register γ1 that stores an unweighted distortion amount.

  Subsequently, in step S201, the shift amount of the encoding target macroblock is estimated. This estimation method is assumed to be given from the outside. For example, H.M. It is also possible to use a motion vector calculated by the H.264 reference software JM as an estimated value of the shift amount used below. Alternatively, it is also possible to calculate the estimated displacement amount according to a standard for minimizing the sum of absolute value errors between the encoding target macroblock and the reference macroblock.

  Subsequently, in step S202, the prediction mode, the prediction vector, the quantization parameter, the encoding target frame signal, and the reference frame signal of the register X are input, and the code amount when encoding using the prediction mode is calculated. The calculated value is written to the register α. The specific calculation method is as follows. It follows the method of H.264 reference software JM.

  Subsequently, in step S203, the prediction mode of the register X and the quantization parameter are input, an undetermined multiplier for encoding using the prediction mode is calculated, and the calculated value is written to the register β. The specific calculation method is as follows. It follows the method of H.264 reference software JM.

  Subsequently, in step S204, a weight is first determined based on the estimated displacement calculated in step S201, and then the prediction mode, prediction vector, quantization parameter, encoding target frame signal, reference frame of the register X are determined. Based on the input signal and the determined weight, the weighted distortion amount when encoding using the prediction mode is calculated, and the calculated value is written to the register γ0. A specific calculation method follows the above-described formula (D) (the formula shown in [Equation 9]).

  Subsequently, in step S205, the code amount stored in the register α, the undetermined multiplier stored in the register β, and the weighted distortion amount stored in the register γ0 are read, and the RD cost is calculated. A specific calculation method follows the above-described formula (C) (the formula shown in [Equation 8]).

  Subsequently, in step S206, when the calculated RD cost is compared with the value of the register C0 and it is determined that the calculated RD cost is smaller than the value of the register C0, step S207 is performed. Then, the calculated RD cost is stored in the register C0, and in step S208, the prediction mode identification information stored in the register X is stored in the register M0.

  On the other hand, when it is determined in step S206 that the RD cost calculated in step S205 is larger than the value in the register C0, the process proceeds to step S209, where the calculated RD cost and the register C2 are registered. When it is determined that the calculated RD cost is smaller than the value of the register C2, the process proceeds to step S210, and the calculated RD cost is stored in the register C2. In subsequent step S211, the prediction mode identification information stored in the register X is stored in the register M1. On the other hand, when it is determined in step S209 that the calculated RD cost is higher, the processes in steps S210 and 211 are omitted.

  In this way, the register C0 stores the minimum cost that has been obtained so far, and correspondingly, the register M0 stores prediction mode identification information that realizes the minimum cost. In addition, the register C2 stores a small cost following the required minimum cost of the processing so far, and correspondingly, the register M0 has a prediction mode for realizing the small cost following the minimum cost. The identification information is stored.

  The processes in steps S204x to S208x are executed in parallel with the processes in steps S204 to S211.

  That is, in step S204x, the prediction mode, the prediction vector, the quantization parameter, the encoding target frame signal, and the reference frame signal of the register X are input, and the unweighted distortion amount when encoding using the prediction mode is calculated. The calculated value is written in the register γ1. The specific calculation method is based on the above formula (D) (the formula shown in [Equation 9]) with no weighting, or the above formula (B) (shown in [Equation 2]). Follow the formula.

  Subsequently, in step S205x, the code amount stored in the register α, the undetermined multiplier stored in the register β, and the unweighted distortion amount stored in the register γ1 are read to calculate the RD cost. The specific calculation method is based on the above-described equation (C) (the equation shown in [Equation 8]) using a distortion-free value, or the above-described equation (A) ([Equation 1]). Follow the formula shown in

  Subsequently, in step S206x, when the calculated RD cost is compared with the value of the register C1, and it is determined that the calculated RD cost is smaller than the value of the register C1, step S207x Then, the calculated RD cost is stored in the register C1, and the identification information of the prediction mode stored in the register X is stored in the register M1 in the subsequent step S208x. On the other hand, when it is determined in step S206x that the calculated RD cost is larger than the value of the register C1, the processes in steps S207x and 208x are omitted.

  When the processing of step S208, step S208x, and step S211 is completed, then, in step S212, it is determined whether or not all prediction modes have been processed, and it is determined that not all prediction modes have been processed. Then, the process proceeds to step S213, one prediction mode is selected from the unprocessed prediction modes according to a predetermined order, the identification information of the selected prediction mode is stored in the register X, and the process proceeds to step S202. Return.

  In this manner, by repeating the processing of step S202 to step S213, when it is determined in step S212 that all prediction modes have been processed, the process proceeds to step S214, and the prediction mode stored in the register M0 is determined. It is determined whether or not the identification information indicates a skip mode.

  That is, it is determined whether or not the prediction mode with the minimum RD cost evaluated using the weighted distortion amount is the skip mode.

  When it is determined that the prediction mode identification information stored in the register M0 is not the skip mode according to the determination process in step S214, the image quality deterioration due to the skip mode does not matter, so the process proceeds to step S218, and the register M0 is stored. The stored prediction mode is output as the optimum prediction mode for the encoding target block, and the process is terminated.

  On the other hand, when it is determined that the prediction mode identification information stored in the register M0 is the skip mode according to the determination process in step S214, the process proceeds to step S215, and the prediction mode identification information stored in the register M1 is skipped. It is determined whether or not the mode is indicated.

  That is, it is determined whether or not the prediction mode with the minimum RD cost evaluated using the unweighted distortion amount is the skip mode.

  When it is determined that the prediction mode identification information stored in the register M1 is the skip mode according to the determination process in step S215, the skip mode is set regardless of whether the weighted distortion amount or the unweighted distortion amount is used. Since it has been selected, the process proceeds to step S218, the skip mode, which is the prediction mode stored in the register M0, is output as the optimum prediction mode for the block to be encoded, and the process is terminated.

  On the other hand, when it is determined that the prediction mode identification information stored in the register M1 is not the skip mode according to the determination process in step S215, the process proceeds to step S216, and the estimated displacement amount and the encoding target estimated in step S201 are processed. It is determined whether the degree of deviation from the predicted vector of the block is greater than a threshold and whether the estimated displacement is greater than the threshold.

  That is, since it is determined in step S214 that the prediction mode with the minimum RD cost is the skip mode, the prediction vector obtained from the motion vectors of the surrounding macroblocks is used as the motion vector when actually encoding. Therefore, whether or not the difference between the two displacement amounts is larger than the threshold value and the estimated displacement amount is larger than the threshold value with the estimated displacement amount estimated in step S201 and the predicted vector as a determination target. This determination is made.

  When it is determined that the condition that the degree of deviation between the two displacement amounts is greater than the threshold value and the estimated displacement amount is greater than the threshold value according to the determination processing in step S216, the process proceeds to step S217, where The prediction mode stored in M2 is output as the optimal prediction mode for the encoding target block, and the process is terminated.

  On the other hand, when it is determined that the condition that the degree of deviation between the two displacement amounts is greater than the threshold value and the estimated displacement amount is greater than the threshold value is not satisfied according to the determination process in step S216, the process proceeds to step S218. Then, the skip mode, which is the prediction mode stored in the register M0, is output as the optimum prediction mode for the encoding target block, and the process is terminated.

  Thus, in this embodiment, the prediction mode determination unit 11 determines that the prediction mode with the minimum RD cost obtained using the weighted distortion amount for the encoding target macroblock is the skip mode. However, when it is determined that coding in the skip mode is likely to include large image quality degradation, the prediction mode with the next smallest RD cost is determined as the optimum prediction mode. .

  FIG. 6 illustrates an example of a device configuration of the prediction mode determination unit 11 that realizes the present embodiment.

  The difference between the device configuration shown in FIG. 6 and the device configuration shown in FIG. 4 is that the configuration shown in FIG. A minimum cost determination unit 117α that executes processing different from the minimum cost determination unit 117, a mode update unit 120α that executes processing different from the mode update unit 120 shown in FIG. 4, and an optimum mode output unit 125 shown in FIG. Is provided with an optimum mode output unit 125α that executes different processes.

  Next, each of these processing units will be described. The displacement amount storage unit 100, the prediction vector calculation unit 101, the initial mode setting unit 103, the code amount calculation unit 105, the weighted distortion amount calculation unit 107, and the unweighted distortion amount calculation. The processing of unit 109, undetermined multiplier calculation unit 111, cost calculation unit 113, and unweighted cost calculation unit 115 is the same as that described with reference to FIG.

[Minimum cost determination unit 117α]
The minimum cost determination unit 117α reads out the RD cost stored in the cost storage unit 114 and the minimum cost stored in the minimum cost storage unit 118, and whether the RD cost is smaller than the minimum cost. If the RD cost is smaller than the minimum cost, the RD cost is written to the minimum cost storage unit 118 and control is passed to the mode update unit 120α.

  On the other hand, when the RD cost is larger than the minimum cost, the RD cost stored in the cost storage unit 114 and the quasi-minimum cost stored in the quasi-minimum cost storage unit 200 (second smallest) Cost), and it is determined whether or not the RD cost is smaller than the quasi-minimum cost. If the RD cost is smaller than the quasi-minimum cost, the RD cost is read. Is written to the quasi-minimum cost storage unit 200 and control is passed to the mode update unit 120α. On the other hand, if the RD cost is greater than the quasi-minimum cost, control is passed to the final mode determination unit 123.

  Then, the RD cost stored in the non-weighted cost storage unit 116 and the minimum cost stored in the non-weighted minimum cost storage unit 119 are read, and whether or not the RD cost is smaller than the minimum cost. If the RD cost is smaller than the minimum cost, the RD cost is written to the weightless minimum cost storage unit 119 and control is passed to the mode update unit 120α. On the other hand, if the RD cost is greater than the minimum cost, control is passed to the final mode determination unit 123.

[Mode update unit 120α]
When the minimum cost determination unit 117α writes the RD cost to the minimum cost storage unit 118, the mode update unit 120α stores the prediction mode that is the calculation source of the RD cost (stored in the mode storage unit 104 at that time). Is written to the optimum mode storage unit 121, and control is passed to the final mode determination unit 123.

  Then, when the minimum cost determination unit 117α writes the RD cost to the quasi-minimum cost storage unit 200, the prediction mode that is the calculation source of the RD cost (stored in the mode storage unit 104 at that time) After the identification information of (prediction mode) is written in the suboptimal mode storage unit 201, control is passed to the final mode determination unit 123.

  When the minimum cost determination unit 117α writes the RD cost to the unweighted minimum cost storage unit 119, the prediction mode that is the calculation source of the RD cost (stored in the mode storage unit 104 at that time) Is written to the optimum mode storage unit 122 without weight, and then the control is passed to the final mode determination unit 123.

[Final mode determination unit 123]
When the control is passed from the minimum cost determination unit 117α or the mode update unit 120α, the final mode determination unit 123 determines whether or not the mode setting unit 124 has finished setting all prediction modes, and determines all prediction modes. When determining that the setting of the prediction mode has been completed, the optimal mode output unit 125α is instructed to output the optimal prediction mode. When determining that all the prediction modes have not been set, the mode setting unit 124 Is instructed to set the next prediction mode.

[Mode setting unit 124]
When there is a prediction mode setting instruction from the final mode determination unit 123, the mode setting unit 124 sets the next prediction mode in the mode storage unit 104.

[Optimum mode output unit 125α]
When there is an instruction to output the optimal mode from the final mode determination unit 123, the optimal mode output unit 125α determines whether or not the prediction mode stored in the optimal mode storage unit 121 is the skip mode. In this case, the prediction mode stored in the optimum mode storage unit 121 is output as the optimum prediction mode of the encoding target block.

  On the other hand, when the prediction mode stored in the optimal mode storage unit 121 is the skip mode, it is determined whether or not the prediction mode stored in the unweighted optimal mode storage unit 122 is the skip mode. If it is, the skip mode stored in the optimum mode storage unit 121 is output as the optimum prediction mode of the encoding target block.

  On the other hand, when the prediction mode stored in the optimal mode storage unit 121 is the skip mode and the prediction mode stored in the unweighted optimal mode storage unit 122 is not the skip mode, the estimated displacement stored in the displacement amount storage unit 100 The amount and the prediction vector stored in the prediction vector storage unit 102 are read out, and the degree of deviation between the estimated displacement amount and the prediction vector is calculated, and the magnitude of the displacement estimation amount is calculated, and the degree of deviation is calculated. Is larger than a threshold value, and it is determined whether or not the magnitude of the transition estimation amount is larger than the threshold value. Then, based on the determination result, when the condition that the deviation between the estimated displacement amount and the prediction vector is large and the displacement estimation amount is large is satisfied, the prediction mode stored in the suboptimal mode storage unit 201 Is output as the optimal prediction mode of the encoding target block, and when the condition is not satisfied, the skip mode stored in the optimal mode storage unit 121 is output as the optimal prediction mode of the encoding target block.

  In accordance with the configuration shown in FIG. 6, the prediction mode determination unit 11 executes the processing of the flowchart in FIG. 5 to minimize the RD cost obtained using the weighted distortion amount for the encoding target macroblock. Even if the prediction mode is the skip mode, when it is determined that there is a high possibility that the image is coded in the skip mode, the image quality is likely to contain a large deterioration in image quality. Processing is performed so as to determine the optimum prediction mode.

[3] Experimental results of experiments conducted to verify the effectiveness of the present invention In order to verify the effectiveness of the present invention, It was implemented in H.264 reference software JSMV (version 8.0.1) and compared with the default JSMV.

  The encoding target sequence used in this experiment is “Mobile & Calender”, “City” of size 352 × 288 [pixles]. Each sequence has a frame rate of 30 [fps]. The GOP structure consists of I and P pictures, and the I picture insertion interval is 15 frames. The quantization parameter is QP = 28 for both I and P pictures.

  The following table shows the comparison results of the code amounts obtained by this experiment.

  From this experimental result, it has been confirmed that the present invention realizes a code amount reduction of 1.15 to 4.14% with respect to JSMV. Thus, the effectiveness of the present invention could be verified. Note that it has been confirmed that there is no subjective difference in image quality between the decoded images of both methods obtained in this experiment.

  The present invention can be applied to moving picture coding having a skip mode instructing to use a prediction vector as a motion vector. By applying the present invention, skipping image quality deterioration in the skip mode can be avoided. It is possible to enjoy the maximum benefit of reducing the amount of codes.

It is an apparatus block diagram of the video coding apparatus which comprises this invention. It is a flowchart which an encoding parameter determination part performs. It is a flowchart which a prediction mode determination part performs. It is an apparatus block diagram of a prediction mode determination part. It is a flowchart which a prediction mode determination part performs. It is an apparatus block diagram of a prediction mode determination part.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Video coding apparatus 10 Encoding parameter determination part 11 Prediction mode determination part 12 Estimated displacement amount calculation part 20 Encoding part 100 Displacement amount memory | storage part 101 Prediction vector calculation part 102 Prediction vector memory | storage part 103 Initial mode setting part 104 Mode storage part 105 Code amount calculation unit 106 Code amount storage unit 107 Weighted distortion amount calculation unit 108 Weighted distortion amount storage unit 109 Unweighted distortion amount calculation unit 110 Unweighted distortion amount storage unit 111 Undetermined multiplier calculation unit 112 Undetermined multiplier storage unit 113 Cost Calculation unit 114 Cost storage unit 115 Weightless cost calculation unit 116 Weightless cost storage unit 117 Minimum cost determination unit 118 Minimum cost storage unit 119 Weightless minimum cost storage unit 120 Mode update unit 121 Optimal mode storage unit 122 Weightless optimal mode storage Part 123 Final mode Tough 124 mode setting unit 125 optimum mode output unit 200 sub-minimal cost storage unit 201 suboptimal mode storage unit

Claims (12)

An encoding parameter determination method for determining an encoding parameter used for image encoding for performing information compression by transform encoding and quantization for a prediction error signal obtained by intra-frame prediction and inter-frame prediction,
For a block to be encoded, a process of calculating an estimated displacement amount indicating temporal movement of an image signal prior to the encoding process;
A block whose coding mode is the skip prediction mode selected by the cost calculation using the weighted distortion amount using the sensitivity coefficient indicating the spatio-temporal visual sensitivity is used as a determination target, and the block is encoded. A process of determining whether or not a block is likely to have a high image quality degradation when encoded in the skip mode, based on the displacement amount used at the time and the estimated displacement amount;
For the block to be determined that is determined not to be the corresponding block, the skip mode is determined as the optimum prediction mode of the block, and for the block to be determined that is determined to be the corresponding block, the weighting is determined. Determining a least cost prediction mode selected by calculating a cost using an undistorted amount as an optimal prediction mode of the block,
A characteristic encoding parameter determination method. An encoding parameter determination method for determining an encoding parameter used for image encoding for performing information compression by transform encoding and quantization for a prediction error signal obtained by intra-frame prediction and inter-frame prediction,
For a block to be encoded, a process of calculating an estimated displacement amount indicating temporal movement of an image signal prior to the encoding process;
A block whose coding mode is the skip prediction mode selected by the cost calculation using the weighted distortion amount using the sensitivity coefficient indicating the spatio-temporal visual sensitivity is used as a determination target, and the block is encoded. A process of determining whether or not a block is likely to have a high image quality degradation when encoded in the skip mode, based on the displacement amount used at the time and the estimated displacement amount;
For the block to be determined that is not the corresponding block, the skip mode is determined as the optimum prediction mode of the block, and the cost minimum is determined for the block to be determined that the block is determined to be the corresponding block. Determining a prediction mode selected as the next lowest cost mode as the optimal prediction mode of the block,
A characteristic encoding parameter determination method. The encoding parameter determination method according to claim 1 or 2,
In the determining process, the displacement amount is compared with the estimated displacement amount, and the difference between the two displacement amounts is larger than a prescribed threshold value, and the estimated displacement amount is larger than the prescribed threshold value. In this case, it is determined that the block is likely to have a large image quality deterioration when encoded in the skip mode.
A characteristic encoding parameter determination method. In the encoding parameter determination method according to any one of claims 1 to 3,
For a block whose prediction mode with the lowest cost selected by calculating the cost using the weighted distortion amount is not the skip mode, the step of determining the prediction mode as the optimum prediction mode of the block,
A characteristic encoding parameter determination method. The encoding parameter determination method according to any one of claims 1 to 4,
In the case where the prediction mode with the minimum cost selected by calculating the cost using the unweighted distortion amount is also the skip mode for the determination target block, the skip mode is performed without performing the determination process. Comprising determining the optimal prediction mode for the block,
A characteristic encoding parameter determination method. An encoding parameter determination device that determines an encoding parameter used for image encoding for performing information compression by transform encoding and quantization for a prediction error signal obtained by intra-frame prediction and inter-frame prediction,
For an encoding target block, prior to encoding processing, means for calculating an estimated displacement amount indicating temporal movement of an image signal;
A block whose coding mode is the skip prediction mode selected by the cost calculation using the weighted distortion amount using the sensitivity coefficient indicating the spatio-temporal visual sensitivity is used as a determination target, and the block is encoded. Means for determining whether or not the block corresponds to a block that is highly likely to deteriorate image quality when encoded in the skip mode, based on the displacement amount used at the time and the estimated displacement amount;
For the block to be determined that is determined not to be the corresponding block, the skip mode is determined as the optimum prediction mode of the block, and for the block to be determined that is determined to be the corresponding block, the weighting is determined. Means for determining a prediction mode with the lowest cost selected by calculating a cost using an undistorted amount as an optimal prediction mode of the block,
An encoding parameter determination device as a feature. An encoding parameter determination device that determines an encoding parameter used for image encoding for performing information compression by transform encoding and quantization for a prediction error signal obtained by intra-frame prediction and inter-frame prediction,
For an encoding target block, prior to encoding processing, means for calculating an estimated displacement amount indicating temporal movement of an image signal;
A block whose coding mode is the skip prediction mode selected by the cost calculation using the weighted distortion amount using the sensitivity coefficient indicating the spatio-temporal visual sensitivity is used as a determination target, and the block is encoded. Means for determining whether or not the block corresponds to a block that is highly likely to deteriorate image quality when encoded in the skip mode, based on the displacement amount used at the time and the estimated displacement amount;
For the block to be determined that is not the corresponding block, the skip mode is determined as the optimum prediction mode of the block, and the cost minimum is determined for the block to be determined that the block is determined to be the corresponding block. Determining a prediction mode selected as the next lowest cost mode of the skip mode as an optimal prediction mode of the block,
An encoding parameter determination device as a feature. The encoding parameter determination apparatus according to claim 6 or 7,
The means for determining compares the displacement amount with the estimated displacement amount, and the difference between the two displacement amounts is greater than a prescribed threshold value, and the magnitude of the estimated displacement amount is greater than a prescribed threshold value. In this case, it is determined that the block is likely to have a large image quality deterioration when encoded in the skip mode.
An encoding parameter determination device as a feature. In the encoding parameter determination device according to any one of claims 6 to 8,
For a block in which the prediction mode with the lowest cost selected by calculating the cost using the weighted distortion amount is not the skip mode, means for determining the prediction mode as the optimum prediction mode of the block,
An encoding parameter determination device as a feature. The encoding parameter determination device according to any one of claims 6 to 9,
In the case where the prediction mode with the minimum cost selected by calculating the cost using the unweighted distortion amount is also the skip mode for the determination target block, the skip mode is performed without performing the determination process. Comprising means for determining as the optimal prediction mode for the block,
An encoding parameter determination device as a feature.   An encoding parameter determination program for causing a computer to execute processing used to realize the encoding parameter determination method according to any one of claims 1 to 5.   A computer-readable recording medium on which an encoding parameter determination program for causing a computer to execute processing used to realize the encoding parameter determination method according to any one of claims 1 to 5 is recorded.

Images