JP7068925B2 - Predictive coding method determination device, moving image coding device, and moving image decoding device, and program - Google Patents

Description

INDUSTRIAL APPLICABILITY The present invention is a moving image coding device that performs predictive coding processing of a moving image, and determines a predictive coding method within a predetermined group that divides the moving image before the predictive coding process into a plurality of screen groups. It relates to a method determining device, a moving image coding device, a moving image decoding device, and a program.

As a compression coding technique for moving images, for example, H.264 / MPEG-4 AVC, H.265 / MPEG-H HEVC, etc., a predetermined group divided into a plurality of screen groups for the moving image before predictive coding processing is set. , A technique is known in which an interscreen predictive coding process or an in-screen predictive coding process selected in advance as a predictive coding method is performed within the predetermined group to generate a bit stream for transmission.

In H.264 / MPEG-4 AVC, a GOP (Group of Pictures) is set as the predetermined group by dividing the moving image before prediction coding into a plurality of screen groups. Then, for each screen (frame or slice screen) in the GOP, an I picture that performs in-screen prediction coding processing, a P picture that performs inter-screen prediction coding processing from a past screen in time, or a time before and after The B picture to be subjected to the inter-screen prediction coding processing is selected in advance from the screen located at. In H.264 / MPEG-4 AVC, all screens in the GOP can be compressed by a compression coding method only for in-screen predictive coding processing.

Further, also in H.265 / MPEG-H HEVC, a sequence in which the moving image before prediction coding is divided into a plurality of screen groups is set as the predetermined group. Although the concept of GOP is not introduced in the HEVC standard, this sequence corresponds to GOP. Then, for each screen (frame screen, slice screen, or tile screen) in the sequence, a processing unit that performs in-screen prediction coding processing or inter-screen prediction coding processing is selected in advance (see, for example, Non-Patent Document 1). ). Also in H.265 / MPEG-H HEVC, it is possible to compress all the screens in the sequence by the compression coding method only for the in-screen predictive coding process.

By the way, as one of the frequency analysis in an image, for example, wavelet analysis and wavelet packet analysis are known (see, for example, Non-Patent Document 2).

Supervised by Ei Okubo, Teruhiko Suzuki, 2 outsiders, "H.265 / HEVC Textbook", published by Impress Japan Co., Ltd., first edition published on October 21, 2013 Kohei Arai, "Basic Theory of Wavelet Analysis", Morikita Publishing Co., Ltd., First Edition, November 2000

In the conventional moving image coding device, a fixed predictive coding method is used within the predetermined group without automatically determining what kind of feature amount the moving image to be coded has. It is often determined subjectively based on a combination of a plurality of types of predictive coding methods prepared in advance.

However, the inter-screen predictive coding process has a large effect of reducing the coded signal, but in general, when the inter-screen predictive coding process is applied to a critical moving image having complicated movements, the coding bit rate becomes higher. If not enough, artifacts such as block distortion and flicker are likely to occur. Therefore, in the selection of the fixed predictive coding method or the predictive coding method based on the subjective determination, it is not possible to expect the improvement of the moving image quality after the predictive decoding corresponding to the predictive coding process.

Therefore, in order to determine the predictive coding method within the predetermined group in the moving image coding apparatus that performs the predictive coding process of the moving image, a technique of automatically selecting from the viewpoint of image quality is desired.

Further, even if there is a technique for automatically selecting a predictive coding method within the predetermined group in the conventional technique, the signal value (pixel value, luminance signal, or brightness signal) of a unit block obtained by dividing the screen in the moving image to be encoded It is assumed that the feature quantity is the color difference signal, etc.) itself. However, even in this case, there is room for improvement in the discrimination accuracy. Therefore, a technique for selecting and determining a predictive coding method within the predetermined group with higher accuracy and automatically is desired.

Therefore, an object of the present invention is to determine a predictive coding method for automatically selecting and determining a predictive coding method in a predetermined group divided into a plurality of screen groups for a moving image before predictive coding processing with high accuracy from the viewpoint of image quality. It is an object of the present invention to provide an apparatus, a moving image coding apparatus, a moving image decoding apparatus, and a program.

The predictive coding method determining device of the present invention is a predictive coding method determining device that automatically selects and determines a predictive coding method in a predetermined group divided into a plurality of screen groups for a moving image before the predictive coding process. , The reference image and the determination image predetermined as the determination target of the prediction coding method are extracted from the predetermined group, and the reference image and the determination image are obtained by the p (p ≧ 2) order wavelet packet analysis process. Using the wavelet packet analysis unit that calculates the frequency analysis coefficient and the element value of each frequency band at an arbitrary phase position in the frequency analysis coefficient as feature quantities, screen-to-screen matching processing is performed for each of the reference image and the determination image. The matching processing unit that calculates the matching amount based on the predetermined similarity evaluation index, and the inter-screen predictive coding process or the in-screen predictive coding process as the predictive coding method within the predetermined group based on the matching amount. It is characterized by comprising a predictive coding method selection determination unit for selecting and determining whether or not to be used.

Further, in the predictive coding method determining apparatus of the present invention, the predetermined image is based on the matching amount in which the reference image is the first frame in the predetermined group and the determination image is one or more frames other than the first frame. It is characterized in that the predictive coding method of each frame in the group is sequentially selected and determined.

Further, in the predictive coding method determining apparatus of the present invention, the predetermined group is based on the matching amount in which the reference image is the first frame in the predetermined group and the determination image is one frame other than the first frame. It is characterized in that the predictive coding method of all the frames in the box is selected and determined at once.

Further, in the predictive coding method determination device of the present invention, the predictive coding method selection determination unit is premised on applying a predetermined basic predictive coding method to each screen in the predetermined group. It is characterized in that the prediction coding method within the predetermined group is determined by switching to apply a second prediction coding method different from the basic prediction coding method based on the matching amount.

Further, in the predictive coding method determination device of the present invention, the predictive coding method selection determination unit uses a predetermined basic predictive coding method including an interscreen predictive coding process for each screen in the predetermined group. On the premise of application, based on the matching amount, the prediction coding method in the predetermined group is determined by switching to apply only the in-screen prediction coding process to all the screens in the predetermined group. It is characterized by that.

Further, in the predictive coding method determining apparatus of the present invention, the matching processing unit sets a matching weight for each frequency analysis coefficient related to the inter-screen matching processing when calculating the matching amount, and has at least the number of floors 2. It is characterized in that the value of the matching weight in the frequency band is made larger than the value of the matching weight in the frequency band of the first floor, and the matching amount is calculated.

Further, in the predictive coding method determination device of the present invention, the matching processing unit sets a matching weight for each frequency analysis coefficient related to the inter-screen matching processing when calculating the matching amount, and sets the matching weight with the reference image. The larger the time distance from the determination image, the larger the matching weight, and the matching amount is calculated.

Further, the moving image coding apparatus of the present invention includes a group setting unit that inputs a moving image before the predictive coding process, divides the moving image into a plurality of screen groups, and sets a predetermined group divided into the plurality of screen groups. The predictive coding method determination device of the present invention that automatically selects and determines the predictive coding method in the predetermined group, and the prediction instructed by the predictive coding method determination device for a plurality of screen groups in the predetermined group. An inter-screen predictive coding process or an in-screen predictive coding process is performed based on the coding method, and the coded signal subjected to the predictive coding process is instructed by the predictive coding method determination device. A predictive coding unit that generates header information indicating a predictive coding method and a predetermined entropy coding process are applied to the coded signal, and the header information is written to a predetermined stream header to generate a bit stream. It is characterized by including a stream generation unit.

Further, the moving image decoding device of the present invention includes a stream input unit that inputs a bit stream from the moving image coding device of the present invention and extracts the header information and the coded signal from the bit stream in units of the predetermined group. Based on the header information in the predetermined group unit, the predictive coding method by the inter-screen predictive coding process or the in-screen predictive coding process is determined, and the predictive decoding process is performed on the coded signal to restore the coded signal. It is characterized by including a predictive decoding unit that generates a moving image.

Further, the program of the present invention is configured as a program for making the computer function as the predictive coding method determining device of the present invention.

According to the present invention, in the inter-screen predictive coding process, it is possible to suppress coding deterioration that tends to occur in a critical moving image and improve the total subjective image quality.

It is a block diagram which shows the schematic structure of the moving image coding apparatus which comprises the predictive coding method determination apparatus of one Embodiment by this invention. It is a flowchart which shows the process in the predictive coding method determination apparatus of one Embodiment by this invention. It is a figure which shows the frequency band obtained by the 2nd order wavelet packet analysis in the predictive coding method determination apparatus of one Embodiment by this invention. It is explanatory drawing of the interframe matching processing based on the frequency analysis coefficient by the matching processing part in the predictive coding method determination apparatus of one Embodiment by this invention. It is a block diagram which shows the schematic structure of the moving image decoding apparatus of one Embodiment by this invention. (A) is a diagram showing an image example when a signal encoded by interscreen predictive coding processing of a frame of a moving image is decoded, and (b) is an embodiment according to the present invention for the frame. When the in-screen predictive coding process is automatically selected by the predictive coding method determination device of the above, and the signal encoded by the moving image coding device according to the present invention is decoded by the moving image decoding device. It is a figure which shows the image example.

Hereinafter, the predictive coding method determining device 10 according to the present invention, the moving image coding device 1 provided with the predictive coding method determination device 10, and the moving image decoding device 2 will be described in detail with reference to the drawings.

[Motion image coding device]
FIG. 1 is a block diagram showing a schematic configuration of a moving image coding device 1 including the predictive coding method determining device 10 of the embodiment according to the present invention. The moving image coding device 1 is represented by, for example, H.264 / MPEG-4 AVC, H.265 / MPEG-H HEVC, etc., as predictive coding methods such as inter-screen predictive coding process and in-screen predictive coding process. The apparatus includes a predictive coding method determination device 10, a group setting unit 11, a predictive coding unit 12, and a stream generation unit 13.

The group setting unit 11 inputs a moving image before the prediction coding process, divides it into a plurality of screen groups, sets a predetermined group divided into the plurality of screen groups, and sets a plurality of screen groups in the predetermined group. It is output to the prediction coding unit 12. Here, the screen can be a frame screen, a slice screen, or a tile screen.

First, the predictive coding method determination device 10 extracts a plurality of predetermined input images (reference image and determination image) to be determined by the predictive coding method, which are included in the predetermined group. Subsequently, the predictive coding method determination device 10 uses the plurality of input images (reference image and determination image) as an interscreen predictive coding process or in-screen predictive coding for all frames in the predetermined group. Determine the predictive coding method by processing. Then, the predictive coding method determination device 10 instructs the predictive coding unit 12 of the determined predictive coding method.

Here, the reference image is a screen located at the head of the predetermined group, and the determination image is one or more screens other than the reference image in the predetermined group, preferably in the final frame in the predetermined group. It can be one screen.

The prediction coding unit 12 performs inter-screen prediction coding processing or inter-screen prediction coding processing based on the prediction coding method instructed by the prediction coding method determination device 10 for a plurality of screen groups in the predetermined group input from the group setting unit 11. Predictive coding processing by in-screen predictive coding processing is performed, and the coded signal to which the predictive coding processing has been performed is output to the stream generation unit 13. The predictive coding unit 12 also outputs header information indicating the predictive coding method instructed by the predictive coding method determination device 10 to the stream generation unit 13.

The stream generation unit 13 performs a predetermined entropy coding process (CAVLC, CABAC, etc.) on the coded signal obtained from the predictive coding unit 12, and uses the header information obtained from the predictive coding unit 12 as a GOP header or a sequence header. To generate a bit stream for writing to a predetermined stream header such as, etc., and transmitting to the moving image decoding device 2 described later.

[Predictive coding method determination device]
As shown in FIG. 1, the predictive coding method determination device 10 includes a wavelet packet analysis unit 101, a matching processing unit 102, and a predictive coding method selection determination unit 103. FIG. 2 is a flowchart showing processing in the predictive coding method determination device 10, and the predictive coding method determination device 10 shown in FIG. 1 will be described in detail with reference to FIG. 2.

First, the wavelet packet analysis unit 101 extracts a plurality of predetermined input images (reference image and determination image) to be determined by the predictive coding method, which are included in the predetermined group of moving images before the coding process. .. Then, the wavelet packet analysis unit 101 uses Parseval's theorem for each of the extracted plurality of input images (reference image and determination image) based on the externally set wavelet packet analysis order p (p is an integer of 2 or more). The p-th order wavelet packet analysis process is executed so as to satisfy (step S1). As a result, stable high-speed processing is realized by compensating for the phase delay. The wavelet filter used for the p-th-order wavelet packet analysis processing may be any wavelet filter that satisfies Parseval's theorem, and may be, for example, a Haar filter having linear phase and capable of obtaining a difference feature amount.

Here, in order to simplify understanding, it is assumed that p = 2, the predetermined group is divided into frame units (8 divisions in this example), and the reference image is the first frame (time t) in the predetermined group. = T ₁ ), and the determination image is the final frame (time t = t ₈ ).

(Wavelet packet analysis unit)
The wavelet packet analysis unit 101 calculates frequency analysis coefficients C ^{m and n} (t) for each of the reference image and the determination image by the p-th order wavelet packet analysis processing, and outputs the frequency analysis coefficients to the matching processing unit 102.

For example, FIG. 3 shows the frequency analysis coefficients C ^{m, n} (t) obtained by the second-order wavelet packet analysis processing when p = 2. Note that FIG. 3 is an example in which a frame image of a so-called 8K moving image captured by a broadcasting camera (horizontal resolution 8K (7680 pixels) × vertical resolution 4K (4320 pixels)) is used as an input image. By performing the second-order wavelet packet analysis processing on the image in the horizontal and vertical directions, respectively, the image is decomposed into frequency analysis coefficients C ^{m and n} as shown in the horizontal frequency × vertical frequency. Here, m and n are horizontal and vertical band numbers of frequency analysis coefficients C ^{m and n} , respectively. The frequency analysis coefficients C ^{m and n} are composed of element values α ^{m, ni} and _j of each frequency band at an arbitrary phase position {i, j}. Here, i and j are horizontal and vertical phase positions within C ^{m and n} , respectively. Then, it is assumed that Parseval's law holds between the coefficient values of each floor during wavelet packet analysis processing.

In this way, by performing the p-th order wavelet packet analysis processing on each of the plurality of input images (reference image and determination image), each having element values α ^m, _{ni, j} (t) of each frequency band. The frequency analysis coefficients C ^{m, n} (t) can be obtained.

(Matching processing unit)
Next, the matching processing unit 102 sets the element values α ^{m, n} _i of each frequency analysis coefficient C ^{m, n} (t) in each of the plurality of input images (reference image and determination image) obtained from the wavelet packet analysis unit 101. Based on, _{and j} , the matching process for aligning between screens for alignment (interframes in this example) is performed. First, prior to the matching process, the matching processing unit 102 sets the matching weight b for each frequency analysis coefficient C ^{m, n} (t) related to the matching process between the screens (between frames in this example) by an external setting. Set ^{m and n} (step S2).

The setting values of the matching weights b ^{m and n} can always be b ^{m and n} = 1. However, for example, when the frequency analysis coefficient {C ³ , 3, C ⁴ , 3, C 3, ^{4, C 4} , ⁴ } of the frequency band of the second order is calculated through the second-order wavelet packet analysis processing by the Har filter, this is obtained. The frequency analysis coefficient of order 2 is an important second-order differential component for the matching process for selecting and determining the predictive coding method. Therefore, in the matching process of the frequency analysis coefficient of the frequency band of order 1, b ^{m, n} = 1, and in the matching process of the frequency analysis coefficient of the frequency band of order 2, the value of the matching weight is increased to b ^{m, n} >. It can be 1 (for example, b ^{m, n} = 1.7, etc.).

Further, the set values of the matching weights b ^{m and n} are the matching weights b ^{m and n} in the matching process of the frequency analysis coefficient of the band of the second rank even when the wavelet packet analysis process of the third or higher order is performed with p> 3. It is preferable that the value of is maximized as compared with other layers. For example, as the setting values of the matching weights b ^{m and n} , b ^{m and n} = 1 in the matching process related to the first floor, b ^{m and n} = 1.7 in the matching process related to the second floor, and matching related to the third floor. It is preferable to set b ^{m, n} = 0.8 in the processing, and b ^{m, n} = 0.7 in the matching processing related to the rank 4.

Further, as the set values of the matching weights b ^{m and n} , the larger the time distance between the reference image and the determination image is, the more the matching weight b ^m is set according to the time distance between the matching frames in the predetermined group. ^{, N} should be increased (increasing the effect). For example, when the predetermined group is divided into eight in frame units, the matching amount is determined between the first frame (time t = t ₁ ) and the subsequent second frame (time t = t ₂ ). Then, it is assumed that b ^{m and n} are multiplied by the coefficient a ₁ . On the other hand, when determining the matching amount between the first frame (time t = t ₁ ) and the last frame (time t = t ₈ ), it is assumed that b ^{m and n} are multiplied by the coefficient a ₈ . Then, a ₁ <a ₈ is set.

Subsequently, the matching processing unit 102 compares a plurality of input images (reference image and determination image) obtained from the wavelet packet analysis unit 101 with each other, and has a frequency analysis coefficient C ^{m, n} according to the matching weights b ^m, n. The matching amount is calculated by the matching process between screens (frames in this example) for positioning based on the element values α ^m, _{ni, and j} in each frequency standby (step S3).

More specifically, the matching processing unit 102 inputs the frequency analysis coefficients C ^{m and n} in each of the plurality of input images (reference image and determination image) and the matching weights b ^{m and n} set externally, and C. Matching processing between screens (frames in this example) is performed using the element values α ^{m, ni, j of arbitrary phase positions {i, j} in m, n} _as ^feature quantities, and the results are b ^{m, n.} The number of values multiplied by the value exceeding a predetermined threshold is counted as the matching amount.

By the way, the predictive coding method determination device 10 individually performs inter-screen predictive coding process or in-screen predictive coding process for each screen through matching process between screens (frames in this example) in the predetermined group. It can be configured to determine any of the predictive coding methods.

However, in order to speed up the processing of the predictive coding method determination device 10, for example, if GOP is taken as an example of the predetermined group, each picture having a GOP size of 8 and a GOP structure is {I, P, B, B from the beginning. , B, B, B, B, B} Random Access is used as the basic predictive coding method. Then, the predictive coding method determining device 10 is configured to switch the basic predictive coding method to a second predictive coding method different from the basic predictive coding method through the inter-frame matching process by the matching processing unit 102. Suitable.

For example, as the second predictive coding method, a compression coding method of only the in-screen predictive coding process can be used for all the screens in the predetermined group (for example, GOP).

Preferably, from the viewpoint of speeding up the processing, the predictive coding method determining device 10 uses the final frame farthest in the time direction from the first frame, which is the reference image, as the determination image, and the first frame and the final frame. It is preferable to switch the basic predictive coding method for all frames in the predetermined group to a second predictive coding method different from this only by the inter-frame matching process.

Therefore, in this example, when the degree of similarity between the first frame and the final frame having the longest distance in the time direction is high, the predictive coding method determining device 10 uses the basic predictive coding method as the first frame. If the similarity between and the final frame is low, it is switched as the second predictive coding method.

Further, the predictive coding method determining device 10 according to the present invention does not feature the signal value (pixel value, luminance signal, color difference signal, etc.) of a unit block itself as a feature quantity, but has a plurality of dimensions with a layer p ≧ 2. Since the inter-screen matching process is based on the frequency analysis coefficients C ^{m, n} (t) with the element values α ^m, _{ni, j} (t) of the frequency analysis coefficients C ^{m, n} (t) as feature quantities. The accuracy of similarity judgment is high.

FIG. 4 is an explanatory diagram of inter-frame matching processing based on the frequency analysis coefficient by the matching processing unit 102 in the prediction coding method determination device 10. The reference image is the first frame (time t = t ₁ ) in the predetermined group, and the determination image is the last frame (time t = t ₈ ). The phase shift amount from an arbitrary phase position {i, j} is {x, y}, and for example, SAD (Sum of Absolute Difference) is used as the similarity evaluation index of the matching process, and the difference value by the SAD evaluation is evaluated. Let _{di, j} (t). In addition to SAD, SSD (Sum of Squared Difference), NCC (Normalized Cross-Correlation), ZNCC (Zero-means Normalized Cross-Correlation), and the like can be used as the similarity evaluation index.

As shown in FIG. 4, the matching processing unit 102 of this example has an arbitrary frequency analysis coefficient C ^{m, n} (t ₁ ) for the first frame and an arbitrary frequency analysis coefficient C ^{m, n} (t ₈ ) for the final frame. Matching processing is performed between the screens (frames in this example) of the element values α ^m, _{ni, j} (t ₁ ) and α ^m, _{ni, j} (t ₈ ) of the phase position {i, j} of.

The evaluation values di _{and j} (t) of the matching points by this inter-frame matching process are calculated from the following equation (1).

That is, the matching processing unit 102 of this example sets the matching points in the frequency analysis coefficients C ^{m, n} (t ₈ ) of the final frame with respect to the element values α ^m, _{ni, j} (t ₁ ) of the first frame by the equation ( As shown on the right side of 1)
Σ ⁴ _{m = 1} Σ ⁴ _{n = 1} ｜ α ^{m, n} _{i, j} (t ₁ )-α ^{m, n} _{i + x, j + y} (t ₈ ) ｜ b ^{m, n}
Can be determined as α ^m, _{ni + x, j + y} (t ₈ ), which is the phase movement amount {x, y} that minimizes.

Then, the matching processing unit 102 sets the threshold value to τ based on the count function Count shown in the following equation (2), and the number of evaluation values di _{, j} (t) in ∀ {i, j} is τ or less. Calculate the number (t). Here, τ is, for example, 32.

When the matching processing unit 102 calculates the number num (t) at which the evaluation values di _{and j} (t) are equal to or less than the threshold value τ by the equation (2), the num (t) is matched between frames in the predetermined group. It is output to the prediction coding method selection determination unit 103 as the matching amount of the processing. If the range of the phase movement amount {x, y} is adjusted to the range of the motion search range related to the inter-screen (inter-frame in this example) prediction of the moving image coding, it can be further adjusted. High-speed processing is possible.

(Predictive coding method selection determination unit)
The prediction coding method selection determination unit 103 inputs the matching amount num (t) in the predetermined group (for example, GOP) from the matching processing unit 102, and based on the number of the matching amount num (t), the predetermined group. As the predictive coding method in the screen, it is selected and determined whether to use the interscreen predictive coding process or the in-screen predictive coding process, and the predictive coding unit 12 is instructed to perform the determined predictive coding method (step). S4).

More specifically, the prediction coding method selection determination unit 103 performs inter-screen prediction coding processing if the number of matching amounts num (t) is equal to or greater than the second threshold value, and in-screen if the number is less than the second threshold value. Select and determine as predictive coding processing.

In the predetermined group (for example, GOP), the matching process is performed by comparing only the first frame (t ₁ ) which is the reference image and the final frame (t ₈ ) which is the farthest in the time direction from the first frame (t 1). In addition to this, matching processing can be performed on more frames within the predetermined group (for example, GOP).

Then, even when matching processing is performed on more frames in the predetermined group (for example, GOP), when even one selection determination is made as the inter-screen prediction coding processing, the prediction coding method selection determination unit 103 , It is preferable to select and determine all the frames in the predetermined group (for example, GOP) as the in-screen prediction coding process. As a result, the subjective image quality is improved and the processing speed is also improved.

Further, the plurality of input images (reference image and determination image) processed by the predictive coding method determination device 10 can be set to each RGB color space cut out from the moving image before the predictive coding process. However, it can be an image that has been grayed out or converted into a luminance signal. Then, in the case of determining the similarity between the reference image and the determination image for each RGB color space, when selecting and determining even one of the color spaces as the inter-screen prediction coding process, the prediction coding method selection determination unit It is preferable that the 103 selects and determines all the frames in the predetermined group (for example, GOP) as the in-screen predictive coding process. As a result, the subjective image quality is improved and the processing speed is also improved.

As described above, the predictive coding method determining device 10 has a multidimensional frequency analysis coefficient C ^{m, n} (t) in which the basic predictive coding method for the predetermined group is predetermined and the layer p ≧ 2. The basic predictive coding method based on the matching process based on the frequency analysis coefficients C ^{m, n} (t) in the reference image and the judgment image of 1 or more, with the element values α ^m, _{ni, j} (t) as feature quantities. Is preferably switched to a second predictive coding method different from this. In particular, it is preferable that the basic predictive coding method includes inter-screen predictive coding processing, and the second predictive coding method includes only in-screen predictive coding processing.

As a result, the predictive coding method determining device 10 estimates the difficulty level of the inter-screen predictive coding in the previous stage of the predictive coding unit 12 of the moving image, and if the difficulty level is high, the prediction coding method determination device 10 is within the predetermined group (for example, GOP). Then, it is possible to switch to the in-screen prediction coding process.

Then, the predictive coding unit 12 in the moving image coding device 1 indicates a predictive coding method instructed by the predictive coding method determining device 10 for a plurality of screen groups in the predetermined group input from the group setting unit 11. The predictive coding process based on the above is performed, and the coded signal subjected to the predictive coding process and the header information indicating the predictive coding method are output to the stream generation unit 13.

The stream generation unit 13 in the moving image coding device 1 performs a predetermined entropy coding process (CAVLC, CABAC, etc.) on the coded signal obtained from the predictive coding unit 12, and the header obtained from the predictive coding unit 12. Information is written in a predetermined stream header such as a GOP header or a sequence header, and a bit stream for transmission to the moving image decoding device 2 described later is generated.

[Motion image coding device]
FIG. 5 is a block diagram showing a schematic configuration of a moving image decoding device 2 according to an embodiment of the present invention. The moving image decoding device 2 includes a stream input unit 21 and a predictive decoding unit 22.

The stream input unit 21 inputs a bit stream transmitted from the moving image coding device 1, extracts the header information and the coded signal from the bit stream in units of the predetermined group, and outputs the coded signal to the predictive decoding unit 22.

The predictive decoding unit 22 determines a predictive coding method by inter-screen predictive coding processing or in-screen predictive coding processing based on the header information in the predetermined group unit, and predictive decoding processing for the coded signal. To generate a restored moving image and output it to the outside.

As a result, it is possible to suppress the coding deterioration that tends to occur in a critical moving image in the inter-screen prediction coding process such as having a complicated movement, and it is possible to improve the total subjective image quality.

FIG. 6A is a diagram showing an image example when a signal encoded by the inter-screen prediction coding process of a frame of a certain moving image is decoded. On the other hand, in FIG. 6B, the in-screen predictive coding process is automatically selected for the frame by the predictive coding method determining device 10 of the embodiment according to the present invention, and the moving image coding device 1 according to the present invention automatically selects the in-screen predictive coding process. It is a figure which shows the image example when the coded signal is decoded by the moving image decoding apparatus 2.

As shown in FIG. 6 (b), which is shown in comparison with FIG. 6 (a), the coding deterioration that tends to occur in a critical moving image is suppressed in the inter-screen predictive coding process such as having a complicated movement. You can see that.

The predictive coding method determination device 10 and the moving image coding device 1 and the moving image decoding device 2 in the above embodiments can be configured by a computer, respectively, and each processing unit of the predictive coding method determining device 10 functions. You can use a program to make it happen. Specifically, a control unit for controlling each processing unit of the prediction coding method determination device 10, the moving image coding device 1 or the moving image decoding device 2 is configured by a central processing unit (CPU) in a computer. It is possible to configure at least one memory as a storage unit that appropriately stores a program required to operate each processing unit. That is, by causing such a computer to execute the program by the CPU, the functions of the predictive coding method determination device 10, the moving image coding device 1 or the moving image decoding device 2 can be realized. Can be done. Further, a program for realizing the functions of each processing unit of the predictive coding method determination device 10, the moving image coding device 1 or the moving image decoding device 2 is placed in a predetermined area of the above-mentioned storage unit (memory). Can be stored. Such a storage unit can be configured by a RAM or ROM inside the device, or can be configured by an external storage device (for example, a hard disk). Further, such a program can be configured as a part of software (stored in ROM or an external storage device) on an OS used by a computer. Further, a program for causing such a computer to function as each processing unit of the predictive coding method determination device 10, the moving image coding device 1 or the moving image decoding device 2 is recorded on a computer-readable recording medium. be able to. Further, each processing unit of the predictive coding method determination device 10, the moving image coding device 1 or the moving image decoding device 2 may be configured as a part of hardware or software, and each of them may be combined and realized.

Although the present invention has been described above with reference to examples of specific embodiments, the present invention is not limited to the examples of the above-described embodiments, and can be variously modified without departing from the technical idea. For example, in the example of the above-described embodiment, an example of calculating the frequency analysis coefficient mainly by using the second-order wavelet packet analysis processing by the Haar filter has been described, but the frequency analysis coefficient may be other than the Haar filter, and further, the third-order or higher-order wavelet packet analysis may be performed. It can also be a process.

Further, in the example of the above-described embodiment, the predictive coding method determination device 10 mainly uses the reference image as the first frame in the predetermined group and the determination image as one frame other than the first frame, and is based on the matching amount. An example of selecting and determining the prediction coding method for all the frames in the predetermined group at once has been described. On the other hand, the predictive coding method determination device 10 uses the reference image as the first frame in the predetermined group and the determination image as one or more frames other than the first frame in the predetermined group based on the matching amount. It is also possible to sequentially select and determine the predictive coding method for each frame.

Therefore, the predictive coding method determining device 10 according to the present invention, and the moving image coding device 1 and the moving image decoding device 2 are not limited to the above-mentioned example of the embodiment, and are described in the scope of claims. Only restricted.

According to the present invention, in the inter-screen predictive coding process, it is possible to suppress the coding deterioration that tends to occur in a critical moving image and improve the total subjective image quality. Therefore, the inter-screen predictive coding process or the in-screen image can be improved. It is useful for selecting a predictive coding method for predictive coding processing.

1 Video coding device 2 Video decoding device 10 Predictive coding method determination device 11 Group setting unit 12 Predictive coding unit 13 Stream generation unit 21 Stream input unit 22 Predictive decoding unit 101 Wavelet packet analysis unit 102 Matching processing unit 103 Prediction Coding method selection decision unit

Claims (10)

It is a predictive coding method determination device that automatically selects and determines the predictive coding method in a predetermined group divided into a plurality of screen groups for the moving image before the predictive coding process.
A reference image and a judgment image predetermined as judgment targets of the prediction coding method are extracted from the predetermined group, and the frequency of each of the reference image and the judgment image is obtained by p (p ≧ 2) order wavelet packet analysis processing. Wavelet packet analysis unit that calculates the analysis coefficient, and
Using the element value of each frequency band at an arbitrary phase position in the frequency analysis coefficient as a feature amount, inter-screen matching processing is performed for each of the reference image and the determination image, and the matching amount based on a predetermined similarity evaluation index is calculated. Matching processing unit and
Based on the matching amount, a predictive coding method selection determination unit that selectively determines whether the predictive coding method within the predetermined group is the inter-screen predictive coding process or the in-screen predictive coding process.
A predictive coding method determining device comprising. The prediction coding method for each frame in the predetermined group is sequentially selected based on the matching amount in which the reference image is the first frame in the predetermined group and the determination image is one or more frames other than the first frame. The predictive coding method determining apparatus according to claim 1, wherein the determination is made. Based on the matching amount in which the reference image is the first frame in the predetermined group and the determination image is one frame other than the first frame, the prediction coding method of all the frames in the predetermined group is collectively performed. The predictive coding method determining apparatus according to claim 1 or 2, wherein the selection and determination are performed. The predictive coding method selection determination unit is premised on applying a predetermined basic predictive coding method to each screen in the predetermined group, and is based on the matching amount with the basic predictive coding method. 3. The predictive coding method determination apparatus according to claim 3, wherein is switched to apply a different second predictive coding method to determine the predictive coding method within the predetermined group. The predictive coding method selection determination unit is based on the matching amount on the premise that a predetermined basic predictive coding method including an interscreen predictive coding process is applied to each screen in the predetermined group. 3. Predictive coding method determination device. When calculating the matching amount, the matching processing unit sets a matching weight for each frequency analysis coefficient related to the inter-screen matching processing, and sets the value of the matching weight in the frequency band of at least the second rank to the frequency band of the first rank. The predictive coding method determining apparatus according to any one of claims 1 to 5, wherein the matching amount is calculated by making it larger than the value of the matching weight in the above. When calculating the matching amount, the matching processing unit sets a matching weight for each frequency analysis coefficient related to the inter-screen matching processing, and the larger the time distance between the reference image and the determination image, the more the matching weight is set. The predictive coding method determining apparatus according to any one of claims 1 to 6, wherein the matching weight is increased to calculate the matching amount. A group setting unit that inputs a moving image before predictive coding processing, divides it into a plurality of screen groups, and sets a predetermined group divided into the plurality of screen groups.
The predictive coding method determining apparatus according to any one of claims 1 to 7, which automatically selects and determines the predictive coding method in the predetermined group.
A plurality of screen groups in the predetermined group are subjected to interscreen prediction coding processing or prediction coding processing by in-screen prediction coding processing based on the prediction coding method instructed by the prediction coding method determination device. A predictive coding unit that generates a coded signal subjected to predictive coding processing and header information indicating a predictive coding method instructed by the predictive coding method determining device.
A stream generator that performs a predetermined entropy coding process on the coded signal, writes the header information to a predetermined stream header, and generates a bit stream.
A moving image encoding device comprising. A stream input unit that inputs a bit stream from the moving image coding device according to claim 8 and extracts the header information and the coded signal from the bit stream in units of the predetermined group.
Based on the header information in the predetermined group unit, the predictive coding method by the inter-screen predictive coding process or the in-screen predictive coding process is determined, the predictive decoding process is performed on the coded signal, and the restored moving image is obtained. A predictive decoding unit that generates an image and
A moving image decoding device. A program for operating a computer as the predictive coding method determining device according to any one of claims 1 to 7.

Images