JPH11298902A - Image coder and image coding method therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance the coding efficiency of an image without deteriorating the image quality. SOLUTION: In the case of a B picture, compression coding processing utilizing inter-frame two-way prediction coding with motion compensation is conducted. In this case, a motion compensation information decision section 39 conducts discrimination processing for deciding in which of the reverse, forward or both directions the prediction coding is to be made, based on the difficulty of prediction in the reverse and forward directions, and outputs data in response to the discrimination to a motion compensation circuit 40 as actual motion compensation data S5 . The motion compensation circuit 40 generates prediction image data from stored reference image data, in response to the motion compensation data S5 decided by the motion compensation information decision section 39. Furthermore, it output motion compensation data S6 actually used for the motion compensation for a variable length coding circuit 34. The variable length coding circuit 34 applies variable length coding to the motion compensation data S6 .

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding apparatus and method for compressing and encoding image data.

[0002]

2. Description of the Related Art A transmitting side compresses image data and transmits the compressed data, and a receiving side compresses and records the compressed image data.
In a compressed image recording / reproducing system or the like that expands and outputs compressed image data during reproduction, a method of compressing image data includes, for example, MPEG (Moving Picture Exposure).
erts Group) standard. In this bidirectional predictive coding method, two types of coding, ie, intra-frame coding and inter-frame coding, are performed. Interframe coding is further divided into interframe forward prediction coding and interframe bidirectional prediction coding. Here, in intra-frame encoding, for example, input image data is divided into macroblocks for each predetermined unit, and after performing frequency transformation such as DCT (discrete cosine transformation) for each macroblock, quantization is performed. Compression is performed. On the other hand, in the inter-frame coding, for example, the input image data is converted into macroblocks for each predetermined unit, and a motion vector for a previously coded prediction reference image is detected for each macroblock. To generate a predicted image corresponding to the original image from the previously coded predicted reference image. Then, compression is performed by predictive coding using the difference between the original image and the predicted image. In the MPEG standard, the difference is subjected to DCT for encoding.

As described above, in the bidirectional predictive coding system,
Three types of encoding are performed: intra-frame encoding, inter-frame forward predictive encoding, and inter-frame bidirectional predictive encoding. Images of each encoding type are I-picture (intra coded picture) and P-picture, respectively. (Predictiv
e coded picture) and B picture (bidirectionally)
Predictive coded picture).

Here, the encoding process for a B picture will be described in further detail. In the encoding process of a B picture,
Forward direction (Forw) with respect to the input original image (B picture)
ard) and at least one of a reference image in the past on the time axis for performing prediction in the backward direction and a reference image in the future on the time axis for performing prediction in the backward direction (Backward). A predicted image corresponding to the original image is generated by predicting the motion of the original image and performing motion compensation, and prediction encoding is performed using a difference between the generated predicted image and the original image. At this time, in a stage before performing substantial motion compensation, information used for motion compensation is based on information from a reference image in any of the prediction directions of the backward, forward, and both sides (both backward and forward). A process is performed in a so-called reference plane determination step of deciding whether to create the image. In the reference plane determination step, the difficulty of prediction from the reference image in the forward direction and the backward direction is calculated, and based on the calculated difficulty of prediction, any one of the prediction directions of the backward direction, the forward direction, and the bilateral direction is calculated. It is determined whether to use the image information.

[0005] The difficulty of prediction is, for example, M
The E residual is used. The ME residual is a sum of absolute values or a sum of squares of a difference between pixel values of a target macroblock of an original image to be encoded and a target macroblock of a reference image.

FIG. 9 is an explanatory diagram for explaining an example of a conventional technique for determining a prediction direction in a reference plane determination step. In the figure, the horizontal axis represents the ME residual value ME _BWD in the backward direction, and the vertical axis represents the ME residual value ME _FWD in the forward direction. Further, in FIG.
Is a region where both directions are used as the prediction direction,
The area B is an area where only the reverse direction is used. Also,
The area F is an area where only the forward direction is used.

As shown in FIG. 1, when the difficulty of prediction in the forward and backward directions (residuals ME _FWD , ME _BWD ) is substantially equal, that is, the residuals M in the forward and reverse directions are equal.
If there is no difference of more than twice between E _FWD and ME _BWD (region A), encoding processing involving motion compensation prediction using information from both sides is performed regardless of the magnitude of its absolute value. Will be Also, the forward and backward residuals ME _FWD , M
When there is a difference of more than twice in E _BWD , an encoding process involving motion compensation prediction using information of only the prediction reference plane having the smaller ME residual is performed. That is, the forward ME
If there is a relationship of ME _BWD > 2ME _FWD between the residual value ME _FWD and the backward ME residual value ME _BWD (region F), the motion compensation using information only in the forward direction is performed. An encoding process involving prediction is performed. Further, M is set between the forward ME residual value ME _FWD and the backward ME residual value ME _BWD.
If there is a relationship of E _FWD > 2ME _BWD (region B),
An encoding process involving motion compensation prediction using information only in the reverse direction is performed. At this time, the amount of information to be transferred as information for motion compensation after encoding is half when using information in only one direction in the backward or forward direction, compared to using information in both directions.

[0008] For example, in an image with little movement such as a still image or a so-called pan image in which an image moves slowly, the ME residual in the backward direction and the forward direction is small, and has substantially the same value. It is known. In such a case, the ME residual is distributed substantially in a region indicated by a region a in FIG. An image material distributed in the area a has a strong correlation between adjacent image frames, and originally belongs to a material that is good at encoding by MPEG.

[0009]

However, according to the above-mentioned conventional techniques, for example, a still image, a so-called pan image in which an image moves slowly, etc., is relatively easy to accurately predict, and the information amount in one direction only is originally required. However, even if the image can be coded sufficiently, the difficulty of prediction in the backward and forward directions is almost equal (region a in FIG. 9). A prediction is made. Therefore, in this case, it is necessary to encode and transfer twice the amount of information, although the prediction encoding can be sufficiently performed with the information amount based on the image in one direction, and the encoding efficiency of the image is increased. Is low, which is uneconomical.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and an object of the present invention is to provide an image encoding apparatus and a method capable of increasing the image encoding efficiency without deteriorating the image quality. .

[0011]

According to the present invention, there is provided an image coding apparatus for an input image, comprising two types of reference images which are temporally past and future in accordance with the type of the input image. Encoding means for performing an encoding process including a predictive encoding process with one-sided motion compensation using one of them and a predictive encoding process with a bidirectional motion compensation using both of two reference images; During encoding of an image to be subjected to predictive encoding in the direction, a value representing the difficulty of prediction for each reference image is within a predetermined range in which it is determined that the prediction encoding using only one reference image is sufficient. If it is within the range, the prediction means is provided with a prediction direction restriction means for causing the encoding means to perform a unidirectional prediction encoding process using only one reference image.

Further, the image encoding method according to the present invention uses one of two temporally past and future reference images for an input image in accordance with the type of the input image. In addition to performing the encoding process including the prediction encoding process with motion compensation in one direction and the prediction encoding process with motion compensation in both directions using both reference images, the prediction encoding process in both directions is performed. At the time of the encoding process of the image to be performed, when the value representing the prediction difficulty for each reference image is within a predetermined range that is determined to be sufficient by the prediction encoding process using only one reference image, The prediction direction is restricted so as to perform a unidirectional prediction encoding process using only one reference image.

In the image coding apparatus according to the present invention, the value indicating the difficulty of prediction for each reference image is set to one of the values by the prediction direction limiting means when the image to be subjected to the bidirectional predictive coding is coded. When the prediction direction using only the reference image is within a predetermined range that is determined to be sufficient, the prediction direction is set so that one-sided prediction encoding process using only one reference image is performed. Limited.

Further, in the image encoding method according to the present invention,
Predictive coding processing for one-way motion compensation using one of two temporally past and future reference images for an input image in accordance with the type of the input image; In addition to performing the encoding process including the predictive encoding process with bilateral motion compensation using both of the two reference images, the encoding process of the image to be subjected to the bidirectional predictive encoding process is performed for each reference image. If the value indicating the difficulty of prediction is within a predetermined range determined to be sufficient by the predictive encoding process using only one reference image, the one-way predictive code using only one reference image is used. The prediction direction is limited so that the conversion process is performed.

[0015]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of an image coding apparatus according to one embodiment of the present invention. The image coding apparatus 1 according to the present embodiment receives an input image signal S ₁ and outputs a picture (I picture, P picture) according to the coding order.
(Picture, B picture), and an image rearranging circuit 21 for inputting the output data of the image rearranging circuit 21 and discriminating between a frame structure and a field structure.
Scan conversion / macroblock circuit 2 for performing scan conversion and macroblock formation of 16 × 16 pixels according to the determination result
2, a motion detection circuit 30 for detecting data for motion compensation such as a motion vector based on the output data of the scan conversion / macroblock conversion circuit 22, and a motion detection circuit 30
Conversion and macroblocking circuit 2 input via
2, a DCT circuit 32 that performs DCT on the output data of the subtraction circuit 31 in DCT block units and outputs DCT coefficients, A quantization circuit 33 for quantizing output data of the circuit 32 based on a predetermined quantization step (a coefficient of division for quantization), and a variable-length code for variable-length encoding the output data of the quantization circuit 33 of the circuit 34, and a buffer memory 35 for outputting the output data of the variable-length encoding circuit 34 temporarily holds, as compressed image data S ₂ made from the bitstream.

The image encoding device 1 further includes a quantization circuit 3
3, an inverse quantization circuit 36 for inversely quantizing the output data, an inverse DCT circuit 37 for performing an inverse DCT on the output data of the inverse quantization circuit 36, an output data of the inverse DCT circuit 37 and predicted image data. Addition circuit 3 for adding and outputting
8, the motion compensation information determination unit 39 determines and outputs data of the motion compensation used from the output data S ₃ of the motion detection circuit 30 for generating the predicted image data, and holds the output data of the adding circuit 38, the motion A motion compensation circuit 40 that performs motion compensation on a reference image corresponding to the motion compensation data sent from the compensation information determination unit 39 to generate predicted image data, and outputs the predicted image data to the subtraction circuit 31 and the addition circuit 38 And In the motion compensation circuit 40, the data S ₆ of the motion compensation actually used for generating the predicted image data is output to the variable length coding circuit 34. The variable length coding circuit 34 also performs variable length coding on the motion compensation data S ₆ output from the motion compensation circuit 40. Here, the motion compensation information determining unit 39 corresponds to the prediction direction limiting unit in the present invention.

The image coding apparatus 1 further determines a quantization step in the quantization circuit 33 so that the generated code amount (generated bit amount) becomes a target predetermined code amount, for example. An encoding control unit 41 that provides the quantization step to the quantization circuit 33 is provided.

The motion detecting circuit 30 detects the data required for motion compensation in the motion compensation circuit 40 is adapted to transmit data S ₃ for detecting the motion compensation to the compensation information determination unit 39 . As data for motion compensation detected by the motion detection circuit 30, for example,
Information about the ME residual and the motion vector for the frame structure and the field structure is included. Here, the ME residual refers to each macroblock of the original image to be encoded,
This is the sum of absolute values or the sum of squares of the luminance difference of each pixel between each macroblock of the original image and each macroblock of the reference image, and is defined by the following equation, for example. In the following equation, Ma is 1 of the original image.
Mb represents a macroblock of a backward or forward reference image corresponding to the macroblock Ma. The subscript i represents, for example, an integer from 0 to 255, and Σ represents the sum of the values | Ma _i −Mb _i | for all the values of _i .

ME = Σ | Ma _i -Mb _i | FIG. 2 is an explanatory diagram for describing data for motion compensation detected by the motion detection circuit 30.
In FIG. 2, P ₁ and P ₂ indicate P pictures, and B ₁ and B ₂ indicate B pictures. The picture data structure in this figure is such that the period (M) at which a P picture appears is 3, and two B pictures B ₁ and B ₂ are inserted between adjacent P pictures P ₁ and P _2. ing. Note that, in this example, the case where the picture adjacent to the B picture is a P picture has been described, but the picture adjacent to the B picture may be an I picture. In the figure, it is assumed that the data of the original image currently input to the motion detection circuit 30 is a picture B ₁ , and the data in the past on the time axis for performing forward prediction is a picture P ₁ . There, the data in the future on a time axis for performing the backward prediction is assumed to be the picture P _2. FIG. 2 illustrates an example in which a rectangular object moves from the lower left to the upper right as time passes.

When the picture B ₁ is input, the motion detecting circuit 30 compares each of the macroblocks of the picture B ₁ with respect to two reference pictures, a past picture P ₁ and a future picture P _2. Is searched from within a predetermined search range of the pictures P ₁ and P ₂ , and the residual ME for each macro block is found.
_FWD and ME _BWD are required. In the example of FIG. 2, the macro block of the picture P _1, P ₂ that are most consistent macroblock M _c of the picture B ₁ is, has each macro-block M _F, and M _B. Here, the coordinates of the center of the macro block M _c a (0, 0), macroblock M _F, the coordinates of the center of the _{M B, (- X F,} -X
_F), and a (X _B, X _B). The motion detection circuit 30, most Consistent based on the center position of the macro block motion vector V _F, so as to detect the V _B. Motion vector V _F, V _B is a motion vector of the macro block M _F, M _B for the macroblock M _c.

FIG. 3 is a block diagram showing the configuration of the motion compensation information determining section 39. Before performing substantial motion compensation in the motion compensation circuit 40, the motion compensation information determination unit 39 performs predictive coding in any of the backward, forward, and bilateral (both backward and forward) prediction directions. A determination process is performed to determine whether or not to perform. For example, the motion compensation information determining unit 39 determines that the value indicating the difficulty of prediction for each reference image is a prediction code using only one reference image during the encoding process of the image to be subjected to the bidirectional prediction encoding process. If it is within a predetermined range determined to be sufficient by the coding process, a determination process is performed such that a unidirectional prediction coding process using only one reference image is performed.

The motion compensation information determining section 39 includes a reference value holding section 51 for storing a reference value serving as a reference for the determination processing, a reference value stored in the reference value holding section 51,
By comparing the data S ₄ about the difficulty in the past has been extracted from the data S ₃ and the reference image data for future prediction from 0, determines the predictive encoding process in any prediction direction should be performed a determination unit 52, in accordance with the determination result of the determination unit 52, selection unit that selects data S ₅ of the motion compensation to be sent to the motion compensation circuit 40 from the data S ₃ detected in the motion detection circuit 30 outputs 53. The reference value holding unit 51 stores, for example, a reference value for defining a predetermined range in which the value indicating the difficulty of prediction is determined to be sufficient in the predictive encoding process using only one reference image. The determination unit 52 compares, for example, the reference value stored in the reference value holding unit 51 with a value representing the prediction difficulty for each reference image, and determines that the value representing the prediction difficulty for each reference image is
It is determined whether it is within a predetermined range where it is determined that the prediction encoding process using only one reference image is sufficient. As the data S ₄ about the difficulty of prediction, which is to be compared in the determination unit 52, for example, ME residual is used. The function of the motion compensation information determination unit 39 may be realized by software, or may be realized by hardware by forming a dedicated circuit.

FIGS. 4 and 5 show the motion compensation information determining unit 3.
FIG. 9 is an explanatory diagram for describing a determination process performed in No. 9; 4 and 5, the horizontal axis represents the ME residual value ME _BWD in the reverse direction with respect to the input original image data, and the vertical axis represents the ME residual value ME _FWD in the forward direction. I have.

FIG. 4 shows a CCIR (International Radio)
Consultative Committee: International video communications standardized by diva with no
It is a figure which shows distribution of ME residual obtained by actually inputting ise (hereinafter, described as diver) and flower garden (hereinafter, described as fwgn) as image data. The distribution of the ME residual in FIG. 4 shows a case where the luminance of each pixel is represented by 8 bits. In FIG. 4, the area inside the two wavy lines corresponds to the area in which bidirectional predictive encoding has been performed conventionally, and the part shown by a solid line corresponds to the part shown by a solid line in FIG. Yes, it is. In FIG. 4, a white square □ indicates an ME residual regarding diver, and a black square ■ indicates an ME residual regarding fwgn. fwgn is div
This is a so-called pan image in which the image has less movement than in FIG. For this reason, fwgn
Has a strong correlation between adjacent image frames. f
Looking at the distribution of the ME residual for wgn, it can be seen that there are a data group concentrated near the origin (the area indicated by the symbol D) and a data group distributed somewhat apart therefrom. .

In the present embodiment, the data group concentrated in the area D near the origin as described above can be easily and accurately predicted by motion compensation, and can be sufficiently encoded by prediction in only one direction. The image is
The motion compensation prediction is performed based on the motion compensation data using one reference image. More specifically,
In the motion compensation information determination unit 39, first, the determination unit 52
However, when the value of the ME residual is a value in a predetermined area near the origin, it is determined that prediction of motion compensation should be performed based on motion compensation data using one reference image, and this determination is made. The result is notified to the selection unit 53. When the determination unit 52 determines that the prediction of the motion compensation should be performed based on the data of the motion compensation using one reference image, the selection unit 53 outputs the data of the motion compensation of only one reference image. Is selectively sent to the motion compensation circuit 40. In this case, the motion compensation circuit 40
Is configured to perform motion compensation on the basis of the data of the motion compensation in only one direction selected by the selection unit 53 to generate predicted image data. In the motion compensation circuit 40, the data S ₆ of the motion compensation on one side only in the direction that is actually used for generating the predicted image data is output to the variable length coding circuit 34.

Here, when the value of the ME residual is a value in a predetermined area near the origin, whether to perform motion compensation in the forward direction or the backward direction will be considered. When the value of the ME residual is a value in a predetermined area near the origin,
Since there is a strong correlation between frames of adjacent images, basically, the prediction direction may be either the forward direction or the reverse direction. However, it is desirable to perform the motion compensation in the reverse direction in consideration of the characteristics of the variable length encoding process in the variable length encoding circuit 34 later. For example, "ISO /
According to the standard document in “IEC 13818-2: 1995 (E)”, when a macro block is a standard macro block of pattern 1 and has been subjected to inter-frame coding, it indicates a macro block type. If the value of the code is in the forward direction, it is 010100 to 0011.
While the one in the opposite direction is 0011
Since it is expanded from 00 to 011, performing predictive coding processing with motion compensation in the reverse direction reduces the code amount after variable-length coding and improves coding efficiency.

Next, with reference to FIG. 5, the determination processing of the motion compensation information determination section 39 will be described more specifically. FIG.
, An area A is an area in which bidirectional predictive encoding is performed using data of two past and future reference images, and an area B including an area a uses only data of a future reference image. This is an area where prediction coding in the reverse direction is performed.
An area F is an area where forward prediction encoding is performed using only the data of the past reference image. The area a is
This is a region corresponding to the region D shown in FIG.
This is the area where the prediction encoding in both directions has been performed. The predetermined threshold value α shown in the figure is set to, for example, a value of about several percent of the maximum value of the residual ME _BWD .

The determination section 52 of the motion compensation information determination section 39
The value of the residual ME _BWD in the reverse direction is larger than the threshold value α, and the residual ME _FWD in the forward direction and the reverse direction is
When the ME _BWDs are substantially equal, that is, when the difference between the ME residual in the backward direction and the forward direction does not exceed twice (region A), motion compensation from both sides is performed regardless of the magnitude of the absolute value. Is determined to be performed. In addition, the determination unit 52 determines that the value of the residual ME _BWD in the backward direction is larger than the threshold value α, and that the value of ME is between the residuals ME _FWD and ME _{BWD in the} forward and reverse directions.
When there is a relationship of _BWD > 2ME _FWD (region F), it is determined that prediction of forward motion compensation should be performed. The determination unit 52 determines _whether the value of the residual ME _BWD in the backward direction is equal to or _smaller than the threshold value α, the value of the forward ME residual ME _FWD and the value of the backward ME residual ME _BWD.
If there is a relationship of ME _FWD > 2ME _BWD (region B), it is determined that prediction of motion compensation in the reverse direction should be performed.

Next, the operation of the image coding apparatus 1 according to the present embodiment will be described. The following description also serves as a description of the image encoding method according to the present embodiment.

First, the overall operation will be described. In the image encoding apparatus 1, the image rearranging circuit 21, the input image signal S _1, a picture in the order of coding (I picture, P picture, B picture) the order of the sorted. Next, the scan conversion / macroblocking circuit 22 determines whether the frame structure or the field structure is used, and performs scan conversion and macroblocking according to the determination result. Scan conversion / macroblock circuit 22
Is sent to the motion detection circuit 30. Motion detecting circuit 30, and sends the output data of the scan conversion macroblock forming circuit 22 to the subtraction circuit 31, the data S ₃ used from the output data of the scan conversion macroblock forming circuit 22 to the motion compensation, such as the motion vector Detected and sent to the motion compensation information determination unit 39.

Then, I picture, P picture and B picture
An encoding process corresponding to each of the pictures is performed.

In the case of an I picture, an encoding process by intra-frame encoding is performed. That is, the subtraction circuit 31
, Without taking the difference from the predicted image data, the output data of the scan conversion / macroblocking circuit 22 is directly input to the DCT circuit 32 to perform DCT. The output data of the quantization circuit 33 is variable-length coded by the conversion circuit 34, and the variable-length coding circuit 3 is
4 after temporarily holding the output data of the compressed image data S ₂
Is output. The output data of the quantization circuit 33 is inversely quantized by the inverse quantization circuit 36, and the output data of the inverse quantization circuit 36 is inversely transformed by the inverse DCT circuit 37.
And the output image data of the inverse DCT circuit 37 is input to the motion compensation circuit 40 via the addition circuit 38 and held.

In the case of a P picture, a compression encoding process using inter-frame forward prediction encoding is performed. That is, the data S ₃ input to the motion compensation information determination unit 39
But it is actually input to the motion compensation circuit 40 as the data S ₅ used in the motion compensation. Motion compensation circuit 40
Then, predicted image data is generated from the stored reference image data corresponding to the past I picture or P picture, and the predicted image data is subtracted from the subtraction circuit 31 and the addition circuit 38.
Output to Further, the motion compensation circuit 40 outputs the data S ₆ of the motion compensation used for generating the prediction image data to the generation variable length coding circuit 34. Further, the difference between the output data of the scan conversion / macroblocking circuit 22 and the predicted image data from the motion compensation circuit 40 is calculated by the subtraction circuit 31, the DCT is performed by the DCT circuit 32, and the DCT coefficient is calculated by the quantization circuit 33. The output data of the quantization circuit 33 is variable-length coded by a variable length coding circuit 34, and the variable length coding circuit 3 is
Temporarily holding the fourth output data is output as compressed image data S _2. Further, the output data of the quantization circuit 33 is inversely quantized by the inverse quantization circuit 36, inverse DCT is performed on the output data of the inverse quantization circuit 36 by the inverse DCT circuit 37, and the inverse DCT circuit 37 The output data and the predicted image data are added and input to the motion compensation circuit 40 to be held.

In the case of a B picture, a compression encoding process utilizing inter-frame bidirectional prediction encoding is performed. First, the motion compensation information determination unit 39 performs a determination process for determining which of the prediction directions of the backward, forward, and both sides (both the backward and forward directions) to perform prediction encoding. the data corresponding to the determination, actually output to the motion compensation circuit 40 as the data S ₅ used in the motion compensation. In the motion compensation circuit 40, at least one reference image data corresponding to the past I picture or P picture held and the motion compensation information movements on the basis of the data S ₅ of the motion compensation vector such as to generate the predicted image data, and outputs the predicted image data to the subtraction circuit 31 and the adding circuit 38. Also, the motion compensation circuit 40 converts the motion compensation data S ₆ into the variable length coding circuit 3.
4 is output. Further, the difference between the output data of the scan conversion / macroblocking circuit 22 and the predicted image data from the motion compensation circuit 40 is calculated by the subtraction circuit 31, the DCT is performed by the DCT circuit 32, and the DCT coefficient is calculated by the quantization circuit 33. quantizing the output data of the quantization circuit 33 by the variable-length coding circuit 34 variable-length coding, temporarily outputs held as compressed image data S ₂ output data of the variable length coding circuit 34 by the buffer memory 35. The B picture is not stored in the motion compensation circuit 40.

Next, the operation of the determination process in the motion compensation information determination section 39 will be described with reference to the flowchart of FIG.

First, the determination unit 5 of the motion compensation information determination unit 39
2 determines whether or not the value of the residual ME _BWD in the reverse direction with respect to the input original image data is smaller than a threshold value α (step S101). Here, when the residual ME _BWD is equal to or _smaller than the threshold value α (step S101; Y),
The determination unit 52 determines that the motion compensation circuit 40 should perform prediction coding only in the reverse direction (step S).
106).

If the value of the residual ME _BWD is not equal to or _smaller than the threshold value α (step S101;
N) Next, it is determined whether or not the residual ME _{FWD in the} forward direction with respect to the input original image data is larger than twice the residual ME _BWD (step S102). Where the residual ME
_{When FWD} is larger than twice the residual ME _BWD (step S102; Y), the determination unit 52 determines that predictive encoding using data only in the reverse direction should be performed (step S106). ).

On the other hand, when the residual ME _FWD is not larger than twice the residual ME _BWD (step S102;
N) Then, the determination unit 52 determines that the residual ME _BWD is
_It is determined whether or not it is larger than twice the _FWD (step S1)
03). Here, when the residual ME _BWD is larger than twice the residual ME _FWD (Step S103; Y), the determination unit 52 determines that prediction coding only in the forward direction should be performed (Step S103). Step S104).

If the residual ME _BWD is not larger than twice the residual ME _FWD (step S103; N
O), the determination unit 52 determines that predictive encoding in both directions should be performed (step S105).

FIG. 7 is an explanatory diagram showing an example of improvement of the coding efficiency when the coding process is actually performed by the above operation. In this figure, the horizontal axis represents time (se).
c), and the vertical axis indicates the rate of change (%) in the amount of generated bits generated by the encoding process. The rate of change of the amount of generated bits in this figure indicates the rate of change of the amount of generated bits generated by the processing of the present embodiment as a relative value to the amount of generated bits generated by the conventional encoding processing. . In the figure, if the value on the vertical axis is negative, it indicates that the coding efficiency is improved as compared with the conventional case. In this figure, fwgn is used as image data, and the threshold α used for the determination processing in the motion compensation information determination unit 39 is set to a value of 2000 as the value of the residual ME _BWD during the encoding processing. This is the result obtained. From this figure, it can be seen that the coding efficiency is improved by about 5% on average as compared with the conventional coding processing, and in some cases, the coding efficiency is improved by 10% or more.

FIG. 8 is an explanatory diagram showing the values of the coding efficiency obtained by varying the threshold value α used in the determination processing in the motion compensation information determination section 39. In this figure, the horizontal axis represents the threshold value α (residual ME _BWD ), and the vertical axis represents the rate of change (%) in the amount of generated bits generated by the encoding process. Also, this figure is data obtained when an actual encoding process is performed using fwgn as image data. If the value on the vertical axis in the figure is negative, the encoding efficiency is improved. Indicates that From this figure, as the value of the threshold α is increased from 0, the coding efficiency is improved. After the coding efficiency is maximized at about 2000, the improvement rate of the coding efficiency gradually increases. It can be seen that the coding efficiency becomes worse at a certain value or more.

As described above, according to the image encoding apparatus 1 of the present embodiment, a still image or a gradual image moves during the encoding of an image to be subjected to the bidirectional predictive encoding. For an image such as a so-called pan image, which is relatively easily predicted by motion compensation and can be satisfactorily encoded using data of a reference image in only one direction, motion compensation is performed only in one direction to obtain predicted image data. generated. Thus performing encoding, the amount of data S ₆ of the motion compensation of the motion vector or the like to be output to the variable length coding circuit 34 is reduced from the motion compensation circuit 40, reducing the image quality As a result, a high-quality image equivalent to that of the related art or without distortion can be obtained, and the encoding efficiency of image data can be increased. Also, in consideration of the characteristics of the variable-length encoding process after performing the motion compensation, if the value of the ME residual is a value in a predetermined area near the origin, the input original image is Thus, since motion compensation is performed only in one direction using a future reference image, the encoding efficiency of image data can be maximized.

Note that the present invention is not limited to the above embodiment, and various modifications can be made. For example, in the above embodiment, the motion compensation information determination section 39, but so as to output the determined data S ₅ of the motion compensation for use in actually generating the predicted image data to the motion compensation circuit 40, motion compensation Data S of motion compensation performed in the information determining unit 39
The function of determining ₅ may be provided in the motion compensation circuit 40, and the motion compensation information determination unit 39 may be omitted from the configuration.

[0045]

As described above, claims 1 to 3 are described.
According to the image encoding apparatus described in any one of the above or the image encoding method according to the fourth aspect, when encoding an image to be subjected to bidirectional predictive encoding, the prediction difficulty for each reference image is reduced. If the value to be represented is within a predetermined range determined to be sufficient by the predictive encoding process using only one reference image, the unidirectional predictive encoding process using only the one reference image is performed. Therefore, it is possible to increase the coding efficiency of image data without deteriorating the image quality.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an image encoding device according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram for describing data for motion compensation detected by a motion detection circuit in the image encoding device shown in FIG.

FIG. 3 is a block diagram illustrating a configuration of a motion compensation information determination unit in the image encoding device illustrated in FIG. 1;

FIG. 4 is an explanatory diagram for describing a motion compensation determination process performed by a motion compensation information determination unit illustrated in FIG. 3;

FIG. 5 is another explanatory diagram for describing a motion compensation determination process performed in the motion compensation information determination unit illustrated in FIG. 3;

FIG. 6 is a flowchart showing an operation of a motion compensation information determination unit in the image encoding device shown in FIG. 1;

FIG. 7 is an explanatory diagram showing an example of improvement in coding efficiency by the image coding device shown in FIG. 1;

FIG. 8 is another explanatory diagram showing an example of improving the coding efficiency by the image coding device shown in FIG. 1;

FIG. 9 is an explanatory diagram for describing a motion compensation determination process performed in a conventional image encoding device.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Image coding apparatus, 21 ... Image rearrangement circuit, 22 ...
Scan conversion / macroblocking circuit, 30: motion detection circuit, 31: subtraction circuit, 32: DCT circuit, 33: quantization circuit, 34: variable length coding circuit, 35: buffer memory, 36: inverse quantization circuit, 37 ... Inverse DCT circuit, 38 ...
Addition circuit, 39: motion compensation information determination unit, 40: motion compensation circuit, 41: coding control unit.

Claims (4)

[Claims] 1. A prediction that involves one-way motion compensation using one of two temporally past and future reference images for an input image in accordance with the type of the input image. Coding means for performing coding processing including predictive coding processing with bidirectional motion compensation using both coding processing and two reference images; and coding of an image to be subjected to bidirectional predictive coding processing. At the time of processing, when the value representing the difficulty of prediction for each reference image is within a predetermined range that is determined to be sufficient by the predictive encoding process using only one reference image, And a prediction direction limiting unit for performing a one-way prediction encoding process using only one reference image. 2. The method according to claim 1, wherein the predicting direction restricting unit transmits, to the encoding unit, only a reference image that is temporally future when the value indicating the prediction difficulty is within the predetermined range. 2. The image encoding apparatus according to claim 1, wherein the used one-way predictive encoding process is performed. 3. The prediction direction restriction means, wherein: a value representing the difficulty of prediction stores a reference value for defining the predetermined range; and a reference value stored in the storage means; Determining means for comparing a value representing the difficulty of prediction for each reference image with a value representing the difficulty of prediction of each reference image within the predetermined range; 2. The image encoding apparatus according to claim 1, further comprising a selection unit that selects information required for the predictive encoding process including the motion compensation according to a result. 4. A prediction involving one-sided motion compensation using one of two temporally past and future reference images for an input image in accordance with the type of the input image. In addition to performing the encoding process including the predictive encoding process with the bidirectional motion compensation using both the encoding process and the two reference images, at the time of encoding the image to be subjected to the bidirectional predictive encoding process, If the value indicating the difficulty of prediction for each reference image is within a predetermined range that is determined to be sufficient by the predictive encoding process using only one reference image, one side using only one reference image An image encoding method, wherein a prediction direction is restricted so as to perform a direction prediction encoding process.