KR100901030B1 - Method for coding moving picture - Google Patents

Abstract

The present invention discloses a moving picture coding method in which coding efficiency is improved by reducing a coded bit rate when a scene transformation codes repetitive pictures. The disclosed invention comprises the steps of: determining whether to transform a scene of a subsequence forward predictive encoded picture when coding a bidirectional predictive encoded picture of a video sequence; Referring to a long term picture or a short term picture according to a scene transformation of the forward predictive coded picture; And when the forward predictive coded picture refers to the long term picture or the short term picture, obtaining each forward motion vector and a reverse motion vector in the direct mode.

Here, when the forward predictive coded picture obtains the macroblock prediction value in the direct mode by referring to the long term picture, the forward motion vector MV _f = MV is calculated in the direct mode, and the reverse motion vector MV _b = 0 is obtained. It is done.

Description

Video coding method {METHOD FOR CODING MOVING PICTURE}

1A is a diagram illustrating an order in which pictures are displayed when a bidirectional predictive coded picture is used in a video coding system according to the related art.

 1B is a diagram illustrating a sequence of coding a digital picture using a bidirectional predictive coded picture in a video coding system according to the prior art.

2 is a view for explaining a method of coding a forward predictive coded picture of a sequence for scene transformation according to the present invention;

3 is a diagram for describing a bidirectional predictive coded picture coding method of a sequence for scene transformation according to the present invention;

4 is a diagram illustrating a method of inserting and coding two bi-predictive coded pictures according to the present invention.

* Description of the symbols for the main parts of the drawings *

I: intra coded picture P: forward predictive coded picture

B: bidirectional predictive coded picture MV: motion vector

The present invention relates to a video coding method. More specifically, when coding a bi-prediction coded picture, the coding efficiency of the bi-prediction coded picture can be improved by judging and coding a scene transform of a forward predictive coded picture which is a subsequence picture. The present invention relates to a video coding method.

In recent years, digital video transmission has been increasingly popular for providing television, audio, video and / or data services to a customer's home. In addition, digital video can provide higher image quality than traditional analog television broadcasts, to provide new and expanded program applications.

In addition, in order to transmit such a digital video picture, it is necessary to use a data compression technique within the available bandwidth. In particular, current techniques utilize the interrelationship between pixels or blocks of pixels, while spatial compression techniques utilize interrelationships between adjacent pixels, or blocks of pixels in a picture frame.

However, in order to optimally compress and code a sequence of video picture frames, it is desirable to be able to detect whether a scene has changed in the sequence. Because digital video pictures, such as close-up conversations such as news, sports broadcasts, and interviews, or multi-point video conferences, the scenes change partly or entirely over time. Because there is.

Each time a scene change is detected, the coding method of the moving picture is changed accordingly. For example, when there is a reduction probability of the current correlation between the first frame of the new scene and the last frame of the previous scene, the motion compensation may be temporarily deferred when the scene change is detected. Moreover, a particular type of picture (e.g., I, P or B picture) can be selected based on the scene change information.

The I, P, and B pictures are used as defined in the MPEG-2 standard, which is defined as an intra coded picture or a coded picture within a picture frame. The P picture is defined as a predictive picture coded picture or a picture inter-frame forward predictive coded picture, and the B picture is defined as a bidirectionally predictive coded picture or a bidirectional predictive coded picture.

FIG. 1A is a diagram illustrating an order in which pictures are displayed when a bidirectional predictive coded picture is used in a video coding system according to the related art.

As shown in FIG. 1A, when an intra picture is displayed for the first time when a scene change occurs overall among pictures displayed in time, an intra picture is used when two bidirectionally predictive coded pictures are used. Continually, the bidirectional predictive coded pictures are displayed in the first and second order. After the bidirectional predictive encoded picture is displayed, the subsequence forward predictive encoded picture of the displayed first or second bidirectional predictive encoded picture is displayed in the third.

In the same manner as described above, the following steps are also performed: when the third subsequence forward predictive coded picture is displayed, the fourth and fifth bidirectional predictive codes and pictures are displayed, and then the subsequence forward predictive coded picture is continued. Is displayed.                         

However, the order of coding the displayed video picture is not the same as the display order. FIG. 1B is a diagram illustrating an order of coding a video picture using a bidirectional predictive coded picture in a video coding system according to the related art.

As shown in Fig. 1B, if an intra picture is coded that does not refer to any other picture after scene transformation, the subsequence forward predictive coded picture of the two bi-predictive coded pictures is always displayed before the two displayed bi-predictive coded pictures. First coded.

When coding using the bi-prediction coded picture, when it is applied to low bit coding, a number of direct prediction modes are selected in which the table of contents information is not required to be coded. In addition, when applying to high bit coding, it is essential to present table of contents information to a picture to be coded, but most bidirectional prediction modes having high prediction accuracy of a picture are selected, and thus have much higher coding efficiency than coding of a forward predictive coded picture. .

 A sub-mode that is located at the same position as the current macroblock of the bidirectional predictive coded picture among intra mode, forward predictive mode, reverse predictive mode, bidirectional predictive mode, and direct predictive mode, particularly in the case of the direct predictive mode used for predicting a compressed video picture. The forward motion vector and the reverse motion vector of the direct mode are calculated from the motion vectors referenced by the macro blocks of the sequential forward predictive coded pictures.

Then, the motion compensation value is obtained according to each direction by using the calculated values of the forward motion vector and the reverse motion vector, and finally the prediction of the macroblock to be coded by averaging or interpolating the two motion compensation values Get the value.

In general, however, when a video picture is displayed, the entire scene is converted and only a specific part of the scene is converted. When the whole scene is converted, all macro blocks of the converted scene have a picture coded in an intra mode. When only a part of the scene is transformed, only the macro block of the region where the transformation occurs is coded in the intra mode.

In the case of coding in the above manner, if the intra mode coding requiring a higher bit amount than the coding in the inter mode is applied to the low bit rate coding of a sequence in which scene transformation occurs frequently, the coding process may cause a defect. A problem arises.

The present invention obtains a forward motion vector and a backward motion vector in a direct mode by referring to long-term holding pictures stored in a frame buffer when a scene transformation occurs in whole or in part in a moving picture compressed picture, and then codes a bidirectional predictive coded picture. The purpose is to provide a video coding method that can increase the efficiency.

In order to achieve the above object, a video coding method according to the present invention,

Determining whether to transform a scene of the sub-sequential forward predictive encoded picture when coding the bidirectional predictive encoded picture of the moving picture sequence;                     

Referring to a long term picture or a short term picture according to a scene transformation of the forward predictive coded picture; And

And obtaining a forward motion vector and a reverse motion vector in the direct mode when the forward predictive coded picture refers to the long term picture or the short term picture.

Here, when the forward predictive coded picture obtains the macroblock prediction value of the direct mode by referring to the long term picture, the forward motion vector MV _f = MV is obtained in the direct mode, and the backward motion vector MV _b = 0 is obtained. When a forward predictive coded picture refers to a short term picture, a time distance T _B of a motion vector from the short sequence picture to the video referenced by the subsequence forward predictive coded picture, and the subsequential forward predictive coding A forward motion vector and a backward motion vector in a direct mode are obtained by using a time distance TR _D from a picture referred to by the picture to the short term picture to a macroblock of the bidirectional predictively encoded picture to be coded.

In order to obtain the forward motion vector and the reverse motion vector in the direct mode, the forward predictive coded picture with scene transformation is used.

MV _f = (N ₁ * MV) / D

With MV _b = N ₂ * MV / D

Where N ₁ , N ₂ , and D are predetermined factors defined to obtain respective motion vectors)

When the picture with the scene change refers to the long term picture, it is determined that N ₁ = 1, N ₂ = 0, D = 1 and transmitted to the decoder to forward motion vector MV _f = MV, and reverse motion vector MV _b = 0. It is characterized by obtaining.

Here, when the long sequence direct forward and reverse motion vectors of the subsequence picture that is the forward predictive coded picture are obtained, a direct forward motion vector MV _f = MV and MV = 0 is obtained, and the subsequence forward predictive coded picture is a short term picture. In the case of obtaining a motion vector from, set N ₁ = TR _B , N ₂ = TR _B -TR _D and D = TR _D

MV _f = (TR _B * MV) / TR _D ,

It is characterized by obtaining MV _b = (TR _B -TR _D * MV) / TR _D.

According to the present invention, when coding a moving picture that is repeatedly transformed by a scene, the subsequential forward predictive coded picture can obtain a motion vector in the direct prediction mode, and can refer to the picture from the long term picture so that the overall There is an advantage to improve the coding efficiency.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Before entering the description of an embodiment of the present invention, a picture in which a transformation occurs in the entire picture in a moving picture that is repeatedly converted is called a scene cut picture, and a picture in which the conversion occurs in a part of the picture is called a partial scene change picture ( partial scene change picture).

In addition, a buffer that holds a moving picture for a long time and stores a scene cut picture or a partial scene conversion picture is called a long-term picture holding buffer (long-term buffer), and conversely, a moving picture picture scene is coded before a relatively short time. The buffer holding the picture is called a short-term picture holding buffer (short-term buffer).

2 is a diagram for describing a method of coding a forward predictive coded picture of a sequence for scene transformation according to the present invention.

As shown in FIG. 2, it is first determined whether scene cut pictures or partial scene transform pictures of all subsequential forward predictive encoded pictures have occurred before coding the bidirectional predictive encoded pictures. This is done by detecting at the encoder of the system and passing the information to the decoding unit.

In the encoder, if the forward predictive encoded picture is determined to be the scene transformed picture or the partial scene transformed picture, the encoder predicts the forward predictive encoded picture or the short term picture retaining buffer (short term buffer) stored in the long term picture maintaining buffer (long term buffer). Coding is performed through motion estimation and motion compensation from the stored forward predictive coded pictures.

The short-term picture holding buffer (short-term buffer) is a space in which already coded pictures are stored according to a FIFO (First In First Out) outputting the first inputted picture first. Stores forward predictive coded pictures.

The long picture retention buffer is a space for storing the first picture of the sequence, that is, an intra picture, the scene cut picture and the partial scene change picture, and if the scene cut picture and the partial scene change picture are not present in the long picture retention buffer, Newly stored. In addition, the scene cut picture is coded as an intra picture, and the partial scene transform picture is coded in an intra mode only in an area in which a transform is performed.

Therefore, referring to FIG. 2, the intra picture, which is the first scene cut picture of scene set A, the first scene cut picture P50, and the first partial scene transform picture P120 of scene set B are the long time picture. It is stored in the retention buffer.

For example, if the sub-sequence forward-prediction coded picture of the bi-prediction coded picture for current coding is the scene cut picture P200 corresponding to the second scene set B in the portion shown in the figure, the short-time picture that was previously coded is maintained. Since the pictures stored in the short-term buffer are present in the second scene set A, the scene cut picture P200 is coded by referring to the forward predictive coded pictures P50 and P120 stored in the picture holding buffer for a long time. Done.

Therefore, in principle, the bit rate should be increased by being coded by the intra mode because it is a scene cut picture, but since it refers to the picture stored in the long-term picture holding buffer, it is coded by the inter mode to code the bit. The rate will not increase.

Also, for example, when the forward predictive coded picture for the current coding is the partial scene transform picture P250, since the macro block in which the scene change has not occurred is coded by referring to the picture stored before, the short time picture is maintained. Coding is performed in the inter mode with reference to the forward predictive coded pictures P249, P248, and P247 stored in a short-term buffer.

In addition, the macro block in the region having the scene transformation is coded in the inter mode with reference to the forward predictive coded pictures P50 and P120 stored in the long-term picture holding buffer.

 If the sub-sequential forward-prediction coded picture of the bi-prediction coded picture to be coded is neither a scene cut picture nor a partial scene transform picture, the sub-sequential forward predictive coded picture is coded by referring to a picture stored in the short-term picture maintaining buffer previously stored.

However, coding is performed separately from the pictures stored in the short-term picture maintaining buffer in the scene cut region of the sub-sequence forward-coded picture of the bidirectional predictive-coded picture to be coded, or when not.

First, when some of the pictures stored in the short-term picture maintaining buffer to be referenced by the sub-sequence forward-prediction coded picture to be coded exist in another scene set, a picture belonging to another scene set is a long picture. Coding is performed by referring to a picture stored in a long-term buffer.                     

For example, in the case of limiting the number of reference picture frames to five in the video coding system, as shown in FIG. 2, P202 in scene set B without scene transformation is stored in the short-term picture holding buffer P201, P200, Reference is made to P199, P198, P197, etc. Of these, since only P201 and P200 belong to the same scene set B, only two of them are coded.

Coding efficiency can be improved by referring to pictures P50 and P120 stored in a picture sustain buffer for a long time in order to comply with the five reference picture limitations. Therefore, as a result, since P202 refers to four pictures stored in the short-term picture retention buffer and the long-term picture retention buffer, the coding efficiency is improved.

Second, when all pictures stored in the short-term picture retention buffer to be referenced by the sub-sequence forward-prediction coded picture to be coded exist in the same scene set, a long-term picture retention buffer Note that only the pictures stored in the short-term picture retention buffer need to be coded with reference to the reference.

As described above, when coding of the forward predictive coded picture of the subsequence is completed, bidirectional predictive coded picture coding is performed.

3 is a diagram for describing a bidirectional predictive coded picture coding method of a sequence for scene transformation according to the present invention.

As shown in FIG. 3, first, a long-term picture retention buffer (long-term buffer) or a short-term picture retention of a subsequential forward-prediction coded picture is coded before coding the coded forward-prediction picture. Detects whether the picture stored in the buffer (short-term buffer) is referenced.

If the subsequence forward-prediction coded picture is coded with reference to a picture stored in the long-term picture retention buffer, when coding the bidirectional predictive-coded picture to be coded, the long-term picture retention buffer is coded. Coding by intra prediction mode, forward prediction mode, backward prediction mode, bidirectional prediction mode, and direct prediction mode is performed including the stored picture.

If the subsequence forward predictive coded picture obtains a motion vector from a picture stored in a long-term picture holding buffer, in the direct prediction mode, a forward motion vector is used to determine the motion vector of the subsequence forward predictive coded picture. Forward motion compensation is used as it is and reverse motion compensation is performed by setting the backward motion vector to zero. Then, finally, two motion compensation values are averaged to predict the prediction value of the macro block.

4 is a diagram for describing a method of inserting and coding two bidirectional predictive coded pictures according to the present invention.

As shown in FIG. 4, when the sub-sequence forward-prediction coded picture P200 detects a scene transformation from the encoder unit and performs coding, the sub-sequential forward-prediction coded picture P200 is stored in a long time picture holding buffer (long-term buffer). If the motion vector is obtained from the picture, when the bidirectional predictive coded picture is B1, as shown in FIG. 3, P200 uses the forward predictive coded picture stored in the picture sustain buffer for a long time.

Accordingly, the direct prediction mode obtains a forward motion vector from the picture P50 stored in the long time picture sustain buffer (long term buffer) and compensates for forward motion. In addition, a backward motion vector and backward motion compensation are performed from the sub-sequence forward-prediction coded picture P200 to obtain a prediction value of a macroblock.

However, since B ₁ is included in a scene set called P200, it will finally select a forward prediction mode in which most macro blocks are motion compensated from the short-term picture holding buffer.

In contrast, when the bidirectional predictive coded picture is B ₂ , as shown in FIG. 3, when it is determined that the sub-sequence forward predictive coded picture P200 uses the picture P50 stored in the long-term buffer, the direct In the prediction mode, a motion vector is obtained from P50, a backward motion vector is obtained from the subsequential forward prediction coded picture, and motion compensation values are averaged from both directions to predict a prediction value of a macroblock.

As shown in FIG. 4, since the bidirectional predictive coded picture B2 exists in the same scene set B as P50 and P200, the direct prediction mode will be selected as the final prediction mode. That is, since the information to be coded is not required for the coded picture, the bit rate can be reduced, thereby improving coding efficiency. In particular, in the video coding method, the forward motion vector and the reverse motion vector are obtained as follows.

In general, the formula for MV _f and MV _b is as follows.

MV _f = (N ₁ * MV) / D ....... (Equation 1)

MV _b = (N ₂ * MV) / D ....... (Equation 2)

Here, MV is a motion vector obtained by the subsequence forward-prediction coded picture (P200) of the bidirectional predictive coded picture to be coded from the stored picture referred to in the long-term picture retention buffer, and MV _f is a long-term picture retained from the bidirectional predictive coded picture to be coded. A forward motion vector obtained in the direct prediction mode up to the reference picture stored in the buffer, and MV _b represents a backward motion vector in the direct prediction mode from the bidirectional predictive coded picture to be coded to the subsequence forward predictive coded picture.

N ₁ = TR _B , N ₂ = TR _B -TR _D , and D = TR _D. The time distance between the reference picture referred to by the TR _D subsequence prediction coded picture and the subsequence forward prediction coded picture, and TR _B is the time from the reference picture referred to by the subsequence forward prediction coded picture to the current bidirectional predictive coded picture. Indicates distance.

However, the TR _D and the TR _B become meaningless when the picture where the scene change has occurred and the picture referred to by the picture are pictures stored in the picture sustain buffer for a long time. Because TR _D and TR _B are generally the forward and reverse motion vectors in the direct prediction mode, the sub-sequential forward prediction coded picture is inserted in consideration of the continuous motion between the sub-frame forward prediction coded picture and the picture frame referred to by the sub-sequence forward prediction coded picture. This is because it is determined that there will still be motion even in the bidirectional predictive coded picture, thereby inducing a temporal distance between the two pictures.

It is true that the motions of two pictures exist when referring to a picture stored in a long-term picture retention buffer due to scene transformation. Cannot be obtained.

Therefore, in the present invention, it is assumed that the similarity with the subsequence forward predictive coded picture is very high when coding the bidirectional predictive coded picture, and the forward and reverse motion vectors of the direct prediction mode are approximated to MV and 0, respectively.

Therefore, when applied to video coding, when a picture that has undergone scene transformation and a picture stored in the picture holding buffer for a long time are referred to, MV _f and MV _b are expressed as follows using Equations 1 and 2 above. Lose.

Here, after determining N ₁ = 1, N ₂ = 0, and D = 1 and transmitting the same to the decoder, the forward and reverse motion vectors may be obtained in the direct mode.

MV _f = MV ......... (Equation 3)

MV _b = 0 .......... (Equation 4)

Next, when a macroblock at the same position as the bidirectional predictive coded picture present in the subsequential forward predictive coded picture is obtained from a picture stored in a short-term picture holding buffer, a general method is used. In accordance with Equation 1 and Equation 2, direct forward and reverse motion vectors are obtained, and then the predicted value of the macroblock is obtained through an average value or interpolation of motion compensation values in each direction.

In general, the sub-sequence forward-prediction coded picture is generated when a picture without scene transformation or a picture without partial scene transformation.

If the subsequence forward predictive coded picture is a partial scene transformed picture, motion compensation is performed from the picture stored in the long-term buffer in the region where the scene change occurs, but without scene change. In the region, motion compensation will be made from the short-term picture holding picture. Therefore, when the macroblock prediction value is obtained through the direct prediction mode of the bidirectional predictive coded picture, the forward motion vector of the direct prediction mode is referred to the picture from the picture stored in the long-term picture retention buffer or the short-term picture retention buffer in the direct mode. Find the motion vector.

As described in detail above, the present invention improves the coding efficiency of a bidirectional predictive coded picture by enabling inter mode coding when a scene cut picture or a partial scene transform picture occurs in a subsequence forward predictive encoded picture. It is effective.

In addition, after coding a sub-sequence forward-prediction coded picture when there is a scene transformation, when coding a bi-prediction coded picture, many direct prediction modes are selected, thereby reducing the coding bit rate.

The present invention is not limited to the above-described embodiments, and various changes can be made by those skilled in the art without departing from the gist of the present invention as claimed in the following claims.

Claims (5)

An encoding method for generating a bit stream including a bidirectional predictive encoding macroblock, When encoding the bidirectional predictive coded macroblock, the prediction mode of the bi-predictive coded macroblock, If the picture referred to by the macroblock existing at the same position of the bidirectional predictive encoded macroblock is a long term reference picture, Obtaining a motion vector of a macroblock existing at the same location; Determining a first motion vector of the bidirectional predictive encoded macroblock as the motion vector of the obtained co-located block, and determining a second motion vector as zero; And Setting the direct prediction mode to be decoded by a decoding method comprising decoding the bidirectional predictive encoded macroblock using the determined first and second motion vectors of the bidirectional predictive encoded macroblock. Characterized by the encoding method. delete delete delete delete

Images