KR19990065274A - Shape Information Coding Method for Progressive Scan - Google Patents

Abstract

The present invention relates to a shape information encoding method for a parallel scan which detects a motion degree of shape information during encoding for a progressive scan of a video and encodes the shape information in a frame unit or a field unit according to the amount of motion. According to the present invention, in shape information encoding for a progressive scan, when determining a motion prediction direction of an odd field of a current frame, a minimum value of sum absorptive difference (SAD) estimated from an odd field of a previous frame is estimated. Determining the direction of motion prediction and determining the motion prediction direction of the even field of the current frame by performing the motion estimation to the minimum value of Sum Absolute Difference (SAD) that estimated the motion in the even field of the previous frame. Encoding shape information.

Description

Shape Information Coding Method for Progressive Scan

The present invention relates to shape information encoding for a progressive scan, and more particularly, to detect the degree of motion of shape information in encoding for a progressive scan of a video and to encode shape information in units of frames or fields according to the amount of motion. A shape information encoding method for row scanning.

In general, the MPEG-4 group has developed and decided on international standards on video and audio coding technology and system construction, and the next generation of video and audio coding technology and system construction to be adopted as an international standard. International Standard (MPEG-4) is being researched and developed. The development of MPEG-4 began with the need to support the next generation of video and audio applications that cannot be supported by existing known standards.

MPEG-4 provides new methods for communicating and accessing and manipulating video and audio data, such as object-oriented interactive functionality and access over networks of differing characteristics.

In addition, it provides a characteristic that operates usefully in a communication environment where errors are easily generated and a low transmission rate communication environment. In addition, it integrates computer graphics technology to provide the ability to encode and manipulate natural and artificial images and sounds together.

In summary, MPEG-4 must support all the features required and expected in many applications. As a result, the expansion and openness of multimedia information is possible to support the functions required for all possible low-cost, high-performance applications that are newly developed or will be developed. Among them are the transmission and storage functions and the improved compression efficiency required for cost reduction.

Applications currently expected to apply MPEG-4 technology include interpersonal communication (IPC) such as Internet multimedia (IMM), interactive video games (IVG), video conferencing and video telephony. Networked Database Services (NDB) using Communications, Interactive Storage Media (ISM), Multimedia Mailing (MMM), Wireless Multimedia (WMM), ATM networks, Remote Emergency Systems (RES) and Remote Video Surveillance (RVS).

In order to support existing applications or applications expected in the future, an image encoding technology that can be edited so that users can communicate only with the desired objects in the image, can be found and read, and can be cut and pasted can be edited.

MPEG-4, a new video and audio encoding technology that is currently under global standardization, is designed to meet this need.

FIG. 1 is a block diagram of a MPEG-4 VOP video encoder first determined by the present International Standards Organization.

This is different from the video encoder structure of H.261, H.263, MPEG-1, MPEG-2, which is the world standardized video encoding. In particular, the introduction of the concepts of Shape Coder and VOP (Video Object Planes) shows the most significant difference. The VOP refers to an object at a point in time on an arbitrary shape of content that can be accessed and edited by the user. The VOP must be encoded for each VOP to support content-based functionality.

When the VOP encoder inputs a VOP for each object image formed in the VOP forming unit 11 to the MOTION ESTIMATION 13, the motion estimating unit 13 performs the macroblock unit from the applied VOP. The motion is estimated.

In addition, the motion information estimated by the motion estimator 13 is input to a motion compensation unit 14 to compensate for the motion. The VOP whose motion is compensated by the motion compensator 14 is input to the subtractor 16 together with the VOP formed by the VOP forming unit 11 to detect a difference value, and the difference value detected by the subtractor 16 is The internal information of the object is encoded by the object internal encoding unit 18 in units of sub blocks of the macro block.

For example, after the X and Y axes of the macroblock are subdivided into 8x8 subblocks having 8 pixels each with M / 2 x N / 2, the object internal information is encoded.

On the other hand, the VOP whose motion is compensated by the motion compensator 14 and the internal information of the object encoded by the object internal encoder 18 are added to the adder 17, and the output signal of the adder 17 is transferred. The previous VOP which is input to the VEV detector 15 (PREVIOUS RECONSTRUCTED VOP) 15 and is the VOP of the image immediately before the current image is detected.

In addition, the previous VOP detected by the previous VOP detector 15 is input to the motion estimator 13 and the motion compensator 14 and used for motion estimation and motion compensation.

The VOP formed by the VOP forming unit 11 is input to a shape coding unit 12, and shape information is encoded.

Here, the output signal of the shape information encoder 12 may vary depending on the field in which the VOP encoder is applied, and as shown by a dotted line, the output signal of the shape information encoder 12 may be used as the motion estimation unit 13. ), And may be input to the motion compensator 14 and the object internal encoder 18 to be used for encoding motion estimation, motion compensation, and internal information of the object.

In addition, motion information estimated by the motion estimation unit 13, object internal information encoded by the object internal encoding unit 18, and shape information encoded by the shape information encoding unit 12 are applied to the multiplexer 19. After being multiplexed and multiplexed, the bitstream is transmitted to a multiplexer which multiplexes and outputs the outputs of the plurality of encoders, although not shown in the figure.

One of the biggest features of this concept of Moving Picture Expert Group-4 (MPEG-4) is that it is based on objects. In other words, one image can be divided into several objects and each object can be individually encoded and processed. So you need to know the shape information to make an object.

The shape information referred to above is commonly referred to as a mask. In the image, the object part is represented by '1' and the outside part of the object (background part) is represented by '0'. This shape information can be used to obtain an object from the image. Since the decoder decodes the object part using the shape information, the shape information should be encoded and transmitted to the decoder side.

Shape information for separating an object from an image, shape motion information and image information of the object must be included, and the information must be encoded and transmitted. Motion estimation is performed in units of blocks, and motion coding is performed using the motion vector generated therein, and then image coding is performed on a portion where motion compensation is not performed. Shape coding is performed on the shape information of the object.

In general, there are progressive and interlaced methods for transmitting a motion image. Progressive scanning encodes, transmits and displays images in units of frames.

However, a parallel scan divides one frame into a first field (top field) and a second field (bottom field) and encodes, transmits, and displays the image.

In this case, conventional interpolation scan transmits one frame into two fields, which causes motion on the time axis between two fields, causing a difference in image size by the size of the motion.

Therefore, the present invention has been proposed to solve all the problems occurring during the conventional fleeting scan,

The present invention provides a shape information encoding method for a progressive scan which detects a motion degree of shape information during encoding for a progressive scan of a video and encodes the shape information in a frame unit or a field unit according to the amount of movement thereof. There is a purpose.

Method according to the present invention for achieving the above object,

In shape information encoding for a progressive scan,

When determining the motion prediction direction of the odd field of the current frame, the minimum value of SAD (Sum Absolute Difference) that has been estimated in the odd field of the reference frame is determined to determine the motion prediction direction, and When determining the motion prediction direction of the even field, the shape information is encoded by performing motion estimation to a minimum value of SAD (Sum Absolute Difference) in which the motion estimation is performed in the even field of the previous frame.

Another method according to the present invention for achieving the above object,

In shape information encoding for a progressive scan,

One 16x16 frame motion vector or two 8x16 frame motion vectors are obtained by performing motion estimation on the binary alpha block (BAB) using the information of the previous frame, and then one if the MVDs are not 0. The shape information is encoded by determining shape_field_prediction (ON: field prediction, OFF: frame prediction), which is used as a 16x16 motion vector or two 8x16 motion vectors.

In the above, when shape_field_prediction is field prediction, 1 is transmitted, and when frame prediction is transmitted, 0 is transmitted.

In addition, if shape_field_prediction is field prediction, 0 is transmitted. If frame_prediction is frame prediction, 1 is transmitted.

Another method according to the present invention for achieving the above object,

In shape information encoding for a progressive scan,

One 16x16 frame motion vector or two 8x16 frame motion vectors are obtained by performing motion estimation on the binary alpha block (BAB) using the information of the previous frame, and then one if the MVDs are not 0. Determining shape_field_prediction (ON: field prediction, OFF: frame prediction), which is whether a 16x16 motion vector or two 8x16 motion vectors are used; Determining a BAB_type that is a shape information additional information mode; If the type of the binary shape information is All_0, All_255, or MVDs = 0No_update, performing encoding on the BAB; If the BAB type is not the three cases, determining CAE_type by first checking whether the CAB mode or the field CAE mode is used to perform selective interpolation scan shape information encoding; When the field CAE mode is selected, two 8x16 field BABs are encoded by applying the same BAB_type, and then the BAB_type is determined by determining IntraCAE or MVDs! = 0 No_Update or MVDs == 0 InterCAE or MVDs! = 0InterCAE. Characterized in that the decision making process is carried out sequentially.

In the above, the determination of whether the CAE_type is a field CAE or a frame CAE is characterized in that the CAE_Type is determined so that the number of bits generated after performing CAE on a field or frame basis is small.

In the above, only one ST (Scan_type) is used when the CAE_type is the field mode.

In the above, when CAE_type is in the field CAE mode, in order to increase coding efficiency, after generating the top field BAB without encoding the Flusih bit for each field BAB, the even field BAB is sent without sending the flush bit. After encoding, the flush bit is generated and transmitted only once.

In the above, BAB_type determination is performed in association with CR to determine whether All_0, All_255 is applied by applying CR to BAB of the original form in lossy coding, and CR is applied to relocated BAB (Permuted BAB). After determining whether or not, if either case is All_0 or All_255, BAB_type is characterized as All_0, All_255.

In the above, by performing BAB_type determination in association with CR, CR is applied to BAB in lossy coding, and MVDs == 0No_Update, MVDs! = 0 No_Update is determined, and CR is applied to relocated BAB (Permuted BAB). After deciding whether MVDs == 0 No_Update, MVDs! = 0 No_Update, if either case MVDs == 0No_Update or MVDs! = 0 No_Update Characterized in that.

1 is a block diagram of a MPEG-4 VOP video encoder primarily determined by the present International Standards Organization;

FIG. 2 is a flow chart of time frames and fields in progressive and interlaced scans. FIG.

3 is a flow chart of adaptive interlaced shape coding.

4 is an exemplary view of a direction of referring to a previous field when performing field prediction;

5 is a configuration diagram of a frame MB and a field MB to be encoded in units of macro blocks (MB);

FIG. 6 is a view showing an image with little motion when binary shape information Mask is present in FIG. 5;

FIG. 7 is a diagram illustrating an image having a lot of motion when binary shape information Mask is present in FIG. 5;

8 is a motion compensated binary shape block diagram (Bordered MC BAB) with edges represented by 16 × 16 macroblocks,

9 is a current binary ordered block diagram (Bordered Current BAB) showing the edges as 16x16 macroblocks,

10 is a flowchart of a process of determining a conversion ratio (CR) for a frame of an original form (in this case, one CR is determined),

FIG. 11 is a flow chart of reordering BAB after field reordering before determining CR, and determining a conversion ratio (CR) in this case (1 CR is determined in this case).

12 is a flowchart for a process of determining a conversion ratio (CR) for each of the top field BAB and the even field BAB (in this case, two CRs determined for each field BAB);

13 is a flowchart for determining a conversion ratio (CR) for a relocation BAB when CAE_type is in field CAE mode (in this case, one CR is determined),

FIG. 14 is a 1-bit Syntax configuration diagram of CAE_type added in FIG. 3; FIG.

15 is an additional motion vector syntax structure of 1 bit of CAE_type and 1 bit of shape_field_prediction of shape information motion prediction and Bottom field added in FIG.

16 is a variable length coding table of an existing motion vector;

Fig. 17 is a variable length coding table applied when the coordinate in the horizontal direction (x axis) of a motion vector is zero.

Explanation of symbols for the main parts of the drawings

11: VOP forming part

12: shape information encoder

13: Motion estimation part

14: Motion compensation department

18: object internal encoding unit

19: multiplexing unit

20: buffer

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

3 shows a flow diagram for selective encoding of the scan information.

One 16x16 frame motion vector or two 8x16 frame motion vectors are obtained by performing motion estimation on the binary alpha block (BAB) using the information of the previous frame, and then one if the MVDs are not 0. The shape_field_prediction (ON: field prediction, OFF: frame prediction) is determined whether the 16x16 motion vector or two 8x16 motion vectors are used.

Thereafter, BAB_type as the shape information additional information mode is determined.

The BAB_type is classified into All_0, All_255, MVDs = 0No_update, MVDs! = 0No_update, IntraCAE, MVDs = 0InterCAE, MVDs! = 0 InterCAE.

Here, if the shape information of the block BAB is all 0, it is defined as All_0. If the shape information of the block BAB is all 255, it is defined as All_255. The definition of MVDs = 0No_update means that the shape information motion vector of BAB is 0 and is not encoded because it is the same as the previous BAB. This is the case when the same code is not used.

Intra CAE refers to Intra CAE encoding of BAB. The definition of MVDs = 0InterCAE is when the shape information motion vector of BAB is 0 and it is not the same as the previous BAB. This is the case where Inter CAE coding is required because the information motion vector is not 0 and is not the same when the motion compensation is performed with the previous BAB.

All_0, All_255, MVDs = 0 No_update, in three cases, the BAB_type is encoded, so that the encoding for the BAB is terminated. If the above three are not performed, CAE_type is determined by first checking whether the frame CAE mode or the field CAE mode is used to perform selective encoding of the scan information.

There may be various methods for determining the CAE_type.

When the frame CAE mode is selected, after encoding in units of 16 × 16 BAB, BAB_type is determined by determining whether IntraCAE or MVDs! = 0No_Update or MVDs == 0InterCAE or MVDs! = 0InterCAE in the less coded bits.

When the field CAE mode is selected, the 16x16 BAB is divided into two 8x16 subblocks (hereinafter referred to as field BABs). The same BAB_type is applied to two 8x16 field BABs, and then IntraCAE or MVDs ! = 0 No_Update or MVDs == 0 InterCAE or MVDs! = 0InterCAE to determine BAB_type.

4 is a diagram illustrating a shape information motion estimation direction of an odd field and an even field of a current frame.

When determining the direction of motion estimation of the odd field of the current frame, the motion vector obtains the minimum value of sum absolute difference (SAD) by estimating the motion of the odd field of the reference frame. Determine.

In addition, when determining the motion prediction direction of the even field of the current frame, a motion is estimated in the even field of the previous frame to obtain a minimum value of the sum absolute difference (SAD). Determine the vector.

FIG. 5 is a diagram illustrating an image configuration of whether two fields are configured in one frame and then encoded in frame units or field units when encoded in a macroblock size, for example.

The white line represents the odd field and the gray line represents the even field.

FIG. 6 illustrates a case in which binary shape information is configured and encoded in a frame unit and a field unit when the moving image is a still image or a still image.

Since the change in time is very small, there is almost no change in image shape between the two fields. In this case, it is advantageous to encode shape information in units of frames. This is because binary shape information is not severely changed when encoded in frame units.

FIG. 7 illustrates a case in which binary shape information is configured and encoded in a frame unit and a field unit when an image having a lot of motions. Since there are many changes in time, the change in image shape between two fields is severe. In this case, it is advantageous to encode shape information in field units. This is because the binary shape information is not severely changed when coding in field units.

As shown in the examples of FIG. 6 and FIG. 7, the encoding efficiency of the interlaced appears differently as the motion is large and small. Optimum coding efficiency can be obtained by separately coding the frames and the fields. As the encoding method, CAE (Context-based Arithmetic Encoding) adopted in the international standard MPEG-4 CD (Committee Draft) is used.

The criterion that there is a lot of motion in the macroblock is small and can be determined by the following equation. As shown in Fig. 6, a pixel having binary shape information is set to 1, and a pixel having no binary shape information is set to 0.

| P _{2i, j} -P _{2i + 1, j} | + | P _{2i + 1, j} -P _{2i + 2, j} |) | P _{2i, j} -P _{2i + 2, j} | + | P _{2i + 1, j} -P _{2i + 3, j} |)

Where Pi, j is binary shape information data, i.e. 1 if shape information is present and 0 if it is present.

If the left side is large, the field CAE mode is selected. If the right side is large, the frame CAE mode is selected. In addition, whether the field CAE mode or the frame CAE mode is transmitted is transmitted as CAE_type (1 bit). CAE is performed in units of fields or frames according to CAE_type. Instead of using the above formula, the CAE_Type may be determined so that the number of bits generated after performing CAE on a field or frame basis is small.

The left figure of FIG. 8 shows a Bordered MC BAB (Bordered Motion Compensated Binary Alpha Block) showing the edge of a 16x16 macroblock. The thick line inside means a macroblock 16x16 to be encoded. The right figure of FIG. 8 shows that when the Field CAE mode is selected, the 8x16 odd field and the even field are divided, and the data of the edge of each 8x16 subblock is obtained from the part associated with each field. In other words, Bordered MC BAB is divided into two fields.

The left figure of FIG. 9 shows a Bordered Current BAB (Bordered Current Binary Alpha Block) showing the edge of a 16x16 macroblock. In the right figure of Fig. 9, when the Field CAE mode is selected, the 8x16 odd field and the even field are divided and the data of the edge of each 8x16 subblock is displayed in each field.

It is taken from the part related to In other words, Bordered Current BAB is divided into two fields. Only the top and left borders get their data from each field.

10, 11, 12, and 13 illustrate how the target BAB is configured to determine the conversion ratio (CR) in the shape information encoding process for lossy coding.

There are two ways to construct the target BAB.

Referring to FIG. 5, the original form of the BAB may be maintained as shown in the left figure, or the new type of BAB may be configured by rearranging and rearranging by field as shown in the figure on the right. Hereinafter, the BAB formed by collecting and rearranging by field is referred to as relocated BAB (Permuted BAB).

FIG. 10 is a block diagram illustrating a process of determining a conversion ratio (CR) for a frame having an original shape. In this case, one CR is determined.

FIG. 11 is a block diagram of relocation BAB after field reordering prior to determining the CR, and determining a conversion ratio (CR). In this case, one CR is determined.

12 is a block diagram of a process of determining a conversion ratio (CR) for an odd field BAB and an even field BAB, respectively. In this case, two CRs are determined for each field BAB.

FIG. 13 determines a conversion ratio (CR) for relocation BAB when CAE_type is in field CAE mode. If CAE_type is frame CAE mode, CR is determined for BAB. In this case, one CR is determined.

In addition to the CR determination method described above, a method of obtaining coding gain in lossy coding by performing BAB_type determination in association with CR is as follows.

After CR is applied to BAB of original form, All_0, All_255 is determined, CR is applied to relocated BAB (Permuted BAB), and All_0, All_255 is determined. If either case is All_0 or All_255, BAB_type Determine as All_0, All_255.

Apply CR to BAB to determine whether MVDs == 0No_Update, MVDs! = 0 No_Update and apply CR to Permuted BAB to determine whether MVDs == 0 No_Update, MVDs! = 0 No_Update If either of MVDs = 0 No_Update or MVDs! = 0 No_Update, BAB_type is determined as MVDs = 0 No_Update or MVDs! = 0 No_Update.

In addition, the following items should be considered to improve coding efficiency.

Only one ST (Scan_type) is used when CAE_type is in field mode. That is, the same ST is used for the odd field BAB and the even field BAB. This is because sending a total of two in each 8x16 field BAB unit increases the additional information, which causes a loss in bits.

In general, in case of CAE (Context Arithmetic Encoding), Arithmetic Coding is performed by using occurrence probability of each pixel in each coding unit (BAB unit), and then Flush bit (2-3 bits) is transmitted. In order to improve coding efficiency when CAE_type is in field CAE mode, after encoding the top field BAB without encoding the Flusih bit for each field BAB and encoding the even field BAB without sending the flush bit after encoding the top field BAB, Generate only once.

FIG. 14 shows the syntax position of additional information of CAE_type 1 bit when the motion vector is transmitted in one 16x16 vector.

Figure 15 shows the position of the syntax of the CAE_type 1 bit additional information, shape_field_prediction, and the position of the additional motion vector when the motion vector is transmitted as two 8x16 vectors.

FIG. 16 is a conventional variable length coding table of a motion vector, and FIG. 17 is a variable length coding table applied when a coordinate in the horizontal direction (x-axis) of a motion vector is zero.

The variable length coding table of FIG. 16 should be used as the variable length coding table of the motion vector of the x-axis and the y-axis when shape_field_prediction is ON. (At this time, even if the coordinate in the horizontal direction (x-axis) of the motion vector is 0. I use it)

As described above, the present invention has the effect of increasing the invalidation efficiency of shape information during the progressive scan by selecting the encoding mode according to the amount of motion of the shape information between the frame and the field when encoding the optional progressive scan shape information. There is.

Claims (10)

In shape information encoding for a progressive scan, When determining the motion prediction direction of the odd field of the current frame, the minimum value of SAD (Sum Absolute Difference) that has been estimated in the odd field of the reference frame is determined to determine the motion prediction direction, and When determining the motion prediction direction of the even field, the shape information is encoded by performing the motion estimation to the minimum value of SAD (Sum Absolute Difference) which estimates the motion in the even field of the previous frame. Shape information encoding method. In shape information encoding for a progressive scan, One 16x16 frame motion vector or two 8x16 frame motion vectors are obtained by performing motion estimation on the binary alpha block (BAB) using the information of the previous frame, and then one if the MVDs are not 0. Shape information is encoded for a perimeter scan by determining shape_field_prediction (ON: field prediction, OFF: frame prediction), which is whether a 16x16 motion vector or two 8x16 motion vectors are used. Way. The method of claim 2, And if the shape_field_prediction is field prediction, 1 is transmitted, and if frame shape is predicted, 0 is transmitted. The method of claim 2, And if the shape_field_prediction is field prediction, 0 is transmitted, and if frame shape prediction is 1, 1 is transmitted. In shape information encoding for a progressive scan, One 16x16 frame motion vector or two 8x16 frame motion vectors are obtained by performing motion estimation on the binary alpha block (BAB) using the information of the previous frame, and then one if the MVDs are not 0. Determining shape_field_prediction (ON: field prediction, OFF: frame prediction), which is whether a 16x16 motion vector or two 8x16 motion vectors are used; Determining a BAB_type that is a shape information additional information mode; If the type of the binary shape information is All_0, All_255, or MVDs = 0No_update, performing encoding on the BAB; If the BAB type is not the three cases, determining CAE_type by first checking whether the CAB mode or the field CAE mode is used to perform selective interpolation scan shape information encoding; When the field CAE mode is selected, two 8x16 field BABs are encoded by applying the same BAB_type, and then the BAB_type is determined by determining IntraCAE or MVDs! = 0 No_Update or MVDs == 0 InterCAE or MVDs! = 0InterCAE. A shape information encoding method for a percussion scan, characterized in that to execute the determining process sequentially. The method of claim 5, And determining whether the CAE_type is a field CAE or a frame CAE to determine the CAE_Type in the number of bits generated after performing CAE on a field or frame basis. The method of claim 5, And only one ST (Scan_type) is used when the CAE_type is the field mode. The method of claim 5, When the CAE_type is the field CAE mode, in order to increase coding efficiency, after generating the top field BAB without encoding the Flusih bit for each field BAB, after encoding the even field BAB without sending the flush bit, Shape information encoding method for a perimeter scan, characterized in that to generate the flush bit only once. The method of claim 5, The BAB_type determination is performed in association with CR to determine whether All_0, All_255 is applied by applying CR to BAB of the original form in lossy coding, and CR is applied to relocated BAB (Permuted BAB) to determine whether All_0, All_255 And then determine if BAB_type is All_0 or All_255 if either of them is All_0 or All_255. The method of claim 5, The BAB_type decision is performed in association with CR to determine whether MVDs = 0 No_Update, MVDs! = 0 No_Update by applying CR to BAB when lossy coding and MVDs by applying CR to relocated BAB (Permuted BAB). == 0 No_Update, MVDs! = 0 After deciding whether or not to update, if either case MVDs == 0No_Update or MVDs! = 0 No_Update, BAB_type is determined to be MVDs == 0No_Update or MVDs! = 0 No_Update. A shape information encoding method for a percussion scan.