KR20000021867A - Method for encoding motion vector of binary form signal - Google Patents

Abstract

PURPOSE: A method for encoding a motion vector of a binary form signal is provided to encode a moving vector of the upper and the lower field BAB real-time. CONSTITUTION: A method for encoding a motion vector of a binary form signal includes six procedures. A first procedure is to decide a search area of P Q to the present field BAB depending on a before field signal recomposed and the present signal when the present field BAB to encode is applied. A second procedure is to extract an MV to the present field BAB and generate a field recognition flag corresponding to the MV as performing a block matching between the present field BAB and a large number of spare field BAB in the decided search area and deciding one among the large number of spare field BAB as an optimum junction field BAB. A third procedure is to check whether the present field BAB is the upper field BAB or the lower field BAB. A fourth procedure is to provide from MVs, which is in advance encoded and adjacent to the present field BAB, of a large number of the present field BAB and a texture field signal encoder, and decide an MVP(Moving Vector Pre-surmiser) to an MV of the present field BAB using the in advance encoded MV of a large number of the present texture field block which is adjacent to the present texture field block in a position corresponding to the present field BAB, when the present field BAB is the upper field BAB. A fifth procedure is to decide MV of the upper field BAB in a position corresponding to the lower field BAB as the MVP to the MV of the lower field BAB when the present field BAB is the lower field BAB. The sixth is to calculate the MVP for the present field BAB through a reduction between the MV of the present field BAB and the decided MVP and encode the generated field recognition flag and the calculated MVD.

Description

Motion vector coding method of binary shape signal

The present invention relates to a technique of encoding a motion vector of a binary shape signal. More particularly, the present invention relates to extracting a motion vector (MV) on a block-by-block basis for shape information of a perturbed image obtained by dividing a shape frame into an upper field and a lower field. The present invention relates to a technique for encoding a motion vector of a binary shape signal suitable for encoding.

In digital broadcast reception systems such as video telephony, teleconferencing and high definition television systems, the video line signal in the video frame signal includes a sequence of digital data represented as pixel values, so that each video frame signal can The definition requires a large amount of digital data. However, the available frequency bandwidth of a typical transmission channel is limited, so that in order to transmit large amounts of digital data to low bit rate video signal encoders, especially used in video telephony and teleconference systems, It is essential to compress or reduce the volume of data using various data compression techniques.

In a low bit rate coding system, one of several techniques for encoding video signals is a technique called object-oriented analysis-synthesis coding technique, in which the input video image is divided into objects, each object Three sets of parameters (parameters) for defining the motion, the contour and the pixel data of are processed through different coding channels due to their signal characteristics.

One embodiment of such an object-oriented analysis-synthesis coding scheme is referred to as MPEG (Video Expert Group) -4, which provides low bit rate communication, interactive multimedia (eg, games, interactive TV, etc.) and It is intended to provide audio-video encoding standards to allow content-based interactivity, improved coding efficiency and / or universal accessibility in applications such as surveillance equipment.

According to MPEG-4, the input video image is divided into a number of Video Object Planes (VOP's), which correspond to entities that are bitstreams that are accessible and manipulated by the user. The VOP is referred to as an object and is represented by a bounding rectangle, which is the smallest of the rectangles whose width and height are integer multiples of the 16 pixels (macro block size) that surround the object, so that the encoder processes the input video image in units of VOP. can do.

The VOP described in MPEG-4 includes shape information and color information consisting of luminance and chrominance data, wherein the shape information represented by binary shape signals is an alpha plane. It is referred to as (alpha plane). The alpha plane is divided into a number of binary alpha blocks (BABs) each having 16x16 binary pixels. Each binary pixel is classified as either a background pixel or an object pixel and is used to assign an object pixel within the object to have a different binary pixel value, for example a value of 255, while a background pixel located outside the object in the alpha plane. Is used to assign a binary pixel value, for example a value of zero.

On the other hand, each binary pixel in BAB is encoded using a conventional bitmap based shape coding method such as a context based arithmetic encoding (CAE) method. That is, in the intra mode, all the binary pixels of the BAB in the current frame (or VOP) are encoded using the intra CAE method to generate an intra-coded BAB. In the intra CAE method, the binary of the BAB in the current frame The context value of the pixel is obtained using the context value of the binary pixel around the binary pixel of BAB in the current frame (or VOP).

On the other hand, in the inter mode, all the binary pixels of the BAB in the current frame (or VOP) are encoded using the inter CAE method to generate an inter coded BAB. In the inter CAE method, the binary of the BAB in the current frame is generated. The context value of the pixel is obtained using the context value of the binary pixel around the binary pixel of BAB in the current frame (or VOP) and the context value of the binary pixel in the previous frame (see MPEG-4 Video Verification Model Version 7.0). , International Organization for Standardization, Coding of Moving Pictures And Associated Audio Information, ISO / IEC JTC1 / SC29 / WG11 MPEG97 / N1642, Bristol, April 1997, pp 28-30)

On the other hand, in the conventional binary shape signal coding scheme, instead of encoding and transmitting all binary pixels in a BAB to improve coding efficiency, a mode signal or a mode signal and a coded binary signal representing or characterizing an encoding condition for each BAB are encoded. A method of transferring pixel values is employed. For example, instead of encoding binary pixel values of object pixels to generate coded binary pixel values that should be transmitted when all binary pixels in BAB are object pixels, indicating that all binary pixels in BAB are object pixels. It is desirable to encode and transmit only the mode signal. In this way, coding efficiency can be enhanced by transmitting a corresponding encoding mode signal such as binary shape information for BAB.

According to the conventional mode encoding method, in encoding BAB, a mode signal indicating one of seven different modes as shown in Table 1 below is assigned to each BAB, and this is designated as the corresponding corresponding mode signal encoded. Alternatively, the signal is encoded and transmitted.

BAB tybe (mode) coding condition used in 0 MVD = 0 & no update P, B vop One MVD = 0! & no update P, B vop 2 all_0 all vop 3 all_255 all vop 4 intra CAE all vop 5 MVD = 0 & inter CAE P, B vop 6 MVD! = 0 & inter CAE P, B vop

Referring to Table 1, there are seven modes corresponding to binary alpha information of BAB according to a conventional mode encoding method, and mode 0 is defined as 0 as a motion vector predictor (MVP) of a shape for BAB. All binary pixels in BAB need not be encoded, mode 1 means that motion vector difference (MVD) is not defined as 0 and all binary pixels in BAB do not need to be coded, and mode 2 means that Means that all binary pixels are defined as background pixels, mode 3 means that all binary pixels in BAB are defined as object pixels, mode 4 means that all binary pixels in BAB are encoded by intra CAE method, and mode 5 Means MVD is defined as 0 and all binary pixels in BAB are encoded by the inter CAE method, and mode 6 does not define MVD as 0. This means that all binary pixels in the BAB are encoded by the inter CAE method.

Here, the motion vector difference (MVD) for the shape of the frame-based BAB is the motion vector (MV) for the shape of the current BAB and the motion vector predictor (MVP) for the current BAB or N × of the position corresponding to the current BAB. The motion vector predictor (MVP) of the N texture block, i.e., the current neighboring the texture block at the position corresponding to the current BAB or the MVP selected from among the motion vectors of a plurality of neighboring neighboring BABs in the current shape frame adjacent to the current BAB. It represents the difference between the selected MVP among the motion vectors of a plurality of sub-encoding adjacent texture blocks in the texture frame, which MVP is determined using conventional motion determination techniques (MPEG-4 Video Verfication Model Version 7.0, International Organization for Standardization, Coding of Moving Pictures And Associated Audio Information, ISO / IEC JTC / SC29 / WG11 MPEG97 / N1642, Bristol, April 1997, pp 20-23)

In this case, the transmission of the motion vector difference (MVD) between the MV and the MVP of the current BAB as the motion vector for the current BAB based on the frame is to reduce the coding bit rate allocated to the motion vector for each BAB. It can also be used in texture information encoding of the parallel encoding.

More specifically, when the motion vector of the current BAB is extracted in the conventional method, as shown in FIG. 5A, for example, the movement of a plurality of preset BABs (frame-based adjacent BABs) in a current shape frame adjacent to the current BAB. Any one of the vector predictors (MV _S 1, MV _S 2, MV _S 3), or as an example, as shown in FIG. 5B, a preset multiple in the current texture frame adjacent to the texture block at the location corresponding to the current BAB. One of the motion vector predictors (MV _T 1, MV _T 2, MV _T 3) of the adjacent texture block of is determined as the final MVP corresponding to the motion vector (MV) of the current BAB, and determined with the MV of the current BAB. MVD between MVPs is transmitted as motion vector information of the current BAB. At this time, if there is no MVP for the current BAB, the MV of the current BAB is transmitted as it is.

Here, the method for determining the MVP corresponding to the MV of the current BAB is to search for MV _S 1, MV _S 2, and MV _S 3 in the current shape frame in order to determine whether the MVP exists in one of three adjacent BABs. If there is a check, the MV of the neighboring BAB is determined as the MVP. If the check result does not exist, the MV _T 1, MV _T 2, and MV _T 3 in the current texture frame are searched one by one in a similar manner. It is checked whether MVP exists in the network, and if there is, the MV of the adjacent texture block is determined as MVP.

On the other hand, the adaptive texture encoding employs an adaptive encoding technique that inter- or intra-mode encodes a texture image based on a frame, or inter- or intra-codes a texture frame by dividing the texture frame into upper and lower fields. When inter-mode coding is performed in units of frame-based N × N frame blocks (16 × 16 frame blocks), as in the aforementioned frame-based shape coding scheme, one MV exists for each N × N frame block, but the texture image is generated. In case of inter-mode encoding in units of field-based M × N field blocks (8 × 16 field blocks), two MVs per N × N frame block, that is, MVs for upper and lower field blocks, respectively exist.

However, at present, there has not been proposed an efficient coding scheme in which the motion vector (MV) coding of the upper and lower fields BAB with respect to the perforated shape can be performed in real time while maintaining efficient coding performance.

The present invention has been devised in view of the above-mentioned point, and by referring to the correlation between the two field BAB with respect to the motion vector of the upper and lower field BAB generated when encoding the binary binary shape signal, the motion vector of the upper and lower field BAB It is an object of the present invention to provide a motion vector encoding method of a binary shape signal capable of encoding a signal in real time.

In order to achieve the above object, the present invention provides a motion vector (MV) of the upper and lower field BAB in a parallel shape signal obtained by dividing an N × N frame BAB having a binary pixel into an upper and lower field BAB of M × N. A method of encoding, comprising: a first step of determining a search region of P × Q for the current field BAB based on a reconstructed previous field signal and a current field signal when a current field BAB for encoding is applied; By performing block matching between the current field BAB and a plurality of candidate field BABs in the determined search area, one of the plurality of candidate field BABs is determined as an optimal matching field BAB, thereby extracting an MV for the current field BAB, and A second step of generating a corresponding field identification flag; A third step of checking whether the current field BAB is an upper field BAB or a lower field BAB; When the current field BAB is the upper field BAB, the current texture field at a position corresponding to the current field BAB provided from the MVs and texture field signal encoders of the base coded multiple current field BAB adjacent to the current field BAB A fourth process of determining a motion vector predictor (MVP) for the MV of the current field BAB by using MVs of a plurality of sub-encoding current texture field blocks adjacent to the block; When the current field BAB is the lower field BAB, determining the MV of the upper field BAB in the upper field signal at the position corresponding to the lower field BAB as a motion vector predictor (MVP) for the MV of the lower field BAB. 5 courses; And a sixth process of calculating an MVD for the current field BAB by subtracting the MV of the current field BAB and the determined MVP, and encoding the calculated MVD and the generated field identification flag. Provides a vector encoding method.

1 is a block diagram of a motion vector encoding apparatus suitable for applying a method of encoding a motion vector of a binary shape signal according to a preferred embodiment of the present invention;

FIG. 2 is a diagram for explaining a process of estimating a motion based on a search area set for each of upper and lower fields BAB;

3 illustrates an example of a motion predictor of a plurality of base coded adjacent field BABs used for MVD calculation for a current field BAB and a motion predictor of a plurality of base coded field texture blocks adjacent to a current field texture block. .

<Description of the code | symbol about the principal part of drawing>

102: field memory 104: search area decision block

106: motion estimation block 108: MV memory

110: MVD calculation block 112: MVD coding block

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

1 is a block diagram of a motion vector encoding apparatus suitable for applying a method of encoding a motion vector of a binary shape signal according to a preferred embodiment of the present invention, which includes a field memory 102, a search region determination block 104, A motion estimation block 106, an MV memory 108, an MVD calculation block 110, and an MVD coding block 112.

Referring to FIG. 1, the field memory 102 stores a reconstructed field signal, for example, a reconstructed 8x16 previous upper and lower field signal, provided from a not shown parallel shape information encoder. The previous upper and lower field signals stored in the field memory 102 are thus provided to the search region determination block 104 via line L12 and to the motion compensator, not shown, for the generation of the predictive field signals through motion compensation. do. In this case, since the separated two field signals have an order in which the upper field signal is first encoded and then the lower field signal is encoded, in the field memory 102, as the encoding of the shape information proceeds, the reconstructed upper field signal and the lower field signal are processed. The field signal is updated alternately.

Next, the search area determination block 104 determines the corresponding P × Q search area when the current field BAB (upper field BAB or lower field BAB) is provided via line L11, ie, the current field BAB is the upper field BAB. P x Q search area in the previous upper field and P x Q search area in the previous lower field, and extract each of these determined search area data from the field memory 102 and then present the current on line L11. A number of previous field BABs (eg, 8 × 16 previous field BABs) corresponding to field BAB are generated and provided to motion estimation block 106 via line L13. Further, the search area determination block 104 determines the P × Q search area in the current upper field and the P × Q search area in the previous lower field when the current field BAB is the lower field BAB, and each of these determined angular searches. After retrieving the area data from the field memory 102, a plurality of previous field BABs (e.g., 8x16 previous field BABs) corresponding to the current field BAB on line L11 are generated to move the motion estimation block through line L13. Provided at 106.

For example, assuming that the current field BAB on line L11 is the upper field BAB 312 in the upper field 310 of the current frame 300, as shown in FIG. The previous upper search area 414 in the previous upper field 410 of the reconstructed previous frame 400 and the previous lower search area 426 in the previous lower field 420 of the previous frame 400 are present. Determine as the search area for the top field BAB 312, and also assume that the current field BAB on line L11 is the bottom field BAB 322 in the bottom field 320 of the current frame 300, as shown in FIG. In the search region determination block 104, the current upper search region 316 and the previous lower field of the previous frame 400 reconstructed in the current upper field 310 of the current frame 300 encoded immediately before. Search previous lower search area 428 within 420 for current lower field BAB 322. The decision reversed.

Meanwhile, in the motion estimation block 106, a search area (previous upper search area and previous lower search area or current upper search area and previous lower area) provided from the search area determination block 104 through the current field BAB and line L13 on the line L11. Block matching between the plurality of candidate field BABs in the search area) to determine an optimal candidate field BAB having the smallest error value among the plurality of candidate field BABs as an optimal matching block, and determine the optimal matching BAB determined with the current field BAB. Calculate a motion vector (MV) representing the displacement of the liver and generate a field identification flag, the MV of the current field BAB thus calculated is provided to the MV memory 108 and the MVD calculation block 110 via line L14, and field identification The flag is provided to the MVD calculation block 110 via line L14.

In addition, the MV and field identification flags on line L14 are provided to a motion compensator, not shown, for generation of the predictive field signal through motion compensation. Here, the motion vectors MV for the current field BABs, which are provided to the MV memory 102 via the line L14, are motions for the motion vector MV of each field BAB of the next field signal, which is encoded following the current field signal. It will be used as a vector predictor (MVP).

That is, the MV memory 108 stores motion vectors MV for each field BAB of the current upper and lower field signals provided through the line L14, and also provides a texture field block (e.g., a texture encoder provided from a texture encoder not shown). For example, motion vectors MV _T of 8 × 16 texture field blocks) are stored.

On the other hand, the MVD calculation block 110 performs a method for determining MVP for MVD calculation of the upper field BAB and a method for determining MVP for MVD calculation of the lower field BAB.

That is, when the MV of the current upper field BAB is provided from the above motion estimation block 106 via the line L14, the MVD calculation block 110 searches for the MV memory 108, as shown in FIG. 3A as an example. Similarly, MV _S 1, MV _S 2, and MV _S 3, MVs of the base coded adjacent field BABs in the current field signal adjacent to the current field BAB, are sequentially searched and one of them is a motion vector for the MV of the current field BAB. If it is determined by the predictor (MVP) and MVP is not determined among MV _S 1, MV _S 2, and MV _S 3, the texture field in the texture field signal at the position corresponding to the current field BAB as shown in FIG. 3B. MVs of MV _T 1, MV _T 2, MV _T 3 of the base coded adjacent texture field blocks adjacent to the block are sequentially searched to determine one of them as the motion vector predictor (MVP) for the MV of the current field BAB. do.

Further, when the MV of the current lower field BAB is provided from the above motion estimation block 106 via the line L14, the MVD calculation block 110 converts the MV of the current upper field BAB corresponding to the current lower field BAB into the current lower field. Decide immediately as the MVP for BAB's MV. That is, in the MVD calculation block 110, if the MV of the current upper field BAB is provided, the MVP is determined based on the MVs of the plurality of neighboring field BABs and the MVs of the plurality of neighboring texture fields, and the MV of the current lower field BAB. If is provided, the MV of the corresponding upper field BAB is immediately determined as MVP.

Then, the MVD calculation block 110 calculates the MVD for the MV of the current field BAB by subtracting between the MV of the current field BAB and the determined MVP, and the calculated MVD and the field identification flag of the current field BAB are the next MVD. It is provided to the coding block 112.

Therefore, in the MVD coding block 112, the MVD and field identification flags provided from the MVD calculation block 110 are encoded and transmitted to the multiplexer (not shown).

As described above, according to the present invention, in the perforated shape encoding, one of the MVs of the neighboring current upper field or the base-coded multiple of the corresponding position in the MVs of the current upper field BAB Determine one of the MVs of adjacent current texture field blocks as the MVP for the MV of the current upper field BAB, and determine the MV of the corresponding current upper field BAB as its MVP for the MVs of the current lower field BAB. Thus, by calculating the corresponding MVD, it is possible to effectively realize the encoding of the motion vectors MV of the upper and lower fields BAB of the perforated shape in real time.

Claims (2)

A method of encoding a motion vector (MV) of the upper and lower field BAB in a parallel shape signal obtained by dividing an N × N frame BAB having binary pixels into an upper and lower field BAB of M × N, A first step of determining a PxQ search region for the current field BAB based on a reconstructed previous field signal and a current field signal when a current field BAB for encoding is applied; By performing block matching between the current field BAB and a plurality of candidate field BABs in the determined search area, one of the plurality of candidate field BABs is determined as an optimal matching field BAB, thereby extracting an MV for the current field BAB, and A second step of generating a corresponding field identification flag; A third step of checking whether the current field BAB is an upper field BAB or a lower field BAB; When the current field BAB is the upper field BAB, the current texture field at a position corresponding to the current field BAB provided from the MVs and texture field signal encoders of the base coded multiple current field BAB adjacent to the current field BAB A fourth process of determining a motion vector predictor (MVP) for the MV of the current field BAB by using MVs of a plurality of sub-encoding current texture field blocks adjacent to the block; When the current field BAB is the lower field BAB, determining the MV of the upper field BAB in the upper field signal at the position corresponding to the lower field BAB as a motion vector predictor (MVP) for the MV of the lower field BAB. 5 courses; And A motion vector of a binary shape signal comprising a sixth process of calculating an MVD for the current field BAB by subtracting the MV of the current field BAB and the determined MVP, and encoding the calculated MVD and the generated field identification flag Coding method. The method of claim 1, wherein the first process comprises: When the current field BAB is the upper field BAB, the P × Q search region at the corresponding position in the previous upper field signal and the P × Q search region at the corresponding position in the previous lower field signal are searched for the current field BAB. An eleventh process of determining the area; And When the current field BAB is a lower field BAB, the P × Q search region at a corresponding position in the current upper field signal and the P × Q search region at a corresponding position in the previous lower field signal are searched for the current field BAB. And an eleventh process of determining the region. The motion vector encoding method of the binary shape signal.