KR100255756B1 - Method for coding motion vector - Google Patents

Abstract

PURPOSE: A motion vector encoding method is provided to prevent the deterioration of encoding efficiency of a motion vector by performing a dispersion considering that two are similar among prediction candidate data applied. CONSTITUTION: 3 estimation candidate data and a median filter apply a median vector of 3 estimation candidate data(201). The smallest motion vector(MVPs) and the largest motion vector(MVPb) are respectively detected from the 3 estimation candidate data on the basis of the median vector(202). An absolute value(A) of a difference between the median vector and the smallest motion vector is calculated(203). An absolute value(B) of a difference between the largest motion vector and the median vector is calculated(204). An is added to B, therefore C is calculated(205). A selection signal generation unit compares C with a determined threshold(206). A motion vector encoder processes a motion vector encoding according as the degree of dispersion is excessive(207). A motion vector encoding is processed according as it is judged that the degree of dispersion is not excessive(208).

Description

Motion vector coding method {METHOD FOR CODING MOTION VECTOR}

The present invention relates to a method of encoding a motion vector. In particular, the present invention relates to a method for encoding a motion vector and determining a predictor according to a degree of variance of motion vectors of candidate blocks. The present invention relates to a motion vector encoding method for improving compression efficiency.

In a conventionally known digital television system, a video line signal of a video frame signal includes a series of digital data called pixel values, and therefore a considerable amount of digital data is required to define each video signal. On the other hand, since the bandwidth of a commonly used transmission channel is limited, it is impossible to transmit a large amount of data. Therefore, various data compression techniques have been proposed to transmit a large amount of data through a transmission channel having a limited bandwidth. Among the currently proposed compressor methods, a so-called mixed coding method using a temporal and spatial compression method together with a stochastic coding method is known to be efficient.

Most hybrid coding techniques use motion-compensated differential pulse code modulation (DPCM), two-dimensional discrete cosine transform (DCT), quantization of DCT coefficients, and variable length coding (VLC). The motion compensation DPCM estimates the motion of the object between the current frame and the previous frame, predicts the current frame according to the motion of the object, and generates a differential signal representing the difference between the current frame and the prediction value.

In particular, in motion compensation DPCM, current frame data is predicted from previous frame data based on motion estimation between the current and previous frames. The motion estimation result of the predicted frame data is described as a two-dimensional motion vector representing the displacement of the pixel between the current and previous frames. That is, it is described as a motion vector expressed in horizontal and vertical components.

FIG. 1 is a block diagram of an existing apparatus for encoding such a motion vector. The median vector of neighboring candidate blocks is used as a predictor to perform DPCM with motion vectors of a search block. The method of performing differential coding with the motion vector of the search block by using the motion vector having the closest value to the currently applied motion vector among the method and the predicted candidate data selectively according to the variance of the currently applied predictive candidate data. Implemented device.

That is, in order to encode the currently applied motion vector, the apparatus shown in FIG. 1 calculates the degree of variance of the predicted candidate data to be applied. The candidate candidate data applied at this time are motion vectors of candidate blocks suitable for a model set among candidate blocks adjacent to the search block of the currently applied motion vector. For example, if the set model is an MPEG-4 Verification Model, blocks located on the left, top, and top right centers of the search block to be encoded are suitable candidate blocks. When the dispersion degree of these candidate blocks is obtained, the signal is transmitted to the selection signal generator 120.

The selection signal generator 120 compares the predetermined threshold value with the dispersion degree value transmitted from the dispersion calculator 110, and when the dispersion degree value is less than the predetermined threshold value, the MIDI outputted through the median filter 100. A selection signal is generated to allow the frozen vector to be transmitted to the corresponding prediction through the switch 140. Accordingly, the differential encoding unit 150 performs differential encoding between the applied median vector and the current motion vector. In this case, the encoded motion vector output through the multiplexer 160 is composed only of data transmitted from the differential encoder 150.

However, as a result of comparing the predetermined threshold value and the dispersion degree value transmitted from the dispersion calculator 110 by the selection signal generator 120, when the dispersion degree value is greater than or equal to the predetermined threshold value, the optimal predictor value determination unit 130 is determined. The optimum predicted value is output to the differential encoder 150 through the switch 140. The optimal predicted value is a motion vector having a value closest to the currently applied motion vector among the predicted candidate candidate data.

The difference encoder 150 transmits the differential encoder 150 to the multiplexer 160 by performing differential encoding between the applied optimal predicted value and the current motion vector. In this case, the multiplexer 160 performs a multiplexing process so that an additional bit applied through the switch 180 is included in the motion vector transmitted from the differential encoder 150. Here, the additional bit is a header code generated from the header code generator 170, and 1 to 2 bits are generated in each direction. That is, when the magnitudes between the median vector output from the median filter 100 and the optimal prediction value output from the optimal predictor determiner 130 have the same value, '0', the optimal predicted value is greater than the median vector. If it is small, a bit corresponding to '10' is generated. If the optimal prediction value is larger than the median vector, a bit corresponding to '11' is generated. As described above, the generated bits are generated for each of the horizontal and vertical directions as described above. This is because the motion vector consists of the values of the horizontal and vertical components.

As described above, when the encoding of the currently generated motion vector is performed, when the values of two motion vectors among the predicted candidate candidate data have similar values, the motion vectors currently having similar values also have similar values. Is likely to be close to. When the value of the currently applied motion vector is close to the above-described similar values, the predicted value output from the optimum predictor determiner 130 and the median vector output from the median filter 100 may be the same. As such, when the predicted value output from the optimal predictor determiner 130 and the median vector output from the median filter 100 are the same, the median vector output from the median filter 100 is differentially coded. The selection signal must be generated from the selection signal generation unit 120 to be transmitted to.

However, when the difference between similar and dissimilar values is large, the difference between similar values (0 and 0.5) and dissimilar values (7) is severe, for example, the predicted candidate data applied is 0, 0.5, 7. In this case, according to the prior art dispersion calculation method, the dispersion degree is severely obtained. That is, the dispersion calculation of the prior art obtains the square root of the difference between the median value (MVPm) and the smallest value (MVPs), as shown in Equation 1, to be larger than the actual difference, the largest value (MVPb) and the median value (MVPm) Similarly, the square root of the difference is calculated to be larger than the actual difference to calculate the degree of variance between candidate data.

S (Pseudo Variance) = (MVPm-MVPs) ² + (MVPb-MVPm) ²

As a result, when one of the candidate data has a value that is not close at all, a selection signal in which the degree of dispersion between the candidate data is considered to be severe is generated.

In other words, when the predicted candidate data is applied as 0, 0.5, or 7, as in the above-described example, the variance result S is a large value such as 42.5 by the operation of (0.5-0) ² + (7-0.5) ² . This will be saved. Therefore, the predicted value output from the optimal predictor determiner 130 is transmitted to the differential encoder 150 by the selection signal generated by the selector signal generator 120, and the header code generated from the header code generator 170 is The encoded motion vector transmitted to the multiplexer 160 and output from the multiplexer 150 has a form in which a header code is added to the differentially encoded motion vector. If the added header code is 0, this means that the currently applied motion vector had a motion vector value close to the median vector.

When the header code is added in this way, an unnecessary 1-bit header code is added. When this phenomenon occurs for each direction component, a motion vector to which 2 bits of unnecessary bits are added is transmitted. This results in a decrease in the coding efficiency.

An object of the present invention is to provide a motion vector encoding method for preventing the coding efficiency of a motion vector from being degraded by variance calculation considering two cases in which the candidate candidate data are similar. There is this.

In order to achieve the above object, a motion vector encoding method according to the present invention includes a first motion vector predictor (MVPm) consisting of a median value of the plurality of candidate motion vectors and the plurality of motion vectors according to a dispersion degree of a plurality of candidate motion vectors. The first or second motion vector predictor selected from among the second motion vector predictors having a value closest to the currently applied motion vector among candidate motion vectors is modulated with the currently applied motion vector by differential pulse code modulation. In the method of encoding a motion vector, when the plurality of candidate motion vectors are applied, the dispersion of the plurality of candidate motion vectors is severe only when the plurality of candidate motion vectors do not all have similar values. A dispersion degree determination step of judging; If it is determined that the dispersion degree of the plurality of candidate motion vectors is severe in the dispersion degree determination step, the currently applied motion is performed in such a manner that the second motion vector predictor and the currently applied motion vector are differential pulse code-modulated. A first encoding step of performing encoding on the vector; If it is determined in the step of determining the dispersion degree that the dispersion degree for the plurality of candidate motion vectors is not severe, the first applied motion vector predictor and the currently applied motion vector are differentially pulse-coded. And a second encoding step of encoding the motion vector.

1 is a schematic block diagram of a motion vector encoding apparatus of the prior art;

2 is a flowchart illustrating a motion vector encoding method according to the present invention.

<Description of the symbols for the main parts of the drawings>

100: median filter 110: dispersion calculation unit

120: selection signal generator 130: optimal prediction value determiner

140: switch 150: differential encoding unit

160: header code generator 170: multiplexer

140, 180: switch

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

2 is a flowchart illustrating a motion vector encoding method according to the present invention. The operation of FIG. 2 will be described in detail with reference to FIG. 1.

First, in a preferred embodiment of the present invention, the predicted candidate data applied is set as motion vectors of three candidate blocks adjacent to the current search block. In addition, the applied motion vectors are composed of horizontal and vertical component values (x, y), respectively, but for convenience of description, the following description will be made based on one-directional components.

When the motion vectors of the three candidate blocks are applied as the predictive candidate data, the motion vectors of the three candidate blocks are transmitted to the variance calculator 110, the median filter 100, and the optimal predictive value determiner 130 as in the prior art.

Dispersion calculation unit 110 is operated as shown in FIG.

That is, when the median vector for the three predictive candidate data is applied from the three predictive candidate data and the median filter 100 in step 201, the process proceeds to step 202 to refer to the median vector in the three predictive candidate data. As a result, the smallest motion vector MVPs and the largest motion vector MVPb are respectively detected. For example, when three candidate candidate data currently applied are 0, 0.5, and 7, the median vector MVPm transmitted from the median filter 100 becomes 0.5, and thus MVPs extracted in step 202 becomes 0. MVPb becomes seven. This detection is performed by comparing the magnitude of three candidate data applied based on the median vector.

When MVPs and MVPb are detected as described above, the flow proceeds to step 203 to obtain an absolute value A of the difference between the median vector MVPm and the smallest motion vector MVPs as shown in Equation 2 below.

A = | MVPm-MVPs |

In operation 204, the absolute value B of the difference between the largest motion vector MVPb and the median vector MVPm is obtained as shown in Equation 3 below.

B = | MVPb-MVPm |

Proceeding to step 205, the obtained A and B are added to obtain a C value. That is, as in the above example, when MVPm is 0.5, MVPs is 0, and MVPb is 7, the dispersion calculation result C obtained in accordance with the present invention is 7 by | 0.5-0 | + | 7-0.5 | do. This is a considerably smaller value than the conventional 42.5. The dispersion calculation result value C thus obtained is transmitted to the selection signal generator 120.

Since the selection signal generation unit 120 obtains a significantly smaller value than the conventional dispersion calculation result obtained by the modified dispersion calculation as in the present invention, a predetermined threshold value used to determine whether the dispersion degree is severe or not is determined. Set smaller than before. That is, the predetermined threshold is set so that the variance can be determined only when all three predicted values that are applied are not similar.

Accordingly, in operation 206, the selection signal generator 120 compares the dispersion calculation result value C transmitted from the dispersion calculation unit 110 with a predetermined threshold, and as a result of the comparison, the dispersion calculation result value C is predetermined. If it is larger than the threshold, it is determined that the degree of dispersion is severe and the flow proceeds to step 207.

In operation 207, the motion vector encoding apparatus shown in FIG. 1 performs a motion vector encoding process for a case where the degree of dispersion is severe. That is, the selection signal generation unit 120 controls the switch 140 so that the prediction value output from the optimum prediction value determination unit 130 is transmitted to the differential encoding unit 150 and is output from the header code generation unit 170. A selection signal for controlling the switch 180 is generated so that a predetermined header code is transmitted to the multiplexer 160. Accordingly, the difference encoder 150 performs differential encoding between the optimal predicted value applied and the currently applied motion vector, and transmits the difference to the multiplexer 160. The multiplexer 160 is connected to the data transmitted from the differential encoder 150. The header code of a predetermined bit transmitted through the switch 180 is multiplexed and output as encoded motion vector data. It returns when this encoding process is completed.

On the other hand, when the comparison result of step 206, the dispersion calculation result value (C) is not greater than the predetermined threshold value, since the degree of dispersion is not severe, the process proceeds to step 208 to perform the encoding process for the currently applied motion vector. . That is, the selection signal generator 120 controls the switch 140 so that the median vector output from the median filter 100 is transmitted to the differential encoder 150 and is generated from the header code generator 170. The predetermined header code generates a selection signal for controlling the switch 180 such that the predetermined header code is not transmitted to the multiplexer 160. Accordingly, the differential encoding unit 150 performs differential encoding between the applied median vector and the currently applied motion vector to the multiplexer 160, and the multiplexer 160 only transmits the data transmitted from the differential encoding unit 150. Output as encoded motion vector data. It returns when this encoding process is completed.

If the dispersion result value is 7, as in the above-described example, the process proceeds from step 206 to step 208 so that the encoding process of the motion vector is performed according to the determination that the dispersion degree is not severe. The above process is performed for each direction component of the motion vector.

As described above, the present invention encodes a motion vector by performing variance calculation to determine that the degree of variance is severe only when the motion vectors of the selected candidate blocks do not all have similar values. If the vectors have two or more similar values, no additional bits are added, thereby improving the coding efficiency of the motion vector.

Claims (3)

A first motion vector predictor (MVPm) consisting of intermediate values of the plurality of candidate motion vectors and a value having a value closest to a currently applied motion vector among the plurality of candidate motion vectors according to a dispersion degree of the plurality of candidate motion vectors. A method of performing encoding on the currently applied motion vector by differentially modulating a first or second motion vector predictor selected from two motion vector predictors with the currently applied motion vector, A dispersion degree determining step of determining that the dispersion degree of the plurality of candidate motion vectors is severe only when the plurality of candidate motion vectors do not all have similar values when the plurality of candidate motion vectors are applied; If it is determined that the dispersion degree of the plurality of candidate motion vectors is severe in the dispersion degree determination step, the currently applied motion is performed in such a manner that the second motion vector predictor and the currently applied motion vector are differential pulse code-modulated. A first encoding step of performing encoding on the vector; If it is determined in the step of determining the dispersion degree that the dispersion degree for the plurality of candidate motion vectors is not severe, the first applied motion vector predictor and the currently applied motion vector are differentially pulse-coded. And a second encoding step of encoding the motion vector. The method of claim 1, wherein the determining of dispersion degree comprises: Candidate motion vector detection for detecting candidate motion vectors (MVPs) having the smallest value and candidate motion vectors (MVPb) having the largest value among the plurality of candidate motion vectors based on the first motion vector predictor (MVPm). step; The absolute value A of the difference between the first motion vector predictor (MVPm) and the candidate vector (MVPs) having the smallest value detected in the candidate motion vector detection step, and the candidate vector (MVPb) having the largest value; A variance calculation result obtaining step of obtaining a sum of absolute values (B) of differences between the first motion vector predictors as a variance calculation result value (C) for the plurality of candidate motion vectors; And a dispersion degree determining step of comparing the obtained dispersion calculation result value (C) with a predetermined threshold value to determine whether the dispersion degree is severe. 3. The motion vector encoding method of claim 2, wherein the threshold is set to determine that the variance of the plurality of candidate motion vectors is severe when the plurality of candidate motion vectors do not all have similar values.

Images