JP3033291B2 - Inter-pixel correlation adaptive motion detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an inter-pixel correlation adaptive motion detection system at the time of high-efficiency coding of a moving image signal such as a television signal.

[0002] In high-efficiency coding of a moving image signal such as a television signal, a motion compensation method is often adopted to obtain a higher compression rate. In the block matching method among the motion compensation methods, an encoded screen is divided into blocks (encoded blocks) including a plurality of pixels, and an optimal block is detected from a reference screen. As an example of a criterion for judging that it is optimal, the absolute value of the difference between each pixel in the block to be coded and the corresponding pixel of the reference block in the reference screen is determined, and the sum thereof is the smallest (see FIG. 6). .

Here, even if the screen resolution is the same,
It is not efficient to always make the search range the same, considering the fine pattern and the pattern when zoomed up.
In other words, even when the subject moves at the same speed, it moves over many pixels when zooming up. Therefore, if the search range is narrow, even a slight movement has no effect.

[0004] Therefore, there is a demand for a motion detection method in which the motion detection range can be dynamically changed according to the correlation between pixels.

[0005]

2. Description of the Related Art FIG. 6 is a diagram for explaining an example of motion-compensated inter-frame coding. FIG. 7 is a block diagram showing a configuration of a conventional block matching search circuit.

In FIG. 7, an analog image signal is digitized by an analog / digital conversion circuit (hereinafter referred to as A / D) 1, and the digitized data is input to a memory 2 and written to a predetermined address.

Next, the data of the screen (encoded screen) of the current frame stored at the above address of the memory 2 is read out in block units by a signal indicating the address of the output of the address generation unit 7 to calculate the absolute difference value. Input to section 4. Similarly, in response to a signal indicating the address of the output of the address generator 7, for example, the data of the screen (reference screen) of the immediately preceding frame stored at the above address in the memory 3 is read out in block units and input to the absolute difference calculator 4. I do.

Then, the difference absolute value calculation unit 4 calculates a difference (Xi-Yi) between the pixel data (Xi) of a certain block on the encoded screen and the pixel data (Yi) of a certain block on the reference screen. The absolute value of the difference (Xi-Yi) is added to all the pixels in the block. The output of the added value is applied to the error comparison unit 5.

The block position of the reference screen is moved vertically and horizontally by a fixed number of pixels, and the absolute value of the difference (Xi-Yi) is similarly added to all the pixels in the block, and the output of the added value is obtained. It is added to the error comparison unit 5. Then, the value is compared with the value at the block position before the movement, and the block position with the smallest error is sequentially obtained.

The value of the block position at which the error obtained by the error comparing unit 5 is minimized is input to the motion vector generating unit 6,
A vector (motion vector) between the block position at which this error is minimized and the block position of the screen to be coded (current frame) is obtained (see FIG. 6). Thus, a motion vector is obtained by the motion detection method using the block matching method.

[0011]

However, in the above-described motion detection method, even when the resolution of the screen is the same, considering the fine pattern and the pattern when zoomed up, the search range is always the same. Be less efficient. In other words, even when the subject moves at the same speed, it moves over many pixels when zooming up.

Therefore, if the search range is narrow, even small movements have no effect. Further, when the search range is widened, there is a problem that the hardware scale increases.

It is therefore an object of the present invention to provide an inter-pixel correlation adaptive motion detection system in which a motion detection range can be dynamically changed according to the correlation between pixels.

[0014]

The above problem is solved by the circuit configuration shown in FIG. That is, in FIG.
(Claim 1) In a circuit having a motion detecting section 800 for dividing a coded picture of a moving picture signal into blocks composed of a plurality of pixels, and comparing the reference picture with a reference picture to detect a block satisfying a predetermined condition, And a correlation detection unit that detects a correlation between pixels from an input moving image signal and outputs a control signal.

A control signal output from the correlation detecting section 900 allows the motion detecting section 800 to change the detection range of a block satisfying a predetermined condition from the reference screen according to the detection result of the inter-pixel correlation. .

(Claim 2) When the degree of correlation of the inter-pixel correlation according to claim 1 is larger than a predetermined value, the number of search blocks per block of the coded picture is made constant, and the search is performed on the reference picture. The detection area of the block to be detected is made coarse to increase the detection area.

(Claim 3) When the degree of correlation between the pixels according to claim 1 is larger than a predetermined value, the search range is made constant and the detection interval of the block to be searched on the reference screen is increased. It is configured to reduce the number of search blocks.

[0018]

In FIG. 1, a motion detecting section 800 divides a coded picture of a moving picture signal stored in a storage section into blocks composed of a plurality of pixels, and generates a reference picture (for example, the immediately preceding picture). And a block satisfying a predetermined condition is detected. The block that satisfies the predetermined condition is, for example, obtained by calculating the absolute value of the difference between each pixel in the block to be coded and the corresponding pixel of the reference block in the immediately preceding screen, and determining the sum of the absolute values.

On the other hand, a newly provided correlation detecting section 900 detects, for example, a correlation between adjacent pixels from an input moving image signal, outputs a control signal, and adds the control signal to the motion detecting section 800. When reading the reference block from the immediately preceding screen stored in the storage unit, according to the above-described detection result of the inter-pixel correlation,
When the degree of correlation between pixels is larger than a predetermined value, the number of search blocks per block of the encoded screen is kept constant, and, for example, a reference block consisting of a fixed number of pixels every other pixel is read. (Corresponding to claim 2).

Alternatively, when the degree of correlation between pixels is larger than a predetermined value, the search range is fixed, the detection interval of blocks to be searched in the immediately preceding screen is made coarse, and the number of search blocks is reduced. For example, a reference block including a fixed number of plural pixels is read every other pixel. (Corresponding to claim 3).

Then, as described above, for example, the absolute value of the difference between each pixel in the block to be coded and the corresponding pixel in the reference block on the immediately preceding screen is determined, and the block having the minimum sum is detected.

As a result, according to the detection result of the inter-pixel correlation output from the correlation detecting unit 900, for example, when the correlation between adjacent pixels is large, a wider range can be searched to reduce the prediction error. Alternatively, the amount of calculation can be reduced by coarsening the search.

[0023]

FIG. 2 is a block diagram showing the configuration of a circuit according to an embodiment of the present invention. FIG. 3 is a diagram for explaining the motion detection range expanding method according to the first embodiment.

FIG. 4 is a diagram for explaining a method of reducing the number of search vectors according to the second embodiment. FIG. 5 is a block diagram illustrating a configuration of the correlation detection unit according to the embodiment. The same reference numerals indicate the same objects throughout the drawings.

In FIG. 2, the analog image signal is represented by A /
The data is digitized by D 1, and the digitized data is input to the memory 2, written to a predetermined address, and input to the correlation detector 9.

FIG. 5 shows the structure of the correlation detecting section 9.
In (A), for each pixel, input data digitized into, for example, 8 bits is added to a subtractor 16 and added to a delay circuit 15 made of, for example, a flip-flop circuit (not shown). Add to the subtractor 16. By calculating the difference between the two inputs by the subtractor 16, a difference in brightness, color, and the like from an adjacent pixel is determined. This difference output is applied to an absolute value circuit 17, the absolute value of which is calculated, and the absolute value output (d) of the difference is applied to a comparator 18.

The comparator 18 compares the absolute value of the difference (d) with a threshold value (Th) set in advance by the threshold value generating section 19, and the absolute value of the difference (d) is larger than the threshold value (Th). When it is large, it outputs 0, and when it is small, it outputs 1 and adds it to the accumulator 20. The accumulator 20 performs cumulative addition for each block or frame for which the correlation is calculated, and outputs the result to the control signal output unit 21.

The control signal output section 21 outputs a detection range control signal value corresponding to the input cumulative value based on, for example, a characteristic diagram showing a relationship between the cumulative value and the detection range control signal shown in FIG. I do. In the case of FIG. 5B, when the accumulated value is large, that is, when the degree of correlation with an adjacent pixel is high, a control signal having a large value is output. The circuit having the characteristics shown in FIG. 5B can be made by an IC or the like. The detection range control signal of the output of the control signal output unit 21 is applied to the data selection circuit 7 'shown in FIG.

FIG. 3 is a diagram for explaining the method of expanding the motion detection range according to the first embodiment. FIG. 3A shows a search vector for obtaining an optimum block from a reference screen by the above-described detection range control signal. It is an explanatory view about how to obtain.
In FIG. 9A, (x, y) coordinates (0, 0) indicating a predetermined position of each block set in a motion vector table (memory) 10 provided in a data selection circuit 7 ',
Blocks having (0, 1), (0, -1), (1, 0),... Are sequentially read out and input to the multiplier 11 to perform multiplication with the above-described detection range control signal.

The circles and circles shown in (a) and (b) of FIG.
The mark indicates a fixed position of each block including a fixed number of pixels (for example, the position of the upper left pixel in each block). (A) shows a fixed pixel position of each block before multiplication is performed by the multiplier 11, and (b) shows a certain pixel position.
Shows a fixed pixel position in each block after multiplication is performed. FIG. 3B shows that the detection range control signal is 2 (magnification).
Is shown.

In FIG. 3B, among the nine circles (a), the circle at the center position is, for example, coordinates (0, 0), the horizontal direction is x, and the vertical direction is y, and (0) , 0) have coordinates of (1, 0), (−1,
0). The coordinates of ● immediately above and below (0, 0) are (0, 1) and (0, -1), respectively. When these coordinates are multiplied by the multiple indicated by the detection range control signal (in this case, for example, twice), the coordinates become as indicated by the black circles shown in (b) of FIG.

Then, the data of the reference screen of each block indicated by ● in FIG. 2B is read out from the memory 3 shown in FIG. The data of the coded block of the coded screen (the screen of the current frame) read from the memory 2 is also input to the difference absolute value calculation unit 4.

In FIG. 3B, (b) indicates a fixed pixel position in a block that is not read. In other words, (b) of FIG.
The figure shows a case where each block is read out every other pixel.

The difference absolute value calculator 4 calculates the difference between the pixel data (Xi) of the block of the coded picture and the pixel data (Yi) of the block of the reference picture in the same manner as described in the background art. (Xi-Yi) is obtained, and the difference (Xi-Yi) is obtained.
The absolute value of Yi) is added to all the pixels in the block.
The output of the added value is applied to the error comparison unit 5.

As described above, the block position of the reference screen is such that the data of each block in which the fixed position of the block is indicated by ● shown in (b) of FIG. The difference (Xi
−Yi), and the absolute value of the difference (Xi−Yi) is added to all the pixels in the block. The output of the added value is applied to the error comparison unit 5.

Then, the error comparison unit 5 compares the value at the block position before the movement with the value at the block position, and sequentially finds the block position at which the error is minimized. The value of the block position at which the error obtained by the error comparison unit 5 is minimum is input to the motion vector generation unit 6, and the block position of the reference screen and the block of the coded screen (current frame) at which this error is minimum A vector (motion vector) between the positions is obtained (see FIG. 6).

In FIG. 5, the amount of delay by the delay circuit 15 is not limited to one pixel, but is delayed by two pixels (every two pixels) when the correlation between pixels is large according to the predicted state of the input screen. Or 3 pixel delay (every 3 pixels)
Can be set in advance.

In the figure, the output of the threshold value generating section 19 can be appropriately changed and set according to the state of the input screen. Further, the gradient (straight line or curve) of the characteristic diagram showing the relationship between the accumulated value and the detection range control signal shown in FIG. 5B can also be changed as appropriate according to the state of the input screen.

As a result, the motion detection range can be dynamically changed according to the correlation between pixels. Next, a second embodiment shown in FIG. 4 will be described.

When the value of the detection range control signal of the output of the correlation detector 9 shown in FIG. 5 is normally 1, the pointer value (0, 1, 2, 2) of the output of the read pointer generator 14 shown in FIG. ,
..) is directly added to the motion vector table 12 via the multiplier 13, and the data of the reference screen of the block having the pixel of the coordinate corresponding to the pointer value at a fixed position (for example, the position of the upper left pixel) of each block is obtained. , The memory 3 shown in FIG.
And input to the difference absolute value calculation unit 4.

For example, in the case of (a) in FIG. 4B, the coordinates of the position of the pixel indicated by ● in order from the left end of the uppermost row to the right are (−2,
2), (-1, 2), (0, 2), (1, 2), (2,
2), and the corresponding pointers are 0, 1, 2, 3, and 4. Therefore, as the pixel corresponding to the pointer 0 of the output of the read pointer generation unit 14, the upper left pixel (−2, 2)
Is read from the memory 3 shown in FIG. 2 and input to the absolute difference calculating unit 4. Then, the difference (Xi−Yi) between the pixel data (Xi) of the block of the coded screen and the pixel data (Yi) of the block of the reference screen is obtained by the same method as described in the background art, and this difference ( The absolute value of Xi−Yi) is added to all the pixels in the block, and the output of the added value is applied to the error comparison unit 5.

Next, since the second pixel (-1, 2) from the left in the uppermost row of (A) in FIG. 4B corresponds to the pointer 1, a block having this second pixel at a fixed position Is read out from the memory 3 shown in FIG.
The same operation as described above is performed. The pointers 2, 3,
.. Read the data of the reference screen of the corresponding block from the memory 3 and perform the same operation as described above.

Then, the error comparison unit 5 compares the value at the block position before the movement with the value in the same manner as described in the prior art, and sequentially obtains the block position with the minimum error. The value of the block position at which the error obtained by the error comparison unit 5 is minimum is input to the motion vector generation unit 6, and the block position of the reference screen and the block of the coded screen (current frame) at which this error is minimum A vector (motion vector) between the positions is obtained (see FIG. 6).

Now, when the correlation between the pixels of the input signal is large and the value of the detection range control signal output from the correlation detection unit 9 shown in FIG. 5 is larger than 1, the read pointer generation unit in the multiplier 13 shown in FIG. The pointer value of the output of FIG. 14 is multiplied by the value of the detection range control signal, and the data of the reference screen of the block having the pixel at the coordinates specified by the detection range control signal at a fixed position (for example, the upper left corner) is shown. 2 is read from the memory 3 shown in FIG.

For example, when the value of the detection range control signal is 2,
The pointer value of the output of the read pointer generator 14 (0,
, 1, 2, 3,...) Are added to the motion vector table 12, so that the pointer values of the motion vector table 12 correspond to pixels at coordinates corresponding to 0, 2, 4, 6,. Are sequentially read from the memory 3 shown in FIG. 2 and input to the absolute difference calculating unit 4.

That is, a block in which the pixel indicated by ● in (b) of FIG. 4B is a fixed position (for example, the upper left corner) is read from the memory 3. A mark indicates a block that is not read. That is, (b) of FIG. 4B shows a case where each block is read every other pixel with the same area of the block to be read. Thereafter, the same operation as described above is performed.

As a result, the motion detection range can be dynamically changed according to the correlation between pixels.

[0048]

As described above, according to the present invention, the motion detection range can be dynamically changed according to the correlation between pixels. As a result, it is possible to reduce the prediction error in the motion compensation method, or to reduce the amount of calculation.

[Brief description of the drawings]

FIG. 1 is a principle diagram of the present invention,

FIG. 2 is a block diagram showing the configuration of a circuit according to an embodiment of the present invention;

FIG. 3 is a diagram for explaining a motion detection range enlarging method according to the first embodiment;

FIG. 4 is a diagram for explaining a search vector number reduction method according to a second embodiment;

FIG. 5 is a block diagram illustrating a configuration of a correlation detection unit according to the embodiment;

FIG. 6 is a diagram for explaining an example of motion-compensated inter-frame coding;

FIG. 7 is a block diagram showing a configuration of a conventional block matching search circuit.

[Explanation of symbols]

 Reference numeral 800 denotes a motion detection unit, and 900 denotes a correlation detection unit.

Continuation of front page (56) References JP-A-55-162686 (JP, A) JP-A-63-181585 (JP, A) JP-A-2-33286 (JP, A) JP-A-4-150284 (JP) , A) (58) Field surveyed (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68

Claims (3)

(57) [Claims] 1. A circuit having a motion detecting section (800) for dividing a coded picture of a moving picture signal into blocks composed of a plurality of pixels, and comparing the reference picture with a reference picture to detect a block satisfying a predetermined condition. A correlation detection unit (900) for detecting a correlation between pixels from an input moving image signal and outputting a control signal; and a control signal output from the correlation detection unit (900) is used in the motion detection unit (800). An inter-pixel correlation adaptive motion detection method, wherein a detection range of a block satisfying a predetermined condition from the reference screen can be changed according to an inter-pixel correlation detection result. 2. When the degree of correlation of the inter-pixel correlation according to claim 1 is larger than a predetermined value, the number of search blocks per block of the coded picture is made constant, and the search is performed on the reference picture. A pixel-correlation-adaptive adaptive motion detection method characterized by broadening the detection area by coarsening the block detection interval. 3. The search block according to claim 1, wherein when the degree of correlation of the inter-pixel correlation is larger than a predetermined value, the search range is fixed and the detection interval of the block to be searched on the reference screen is coarse. An inter-pixel correlation adaptive motion detection method characterized in that the number is reduced.