JP3149076B2 - Motion vector detection device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector detecting device for detecting a predicted block having a minimum sum of differences between an input block in a frame and a block to be predicted in another frame. .

[0002]

2. Description of the Related Art Conventionally, MPEG and H.264 are used. The motion vector detection used in the technology for compressing / expanding moving images such as H.261 uses a block matching method as shown in FIG. The block matching method shown in FIG.
A difference between a block of a current frame (input block) and a block (predicted block) in a predetermined area of the previous frame is calculated, and a predicted block having the smallest sum of the differences is detected. The distance to the input block of the frame was detected as a motion vector.

[0003]

In the above-described conventional motion vector detection described with reference to FIG. 8, even when the difference between some pixels in a block is very large when the difference is obtained, the motion vector is not considered as a whole. In this case, the difference becomes small, and there is a problem that efficient conversion cannot be performed when the difference between the input block and the prediction block is subjected to DCT and compression.

[0004] The present invention solves these problems,
A difference between corresponding pixels between the input block and the prediction block in the frame is determined, a prediction block having the smallest sum of the differences between the adjacent pixels is detected, and DCT transform is performed based on the difference between the input block and the prediction block. Go and compress, AC
The purpose is to compress moving images with high efficiency by reducing the components.

[0005]

Means for solving the problem will be described with reference to FIG. In FIG. 1, an encoder 2 fetches an input image and stores the fetched image in a memory 3, and calculates a difference between an input block and a prediction block based on a motion vector detected by a motion vector detector 4.
The moving image is compressed by CT conversion.

The motion vector detector 4 detects a predicted block in which the sum of the differences between the input block and the block to be predicted is the smallest. The decoder 5
The image is generated by decoding the signal generated by the encoder 2.

[0007]

According to the present invention, as shown in FIG. 1, after a motion vector detector 4 calculates a pixel-correspondence difference between an input block of a current frame inputted and a block of which motion of a previous frame is to be predicted. , The difference of the difference between the adjacent pixels is obtained, the block having the smallest sum is detected as the prediction block, and the encoder 2 performs DCT conversion on the difference between the input block and the prediction block and compresses the difference.

At this time, a block having the smallest sum of the differences between the adjacent horizontal pixels is detected as a predicted block. Further, a block in which the sum of the differences between the adjacent pixels in the vertical direction is the smallest is detected as a predicted block.

Further, a block in which the sum of the differences between the adjacent pixels in the oblique direction is the smallest is detected as a predicted block. In addition, a block in which the sum of the differences between the pixels in two or more of the horizontal direction, the vertical direction, and the diagonal direction in contact with each other is detected as a predicted block is the smallest.

In addition, an image of a raw frame before compression is used as an input block and a block whose motion is to be predicted. Further, an image of a raw frame before compression is used as an input block, and an image of a frame obtained by compressing and reproducing a block whose motion is to be predicted is used.

Therefore, after calculating the difference between the corresponding pixels between the input block and the prediction block, the difference between the adjacent pixels is calculated, and the prediction block having the minimum sum is detected. Is compressed by performing DCT transform based on the difference
It is possible to perform moving image compression with high efficiency by performing DCT transformation by reducing the components.

[0012]

Next, the structure and operation of an embodiment of the present invention will be sequentially described in detail with reference to FIGS.

FIG. 1 shows a system configuration diagram of the present invention.
In FIG. 1, an input device 1 is for inputting a frame (image), for example, a video camera as shown in FIG.
Examples include a video disk and a laser disk.

The encoder 2 fetches the input image and stores it in the memory 3, and calculates the difference between the input block of the current frame and the prediction block of the previous frame based on the motion vector detected by the motion vector detector 4. D
It performs CT conversion and compresses moving images.

The memory 3 stores and stores frames (images) and the like input from the input device 1. The motion vector detector 4 calculates a pixel-corresponding difference between the input block of the current frame and the block to be predicted in the previous frame, and detects a predicted block in which the sum of the differences between the adjacent pixels is the smallest. is there.

The decoder 5 decodes the signal compressed by the moving image by the encoder 2 to generate an original image (frame). The display 6 displays an image (frame) decoded by the decoder 5.

Next, the operation of the configuration of FIG. 1 will be described in detail according to the order shown in the flowchart of FIG. In FIG. 2, S1 captures image data. this is,
The encoder 2 shown in FIG. 1 takes in the image data inputted from the input device 1, stores it in the memory 3, and stores it.

In step S2, a block i (input block) of the current frame is extracted. In this method, the current frame is divided into blocks (for example, one pixel is represented by 8 bits, and 16 pixels × 16 pixels are divided into one block), and an arbitrary block i is extracted from the current frame.

In step S3, a block j of the previous frame is extracted from the predetermined area. This means that the previous frame is similarly divided into blocks (for example, one pixel is represented by 8 bits, 16 × 16 pixels are divided into one block), and a predetermined area in the previous frame (for example, corresponding to block i of the current frame) An arbitrary block j is extracted from within a predetermined area of 48 pixels × 48 pixels centering on the block position of the previous frame.

In step S4, a pixel-corresponding difference between block i (input block) and block j (prediction block) is calculated. In step S5, the differences between the adjacent pixels in the horizontal, vertical, and oblique directions are calculated, the sum of the differences is calculated, and the sum is calculated. This is schematically illustrated in FIG. 4 described below.
According to the equation in FIG. 5, the difference between the block i (input block) and the block j (prediction block) is obtained, and the sum of the differences between the adjacent pixels in the horizontal direction, the vertical direction, and the diagonal direction is obtained. Calculate the sum of the sums.

In step S6, the pixel is shifted in the horizontal or vertical direction by one pixel. This corresponds to the block j (16
A block j of pixels × 16 pixels is sequentially shifted by one pixel in a horizontal direction or a vertical direction within a predetermined area (for example, a predetermined area of 48 pixels × 48 pixels).

A step S7 decides whether or not the processing in the predetermined area is completed. In the case of YES, the block j (16 pixels × 1) is set within a predetermined area (for example, a predetermined area of 48 pixels × 48 pixels).
(The block of 6 pixels) is moved one pixel at a time in the horizontal direction or the vertical direction, and the processing for obtaining the sum of S5 at that time has been completed. On the other hand, in the case of NO,
S3 and subsequent steps are repeated for the next block j.

In S8, since the total sum has been calculated for all the blocks j in the predetermined area by YES in S7, the moving amount (Δ
x, Δy) are calculated.

In step S9, a difference in pixel units between the current block (block i) and the calculated block (block j having the smallest sum of differences) is calculated. This calculates the difference between the block i (input block) in the current frame obtained in S8 and the block j (prediction block) having the smallest sum in the previous frame.

In step S10, DCT is performed. this is,
Based on the difference obtained in S9, DCT transform described later is performed.
In step S11, quantization is performed.

In step S12, variable length coding is performed. As described above, the block i in the current frame (for example, 16 pixels × 16
Pixel, input block)) and a block i in a corresponding predetermined area (for example, 48 pixels × 48 pixels) in the previous frame is calculated, and the difference between the pixels in horizontal, vertical, and oblique directions is determined. , And the sum of these sums is calculated to obtain the smallest block j (prediction block)
And the block j (prediction block) having the minimum sum of the block i of the current frame (input block) and the previous frame
And performs DCT transformation, quantization, and variable encoding based on the difference between them, and compresses the moving image. Thus, when detecting a block j (prediction block) in a predetermined area in the previous frame with respect to a block i (input block) in the current frame, the sum of the sum of the differences in the horizontal, vertical, and oblique directions is calculated. , A block j (prediction block) having the least undulations (smallest AC component) can be detected, followed by DCT transform, quantization, and variable length. Encoding can be performed more efficiently. This will be described in detail below.

FIG. 3 shows an example of a difference value calculated according to the present invention. FIG. 3A schematically shows an input block (block i of the current frame in FIG. 2). Here, the input block is 4 pixels × 4 pixels, and each pixel is 8 bits (256 bits).
(Gradation). X0, X1, X2, X3, Y0, Y1,
Y2 and Y3 represent pixels (4 × 4), and the vertical direction represents 8-bit gradation.

FIG. 3B schematically shows a prediction block (block j of the previous frame in FIG. 2). Here, the input block is 4 pixels × 4 pixels, and each pixel is 8 bits (25 bits).
6 gradations). X0, X1, X2, X3, Y0, Y
1, Y2 and Y3 represent pixels (4 × 4), and the vertical direction represents 8-bit gradation.

FIG. 3C shows the difference between the differences. That is, the difference between the input block of FIG. 3A and the prediction block of FIG. 3B is obtained, the sum of the differences between the pixels adjacent in the horizontal direction, the vertical direction, and the oblique direction is obtained, and the sum is obtained. It is what was asked. As can be seen from the difference between the differences, since the sum of the differences between the adjacent pixels is calculated, the AC component is subtracted, and a prediction block without gentle undulations can be detected. Conversion can be performed.

FIG. 4 is a diagram for explaining the operation of the present invention. This is a specific explanatory diagram when a difference between the above-described differences is obtained. FIG. 4A shows a predicted frame. This prediction frame is the previous frame and is divided into a plurality of blocks (prediction blocks) as shown in the figure. Each block (prediction block) has a pixel value Y _{00 of} 8 × 8 pixels here. _{~Y 07, Y 10 ~Y 17,} Y 20 ~Y 27, Y
_{30 ~Y 37, Y 40 ~Y 47} , Y 50 ~Y 57, Y 60 ~Y 67, Y 70
And with a ~Y _77.

FIG. 4B shows the current frame. This current frame is similarly divided into a plurality of blocks (input blocks) as shown in the figure, and each block (input block)
Here, for 8 × 8 pixels, each pixel has a pixel value of X _{00 to} X ₀₇ , X _{10 to} X ₁₇ , X _{20 to} X ₂₇ , X _{30 to} X ₃₇ , X
And with _{40 ~X 47, X 50 ~X 57} , X 60 ~X 67, X 70 ~X 77.

FIG. 4C shows how the difference between the horizontal direction and the vertical direction is obtained. (1) The pixel value X of each pixel of the block (input block) of the current frame shown in FIG. 4B and the pixel value Y of each pixel of the block (prediction block) of the predicted frame shown in FIG. The differences Z _ij (i = 0 to 7, j = 0 to 7) shown in the figure are obtained. For example, determined as _{· Z 00 = X 00 -Y 00} .

(2) The difference Δa _ij (i = 0-7, j =
0-7) respectively. For example, Δa ₀₀ = Z ₀₁ −Z ₀₀ is determined. Next, the difference Δa _ij (i
= 0 to 7 and j = 0 to 7) for each horizontal direction Σaj (j =
0 to 7) are obtained as shown in the figure.

(3) The difference Δb _ij (i = 0 to 7, j = 0) between the pixels adjacent in the vertical direction with respect to the difference Z obtained in (1)
0-7) respectively. For example: Δb ₀₀ = Z ₁₀ −Z ₀₀ Next, the difference Δb _ij (i
= 0 to 7 and j = 0 to 7) for each vertical direction Σbi (i =
0 to 7) are obtained as shown in the figure.

(4) The sum of the sums obtained in (2) and (3) is obtained. The summation is determined by shifting one pixel up, down, left and right with respect to the block shown in FIG. 4A by one pixel, and the sum at each time is obtained, and the block having the smallest sum is detected as a predicted block.

As described above, it is possible to obtain the sum of the differences between the horizontal direction and the vertical direction and detect the block having the minimum sum as the predicted block. This allows
By calculating the total sum of the differences in the horizontal direction and obtaining the predicted block with the minimum total, it is possible to detect the predicted block with a small undulation by subtracting the AC component in the horizontal direction and increase the efficiency of DCT transformation and the like. In addition, the sum of the differences in the vertical direction is calculated, and the predicted block having the smallest sum is obtained. Thus, the AC component is subtracted in the vertical direction to detect a predicted block with little undulation, thereby improving the efficiency of DCT transformation and the like. It is possible to increase.

FIG. 4D shows the manner in which the difference in the oblique direction from the upper left to the lower right is obtained. This is obtained by calculating the total sum of the differences in the oblique direction from the upper left to the lower right as shown in FIG. By calculating the sum of the differences of the differences in the diagonal direction from the upper left to the lower right to find the predicted block with the smallest sum, the AC component is subtracted in the diagonal direction from the upper left to the lower right to detect a predicted block with little undulation As a result, the efficiency of DCT conversion and the like can be increased.

FIG. 4E shows the manner in which the difference of the difference in the oblique direction from the lower left to the upper right is obtained. This is obtained by calculating the total sum of the differences in the oblique direction from the lower left to the upper right as shown in the figure, similarly to FIG. 4C described above. By calculating the sum of the differences of the differences in the diagonal direction from the lower left to the upper right to obtain the predicted block having the smallest sum, the AC component is subtracted in the diagonal direction from the lower left to the upper right to detect the predicted block with less undulation. It is possible to increase the efficiency of DCT conversion and the like.

FIG. 5 shows a formula for calculating the sum of the present invention. FIG.
, Smin represents the minimum value of the sum. The right side of the first row is a calculation formula for calculating the sum in the oblique direction from the upper left to the lower right. This is a calculation formula for calculating the sum of (d) in FIG. 4 described above.

The second line is a calculation formula for calculating the sum in a diagonal direction from the lower left to the upper right. This is a calculation formula for calculating the sum of (e) in FIG. 4 described above. The third line is a calculation formula for calculating the sum in the horizontal direction. This is a calculation formula for calculating the horizontal sum in (c) of FIG. 4 described above.

The fourth line is a calculation formula for calculating the sum in the vertical direction. This is a calculation formula for calculating the total sum in the vertical direction in FIG. 4C described above. FIG. 6 shows the second embodiment of the present invention.
FIG. 3 is an explanatory diagram of a dimensional DCT transform.

FIG. 6A shows an image to be compressed by two-dimensional DCT. FIG. 6B shows the state of FIG.
(A), when expressed as a sum of 4 × 4 illustrated images, each coefficient S _ij (i = 0 to 3, j = 0 to
The state of obtaining 3) is shown. These coefficients S _ij (i = 0 to
The image of FIG. 6A is represented by the values of (3, j = 0-3).

FIG. 7 is a diagram for explaining image data compression according to the present invention. FIG. 7A shows image data. FIG. 7B shows an example of DCT coefficient data of the present invention. In the present invention, since the prediction block in the previous frame in which the sum of the differences is the smallest is detected with respect to the input block of the current frame as described above, the AC component is reduced to reduce the prediction block having the least undulation. Since the block is detected, as shown in the figure, the area where the DCT coefficient is small increases, and the part where the DCT coefficient is large concentrates on a part and the number decreases, so that the DCT conversion can be performed efficiently. While you can
In particular, in the case of variable length coding, it is possible to increase the compression rate by adding a short code to a portion having a large DCT coefficient.

[0044]

As described above, according to the present invention,
After calculating a difference between corresponding pixels between the input block and the prediction block of the frame, a difference between the differences between adjacent pixels is calculated, a prediction block having the smallest sum is detected, and the difference between the input block and the prediction block is also calculated. Since the DCT transform is performed to compress the image, the AC component is reduced, and the DCT transform is performed to compress the moving image with high efficiency.

[Brief description of the drawings]

FIG. 1 is a system configuration diagram of the present invention.

FIG. 2 is a flowchart illustrating the operation of the present invention.

FIG. 3 is an example of a difference value calculated according to the present invention.

FIG. 4 is a diagram illustrating the operation of the present invention.

FIG. 5 is an example of a sum calculation formula according to the present invention.

FIG. 6 is an explanatory diagram of a two-dimensional DCT transform according to the present invention.

FIG. 7 is an explanatory diagram of image data compression according to the present invention.

FIG. 8 is an explanatory diagram of a conventional technique.

[Explanation of symbols]

 1: input device 2: encoder 3: memory 4: motion vector detector 5: decoder 6: display

Claims (7)

(57) [Claims] An apparatus for compressing based on a difference between an input block in a frame and a prediction block in another frame, wherein the compression of the input block in the frame corresponds to the block in the frame whose motion is to be predicted. Pixel difference (Zi,
j), and finds the difference (Zi, j + 1, Zi + 1,
j, Zi + 1, j + 1, Zi + 1, j-1) (ai, j, bi, j, c
A motion vector detecting device comprising a motion vector detecting means for detecting a block having the smallest sum of i, j, di, j) as a predicted block. 2. The motion vector detecting device according to claim 1, wherein a block having the smallest sum of the differences between the adjacent horizontal pixels is detected as a predicted block. 3. The motion vector detecting device according to claim 1, wherein a block having the smallest sum of the differences between the adjacent pixels in the vertical direction is detected as a predicted block. 4. The motion vector detecting device according to claim 1, wherein a block having the smallest sum of the differences between the adjacent oblique pixels is detected as a predicted block. 5. A block in which the sum of the differences between the pixels in two or more of the adjacent horizontal, vertical, and oblique directions is detected as a predicted block. 2. The motion vector detection device according to 1. 6. The motion vector detecting device according to claim 1, wherein a raw image before compression is used as said input block and a block whose motion is to be predicted. 7. The apparatus according to claim 1, wherein a raw image before compression is used as said input block, and an image reproduced after compressing said block whose motion is to be predicted is used.
6. The motion vector detecting device according to claim 5, wherein: