JPH0562875B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention is a motion detection device that is applied to detect movement due to panning of a television camera, for example, when transmitting the imaging output of a television camera.

[Conventional technology]

A block matching method is known as one method of detecting a motion vector of a television signal. The block matching method divides the screen into many small areas (blocks), compares the block of interest in the previous frame with the neighborhood of the block of interest in the current frame, and detects the block with the highest correlation.

One method for detecting the magnitude of this correlation is to find the absolute value of the difference (frame difference) between corresponding pixel data between two blocks to be compared, integrate the above absolute values for all pixels in the block, and This method is to find the minimum integrated value. The block with the minimum integrated value is detected as the position after movement of the block of interest. This method can perform motion detection with block precision.

Another method for detecting the magnitude of the correlation is to move a block in the current frame whose center coincides with the block of interest in the previous frame pixel by pixel, and find the position where the frame difference within the moving range, that is, the investigation range, is minimum. This is the way to find out. According to this method, motion can be detected with pixel precision.

[Problems to be solved by the invention] None of the above-mentioned conventional block matching methods
Since this is an exhaustive survey in which all data in areas that can be considered as the amount of movement is compared, there is a drawback that the number of comparisons required for detection is extremely large and it takes time to detect movement. Furthermore, when determining the correlation, if there are a plurality of extreme values of approximately the same size, the smallest value among them is simply selected. This poses a problem in that when motion detection such as panning of a television camera is performed and motion correction is performed based on the result, the image of interest may be degraded.

For example, if you use a TV camera to photograph a scene with a mountain range as a background image and an airplane (the image of interest) flying in front of the mountain range, the TV camera will pan after the airplane so that the image of interest is located in the center of the viewfinder. be done. When such screen motion detection is performed, there is an extreme value near the center of the screen that is commensurate with the area of the image of interest, and on the other hand, there is also a large mountain range (background image) at a position corresponding to the movement of the TV camera. There are extreme values that are suitable for this. In this case, if an extreme value corresponding to the background image is selected, the image of interest will deteriorate as a result of motion correction. In other words, motion compensation detects the movement of the TV camera, shifts the coordinates by the amount of screen movement, and reads the image data of the previous frame from memory. Image movements cannot be reproduced correctly.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a television signal motion detection device that can perform appropriate motion detection without causing deterioration of the image of interest when used for motion detection of a television camera. be.

Further, the present invention provides a motion detection device that reduces processing time and has simple hardware, unlike detecting extreme values by examining all frame differences in a motion investigation range.

[Means for solving problems]

This invention divides one screen into a plurality of blocks, and divides each of the pixel data of a predetermined range corresponding to the representative point pixel data at the approximate center position of the block of the previous frame and the pixel data of the representative point within the block of the current frame. Means 1, 2, 3 for calculating the absolute value of the difference between
has an address corresponding to a predetermined range, writes the absolute value of the difference to the address, integrates the absolute value of the difference for each address over the period of the parsial field, and stores the integration result as the first table. A first storage means 5 having addresses corresponding to a predetermined range, writes the absolute value of the difference in the address, integrates the absolute value of the difference for each address in an even field period, and stores the integration result. a second storage means 6 for storing the table as a second table; detecting the minimum value in the first table stored in the first storage means 5 during the even field period, and detecting the minimum value in the first table stored in the first storage means 5 during the even field period; , and means 7 for detecting the minimum value in the second table stored in the second storage means 6.

[Effect]

Equipped with two block-sized graph memories,
The write operation and read operation of one graph memory 5 and the other graph memory 6 are reversely switched for each field. The CPU 7 accesses the graph memory 5 or 6 during the read period to detect extreme values according to a predetermined program.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings. This embodiment is basically a block matching method, and is an example in which the present invention is applied to motion detection of a television camera. That is, in this embodiment, in order to reduce the number of integrations, the central pixel (representative point) in the space of the block of the previous frame is
This method calculates the difference between each pixel in the corresponding block of the current frame, and integrates the differences for each block thus determined over one frame to detect the movement of one screen, that is, the movement of the television camera. be.

In FIG. 1, reference numeral 1 denotes a representative point memory for storing representative point data of each block in one frame of input image data. In the subtraction circuit indicated by 2, the difference between the representative point of the previous frame and the pixel in the corresponding block of the current frame is calculated. The output of the subtraction circuit 2 is converted into an absolute value by the conversion circuit 3,
The signal is supplied to an adder circuit 4. The output of this adder circuit 4 is written into the first graph memory 5 and the second graph memory 6 alternately for each field.

Graph memories 5 and 6 are coupled to data bus 8 of CPU 7, and outputs of graph memories 5 and 6 are supplied to selector 9. All '0' data is supplied as the other input of this selector 9, and for the pixels of the first block of one field, '0' data is selected by the selector 9 and supplied to the adder circuit 4. Ru. The graph memories 5 and 6 each have an address corresponding to one block, in which the absolute value of the difference data regarding the pixel at the position corresponding to each address is written, and supply the absolute value of the read difference data to the addition circuit 4. By writing the output of this adder circuit 4 to the original address,
An integrated value of difference data for one field (frame difference integration table) is stored in each.

The CPU 7 refers to the frame difference integral tables stored in the graph memories 5 and 6 to detect extreme values and verify whether the detected extreme values are correct. In this case, the graph memory 5 and the graph memory 6 perform the write operation (however, as described above, in order to integrate the frame difference, the read operation is performed before the write operation) and the read operation is performed for each field. These operations are controlled to be performed alternately and in opposite phases.

FIG. 2A shows a flag signal sent from the graph memories 5 and 6 to the CPU 7 during the vertical blanking period. FIG. 2B shows the operation of the graph memory 5, in which the graph memory 5 performs read operations in odd-numbered fields, and the graph memory 5 performs write operations in even-numbered fields. Figure 2 C is
This shows the operation of the graph memory 6, in which the graph memory 6 performs a write operation in odd-numbered fields, and the graph memory 6 performs a read-out operation in even-numbered fields. The frame difference integral data read from each of the graph memories 5 and 6 is supplied to the CPU 7,
Extreme values are detected and verified.

Since it takes one frame time to integrate the absolute value of the frame difference detected for each block over the entire frame by the representative point memory 1, subtraction circuit 2, and conversion circuit 3, the input image data is The motion vector written in the frame memory 10 and determined is supplied from the CPU 7 to the addition circuit 11, and this motion vector is added to the image data and transmitted. On the receiving side, motion correction is performed using the motion vector to shift the coordinate axes.

FIG. 3 shows an example of dividing the screen of one field. In this embodiment, a high-definition television signal is processed, and the size of one block (range of motion amount) is 8 lines in the vertical direction and 32 samples in the horizontal direction. inside
Contains 256 pixels, one field is vertical
Divided into 64 blocks and horizontally divided into 44 blocks.

The frame difference integral table stored in each of the graph memories 5 and 6 is as shown in FIG.
It is defined by X-Y coordinates centered on the origin (0,0). This X-Y coordinate takes values from -16 to 15 on the X axis and -4 on the Y axis.
It takes values from to 3. In the graph memories 5 and 6, a frame difference integral table as schematically shown in FIG. 5 is formed, where the vertical axis represents the absolute value of the difference value. The coordinate data of the minimum value (Xmin,
Ymin) is detected by the CPU 7. The vector connecting the origin and the detected (Xmin, Ymin) is the motion vector.

The motion vector detection and motion vector verification performed by the CPU 7 using the frame difference integral table data stored in the graph memories 5 and 6 will be described below in FIGS. 6, 7, 8, 9, and 9. This will be explained with reference to FIG.

As shown in FIG. 6, motion detection begins from the origin (steps 21 and 22). 7th
The figure shows four pieces of frame difference integral data located adjacent to each other in four directions: above, below, left and right of the origin. The extreme value detection at this time is (0, 0) (0, -1) (1, 0) (0, 1) (-1,
This is done by detecting the minimum value in the data of each coordinate of 0) (in FIG. 6, TABLE is added to represent the data). It is checked whether the coordinates (Xmin, Ymin) of this minimum value match X and Y (in this case, (0, 0)) (step 23). If they match, the process moves to step 25 of extreme value verification. If they do not match, the coordinates of the minimum value (Xmin, Ymin) are
(Step 24), and the minimum value of adjacent data located in four directions centered on these coordinates (X, Y) is detected again (Step 22).

In this way, the extreme value detection starts from the origin and proceeds in the direction of the maximum negative slope among the four directions, up, down, left and right, and finds a location where there is no negative slope in any of these four directions as an extreme value. In step 23, if the coordinates of the minimum value match the previous coordinates, the process moves to step 25 of extreme value verification.

As shown in FIG. 8, this step 25 includes (X+1, Y-1), (X+1, Y+1), (X
-1, Y+1), (X-1, Y-1) coordinates and the frame difference integral data of the extreme value coordinates (X, Y) is detected. . When the coordinates (Xmin, Ymin) of the detected minimum value match (X, Y), it is determined that the detected extreme value is correct.
If they do not match, the process moves to step 24 and extreme value detection is performed again.

FIG. 9 shows a more specific flowchart of step 22 of extreme value detection, and FIG. 10 shows a more specific flowchart of extreme value verification. In FIGS. 9 and 10, ":" means a comparison of data sizes. Furthermore, the display of "TABLE" which means coordinate data is omitted in these figures.

In extreme value detection, first a comparison is made with the pixels on the top, then with the pixels on the right, then with the pixels on the bottom, and finally with the pixels on the left. A comparison is made. In the first step 31 of extreme value detection, it is checked whether the coordinate Y coincides with -4. When they match, it means that there is no data located in the upward direction, so the detection of extreme values in the upward direction is skipped and the process moves to the detection of extreme values in the right direction. In step 32, the coordinate data of (Xmin, Ymin) detected as the minimum value and the coordinate (X, Y-1) located above it are
This is the step of comparison with the data.

When the data at these coordinates (Xmin, Ymin) is larger, frame difference integral data at (Y-1) is replaced in place of Ymin (step 3).
3). If the data at coordinates (Xmin, Ymin) is smaller than the data at coordinates (X, Y-1), Ymin is not changed.

When the comparison with the upward frame difference integral data is completed, a comparison with the frame difference integral data on the right side is started. In this case, it is first checked whether (X=15) (step 34). If this relationship holds true, there is no data to compare on the right side, so step 3 of comparison with the data on the right side is performed.
5 is jumped.

This step 35 is similar to step 32, and compares the magnitude of the frame difference integral data detected as the minimum value in the detection up to now with the frame difference integral data on the right side thereof. And when the data on the right side is smaller, Xmin becomes (X-1)
(step 36). Otherwise, this replacement is not made.

Furthermore, the size is then compared with the lower frame difference integral data consisting of steps 37, 38, and 39, and finally the size is compared with the left frame difference integral data consisting of steps 40, 41, and 42. will be done. In this way, the coordinates of the minimum value are determined among the four directions (up, down, left and right) of the initial coordinate position for extreme value detection.

FIG. 10 is a flowchart showing a more specific procedure of step 25 of extreme value verification in FIG. Extreme value verification is performed using the data of the coordinates (X, Y) of the extreme value detection mentioned above and the coordinate (X+
1, Y-1) is first compared in size with the frame difference integral data. In this case, (X=15)
It is checked whether or not (Y=-4) holds (steps 43 and 44), and when these relationships hold, there is no data diagonally above the right, so comparison step 45 and replacement step 46 are performed. Jumped.

Next, the size is compared with the frame difference integral data diagonally lower right. In this case, (Y=3)
At this time, the comparison step 48 with the data on the lower right side and the replacement step 49 are skipped.

Furthermore, next, the size is compared with the frame difference integral data diagonally lower left. In this case, (X=-
16) or (Y=3) is established (steps 50 and 51), and when these are established, there is no lower left diagonal data to be compared, so this comparison step is 5
2 and replace step 53 is skipped.

Finally, the size is compared with the frame difference integral data diagonally above the left. In this case, (Y=-4)
It is checked whether the following holds true (Step 5)
4). When this relationship is established, there is no data to be compared, so the comparison step 55 and the replacement step 56 are skipped.

In the manner described above, it is verified whether the detected extreme value is really an extreme value, and erroneous extreme value detection is prevented.

〔Effect of the invention〕

According to this invention, unlike detecting extreme values by examining all the data in the frame difference integral table, in a small area connecting the origin and the extreme values with a straight line,
All that is required is a comparison, which greatly reduces processing time and makes it possible to verify detected extreme values. Further, since the present invention does not require high-speed operation, extreme values can be easily detected under CPU control, and the scale of the hardware can be reduced.

Furthermore, when there are a plurality of extreme values in the frame difference integral table, instead of simply selecting the one with the smallest value, a reasonable one is checked. That is, the present invention enables detection according to the movement of the image of interest, avoids deterioration of the image of interest, and detects the movement of the image by detecting extreme values near the origin when the television camera does not move. Can remove the influence of objects.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of an embodiment of this invention, Fig. 2 is a time chart used to explain the operation of a graph memory in an embodiment of this invention, and Fig. 3 is an explanation of screen division in an embodiment of this invention. 4 and 5 are schematic diagrams used to explain a frame difference integral table according to an embodiment of the present invention, and FIG. 6 is a diagram showing extreme value detection and extremum values according to an embodiment of the present invention. A flowchart used to explain the verification, FIGS. 7 and 8 are schematic diagrams used to explain extreme value detection and extreme value verification, and FIG. 9 is a flowchart showing the procedure for extreme value detection in an embodiment of the present invention. FIG. 10 is a flowchart showing the procedure for extreme value verification according to an embodiment of the present invention. 1: Representative point memory, 2: Subtraction circuit, 4: Addition circuit, 5, 6: Graph memory, 7: CPU, 21:
Step of extreme value detection, 25: Step of extreme value verification.

Claims (1)

[Scope of Claims] 1. A screen is divided into a plurality of blocks, and pixels in a predetermined range corresponding to the representative point pixel data at the approximate center position of the block in the previous frame and the representative point pixel data in the block in the current frame are divided. means for calculating the absolute value of the difference between each of the data, and an address corresponding to the predetermined range, writes the absolute value of the difference to the address, and writes the absolute value of the difference for each of the addresses to an odd field. a first storage means for accumulating over a period of a second storage means for accumulating the absolute value of the difference for each in an even field period and storing the accumulation result as a second table; and means for detecting the minimum value in the first table and detecting the minimum value in the second table stored in the second storage means during the odd field period. Features: Television signal motion detection device.