JP3035420B2 - Image motion vector 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 a method for detecting a motion vector of an image, and more particularly to a method for detecting a motion vector of an image which is not affected by a change in brightness or flicker of a light source.

[0002]

2. Description of the Related Art The present inventors disclosed in Japanese Patent Application No. Hei 3-306445 a method of using a motion vector of an image when monitoring unauthorized entry into a restricted area with an image. To the extent necessary for the description of the invention, this approach is abbreviated with reference to FIGS. In Figure 2, 3 and 4, the vector image forming means 1, the image I _t of the frame F for example Figure 6 of the observation area continuous images generated is divided into blocks B of a certain number of pixels, the current image B of each block _The image in _t is replaced with a motion vector V from the previous image _Bt-1 to the current image Bt of the block, and the frame F is converted into a vector image Iv in FIG. 7 and output. After the motion vector V is binarized by the binarization unit 2 and the motion vector image Iv is converted into a binary image, the identification unit 3 determines whether the binary image is a monitoring target in the observation area. Identify. The identification means 3 in the illustrated example removes minute connected components in the binary image by contraction image processing and diffusion image processing using the contraction diffusion means 20, and performs labeling by the labeling means 22 and binary image by the shape feature extraction means 24. After the detection of the shape feature of each connected component and the detection of the motion feature of the pre-transformation motion vector of the connected component by the motion feature extraction means 26, the detected shape and motion feature are converted to the shape feature and motion of the predetermined monitoring target. Features and means of comparison 28
The monitoring target in the observation area is identified by the comparison.

[0003] A motion vector V is obtained by a so-called block matching method. According to this method, as shown in FIG. 3, a frame F of an image is divided into blocks B each having a predetermined number of pixels, for example, m pixels in the horizontal direction and n pixels in the vertical direction. Frame F has x pixels horizontally and y pixels vertically
In the case of having pixels, this frame F is horizontally (x
/ m) blocks and vertically (y / n) blocks. The coordinates X and Y of the block coordinate system when the coordinates of the pixel coordinate system are represented by x and y can be defined as follows.

[0004] X = (x / m), Y = (y / n) ··· (1) to extract the motion vector V, first 4 current specific block in the current frame F _t in (a) For the image _Bt , FIG.
The portion corresponding to the block current image _Bt in the previous frame _{Ft-1 of} (b) (hereinafter sometimes referred to as the previous image It _-1 ) is sequentially moved one pixel at a time in the horizontal and vertical directions as indicated by arrows. Move in the direction. Each time one pixel movement, detects the luminance difference between the image I _t-1 of the pixel before overlapping the pixel of the current image B _t and the pixel in the current image B _t, the current image B _t of the specific block And the sum of the absolute values of the luminance differences for all the pixels within the pixel, that is, the sum of the differences E, is stored together with the movement amount (i, j) at that time. FIG.
In (b), the previous image It _-1 is horizontally shifted from -μ to μ.
(−μ ≦ i <μ, see FIG. 1 (B)) When the pixel is moved one pixel at a time and vertically moved from −ν to ν (−ν ≦ j <ν, not shown), one pixel If the luminance of the image at the coordinates x and y is represented by f (x, y), the luminance difference sum E (i, j) at the movement amount (i, j) can be represented by the following equation.

[0005]

(Equation 1)

In this way, 2 μx2ν difference sums E calculated for each block B _t for each pixel movement are stored, and the minimum difference sum E _m among all the stored difference sums E is stored.
Movement amount corresponding to M (i _m, j _m) the motion vector V (V _i =
i _m , V _j = j _m ). Here, the movement amount M (i, j) indicates the movement of i pixels in the horizontal direction and j pixels in the vertical direction. In short, the motion vector V corresponds to the movement amount from the previous image to the current image which gives the minimum difference sum E _m.

FIG. 5 shows an example of the configuration of the motion vector image forming means 1. Digital input images I which are input continuously in time are alternately stored in the first current image storage means 12a or the second current image storage means 12b by the first switching means 10a.
At the same time, the second current image storage means 1 is switched by the second switching means 10b.
2b or alternatively already-stored image from the first current image storage means 12a is read in blocks as the current image I _t. At this time, the input image is written in the first current image storage unit 12a or the second current image storage unit 12b that is not currently being read, and the second current image storage unit 12b or the first image that is not currently being written is written.
It is read from the current image storage means 12a. Read current image I
_t is applied in parallel to the third switching means 10c and one input of the block matching means 14 described below.

The third switching means 10c is alternately connected to the first previous image storage means 16a or the second previous image storage means 16b, and the fourth switching means 10d is connected to the second previous image storage means 16b or the first previous image storage. It is connected alternately to the means 16a to read the previous image It _-1 .
Also in this case, the data is written to the first previous image storage unit 16a or the second current image storage unit 16b that is not currently read, and the second previous image storage unit 16b or the first previous image that is not currently written is written. It is read from the storage means 16a. The previous image It _-1 read in block units is applied to the other input of the block matching means 14.

The writing time and the reading time of both the current image storage means 12a and 12b and the previous image storage means 16a and 16b are all equal. Further, the processing of the block matching means 14, that is, the calculation processing of the motion vector V for each block is performed within the writing time or the reading time. Under such a configuration, the first to fourth switching means 10a, 10b, 10c, and 10d are alternately operated as described above, so that digital images input continuously in time can be sequentially converted in real time. Processing.

In this example, the motion vector V from the block matching means 14 is passed through the filter means 17 and then used by the image forming means 18 for the motion vector image Iv. this is,
This is to prevent a finite value motion vector V from being erroneously generated due to noise of the image signal or the like when there is no image change and no motion vector V actually exists. For example, the threshold value V _t (V _it, V _jt) with respect to the motion vector V is set to, V _i <V _it
When V _j <V _jt , the motion vector V is forcibly set to V = 0. Here, V _i and V _it is respectively in the horizontal direction component of the motion vector V and the motion vector threshold V _t, the vertical component of V _j and V _jt motion vector V and the motion vector threshold V _t Shown respectively.

[0011]

In the above-described motion vector identification method, there is a problem of vector noise caused by changes in solar radiation. In other words, solar radiation changes over time, such as the difference between daytime and evening, changes due to weather conditions such as the presence or absence of clouds and the intensity of rain and snow, temporal changes in shading by buildings, and other changes. Therefore, if there is a change in the solar radiation before and after the detection of the total difference E of the luminance, there is a possibility that the total sum E of the large luminance is generated and the motion vector is erroneously generated even though the monitoring target does not move. Since the filtering means 17 of the related art selects the motion vector once generated, it cannot prevent the generation of the motion vector due to the change of the solar radiation, and may not be able to perform a sufficient malfunction prevention function. .

In the case of artificial lighting using electric lights or the like, uneven brightness due to the difference in distance from the light source cannot be avoided.
If the light intensity of the light source fluctuates in a short cycle due to so-called flicker or the like, the non-uniform fluctuation in the brightness acts as noise, which may cause an undesirable motion vector.
The conventional filter unit 17 cannot prevent the erroneous generation of the motion vector due to such noise.

Furthermore, objects such as a tree or branches and leaves swaying due to the wind and a flag that repeatedly moves at a fixed position are noises that induce erroneous generation of motion vectors, even though detection is unnecessary. Therefore, it is an object of the present invention to provide a method for detecting a false
An object of the present invention is to provide a method for detecting a motion vector of an image, which can reliably prevent leakage .

[0014]

For simplicity, referring to FIG. 1 showing only horizontal movement, the method of detecting a motion vector of an image according to the present invention employs an observation area image I which is continuously generated for a certain observation area. to the current image B _t of consisting longitudinal n pixels in the horizontal m pixels in _t block, said block current image B _t in the pre-image I _t-1
Is moved one pixel at a time in the vertical and horizontal directions in the vicinity of the portion corresponding to the moving amount M (i, j) consisting of the number of horizontal moving pixels i and the number of vertical moving pixels j from the original position before the movement for each pixel. and the absolute value of the difference between the current image B _t and the luminance f of each pixel between the current image corresponding parts in the previous image I _t-1 after the movement by the movement amount (x, y) (| f t (x _{, y) -f t-1 (} x + i, y + j) |) and stores the sum obtained by adding all the pixels of the current image B _t as the difference sum E, all stored in the difference sum movement amount corresponding to the minimum difference sum _{E m (i m, j m} ) of the block motion vector V (V _i, V _j) a method according to, the moving amount is zero (i = 0, j = 0 ) the difference serving as a difference sum of the initial and minimum deviation ΔE (= E _z -E _m) is constant for each of the blocks between the differential sum is E initial difference sum E _z and the minimum difference sum E _m when the
Meta time less than a predetermined threshold value T (ΔE <T) is made of the motion vector V of the block as zero _{(V i = 0, V j} = 0).

[0015]

The operation will be described for the case of solar radiation change shown in FIG. At the time of the previous image It _-1, the luminance distribution f _t-1 (x, which has the dotted line shape of FIG. 1A and is indicated by the dotted line in the horizontal direction of FIG.
The previous image (not shown) of the block with y) has the shape and luminance distribution indicated by the solid line at this moment due to the increase in luminance due to changes in solar radiation.
Assume that a current image B _t having f _t (x, y) is generated.

[0016] In the prior art, in the case of FIG. 1 showing only horizontal movement for simplicity, the previous image I _t-1 of the luminance distribution f
The high-brightness part in _t-1 (x, y) horizontally _shifts the previous image It _-1 by i.
It is assumed that the pixel is moved to the position closest to the high-luminance portion in the current image _Bt by moving by the pixel. When this, taking the luminance difference between the current image B _t previous image I _t-1 for each pixel, the minimum difference sum E _m is obtained as shown in FIG. 1 (E). In the prior art,
Once the minimum difference sum _Em is found, the minimum difference sum _Em
Movement amount i of the previous image I _t-1 to give directly becomes a motion vector V = V _i. The white arrow in FIG. 1 (C) indicates this movement amount i.
Is shown. The problem with this prior art is that, despite the fact that there is no actual image movement, a motion vector V is generated due to luminance noise.

Referring to FIG. 17A, the block current image B _t
Is not moving from the corresponding block in the previous image It _-1 of the frame, and the luminance of the entire frame changes almost uniformly as shown in FIG. In the figure, only block current image B _t of the frame current image in I _t is assumed luminance of the entire frame current image in I _t is that the same change as well in block current picture B _t. Initial difference sum E
It is clear that _Z becomes as shown in FIG. In practice of the previous frame image I _t-1 of the luminance distribution f _{t over 1} (x, y) and the frame current image I _t of the luminance distribution f _t (x, y), even when they are seemingly part of a flat distribution Since there is also a small amount of noise, there can be a difference sum E smaller than the initial difference sum E _Z ,
Movement amount i to provide an initial difference sum minimum difference sum E _m very close magnitude E _Z as in FIG. 17 (E) may also be present as in FIG. 17 (C). That is, the prior art has a disadvantage that the motion vector V is generated even though the actual image has no motion.

The present inventor has proposed that even in the case of FIG.
By obtaining the initial difference sum E _Z and minimum difference sum initial and minimum deviation ΔE difference sum as the difference between E _m as shown in FIG. 17 (F),
The initial-minimum deviation ΔE represents only the motion of the image, the luminance noise such as solar radiation changes in the previous frame image I _t-1 and the current image I _t was noted to be erased. That is, by using the initial / minimum deviation ΔE of the sum of differences, it was possible to avoid the above-mentioned inconvenience and successfully generate a vector that moves only for the motion of the image.

As can be understood from the description of the case of solar radiation change, the occurrence of the minimum difference sum _Em and the motion vector V due to luminance noise exists even when there is no solar radiation change.

Referring to FIG. 18, it will be described with reference to FIG. 18 that it is difficult to remove luminance noise due to a change in solar radiation by applying a threshold value in a conventional method for detecting a motion of an image only by a luminance change of a frame image I. That is, if a large threshold value T is selected as in the high solar radiation side shown in FIG. 18 so that the luminance change is not caused to move the image only to the solar radiation change amount E _zl during the high solar radiation, the low solar radiation side in FIG. As a result, a slight change in brightness E _zv due to the image movement cannot be detected,
This is contrary to the original purpose of image movement detection. Further, if the threshold value T is reduced to detect a small luminance change _Ezv due to image movement as shown on the low solar radiation side in FIG. 18 during low solar radiation, noise due to the solar radiation change during high solar radiation cannot be prevented.

According to the method of the present invention, even if there is a change in solar radiation, no motion vector is generated as described below. That is, the initial difference sum E _z took luminance difference between the current image B _t before without moving the image I _t-1 is as shown in FIG. 1 (D), the current image B _t luminance distribution f _t It has the same shape as (x, y). This initial difference sum E _z and Figure 1 (E) minimum difference sum E _m and the difference sum of the initial and minimum deviation ΔE as the difference between the (= E _z -E _m)
Is as shown in FIG. 1 (F). In this case, the luminance difference due to solar radiation changes, because it contains the initial difference sum E _z and minimum difference sum equal respectively as E _zl and E _ml to both E _m, the initial-minimum difference sum E is the difference between A luminance difference due to a change in solar radiation does not appear in the deviation ΔE. If the components of the monitored initial difference sum E _z and minimum difference sum due to the movement E _m, respectively E _zv and E _mv, the value of the initial and minimum deviation ΔE difference sum in the illustrated example, FIG. 1 (F ), It is equal to (E _zv −E _mv ), and depends only on the movement of the monitoring target and is not affected by solar radiation. Referring to FIG. 1 (G), a threshold value T for the initial / minimum deviation ΔE of the sum of the differences is shown.
Is set to a value larger than ΔE due to insolation change and other undesired random noise that are insignificant for each block . Reasons for setting a threshold for each block
As described later, when the brightness of the monitoring area is not uniform,
When there is an undetected object that moves repeatedly within the monitoring area
Because there is a case. Therefore, if the initial / minimum deviation ΔE of the sum of differences is equal to or smaller than the threshold value T determined for each block , no motion vector is generated.

The symbol d in FIG. 1 indicates the width of the high-luminance portion in the block _Bt or _Bt-1 in the x-axis direction. FIG.
In the example of (G), the absolute value of the threshold T is smaller than the minimum difference sum E _m, merely the threshold minimum difference sum E _m T
In comparison with, the effects of changes in solar radiation cannot always be avoided.

Thus, the object of the present invention is to provide a "method of detecting a motion vector of an image which can reliably prevent erroneous detection or omission of detection due to the influence of noise". FIG. 11 illustrates the above method of the method for detecting a motion vector of an image according to the present invention.
A configuration is shown in which the reliability determination unit 17a for the motion vector is provided between the block matching unit 14 and the image forming unit 18 in FIG.

FIG. 12 is a flowchart of a method for detecting a motion vector of an image according to the present invention. Step 101 calculates the initial and minimum deviation ΔE difference sum E, the set thresholds T as a constant T _C in this case for each block in step 102, an initial-minimum deviation ΔE starvation of the difference sum at step 103 It is determined whether or not it is equal to or smaller than the threshold value T _C (ΔE> T _C ), and T
Movement amount when it is _C or less is set to zero the motion vector V in step 104, when more than T _C corresponding motion vector V to the minimum difference sum E _m at step 105 (i _m, j _m)
Set to.

[0025]

Examples of EXAMPLE 8, no change in solar radiation, only by the necessary motion detecting brightness change is monitored, no other noise, the minimum difference sum E _m becomes zero, the This is the case where the initial / minimum deviation ΔE of the difference sum obtained in this way is equal to the initial difference sum E _z .

When the monitoring area is illuminated by an artificial light source located inside or on one side of the monitoring area, the brightness of the monitoring area is not uniform as shown in FIG. FIG. 9 shows a case where a luminance gradient occurs along the x-axis. When flicker occurs in the artificial light source in this state, the luminance distribution f of the previous image
_{A change from t-1} (x) to the luminance distribution f _t (x) of the current image occurs.
The change is an increase or decrease in the inclined brightness distribution as shown in FIG. Also in this case, the initial difference sum E _z is determined as shown in FIG. 9C, and the minimum difference sum _Em is moved by, for example, i pixels of the previous image It _-1 as shown in FIG. 9B. It is determined as shown in FIG. Moreover, the initial and minimum deviation of the difference serving as a difference sum of the initial difference sum E _z and minimum difference sum E _m delta
E is determined as shown in FIG. In this case, it is necessary to avoid generating the motion vector V because there is no movement of the object in the observation area and only a luminance change due to flicker has occurred. For this purpose, the threshold value T may be selected to be large. There are three methods.

The first approach, the initial difference sum E product aE _z select coefficient a as a function of _z is to the threshold T. As estimated from a comparison of FIGS. 8 and 9, when there is motion in the observation target, the initial and minimum deviation of the minimum difference sum E _m is relatively small becomes difference sum E of the initial difference sum E _z ΔE becomes relatively large. In contrast, in the case of the flicker, the initial difference sum E _z and minimum difference sum E _m are both relatively large, the initial and minimum deviation ΔE difference sum is relatively small. Focusing on this, for example, a threshold T that does not generate a motion vector is selected by the product aE _z for the initial / minimum deviation ΔE of the difference sum E of several tens% or less of the initial difference sum E _z. . Thus a threshold value determined as a function of the initial difference sum E _z may be referred to as a T _a in the description below.

The second approach is a product bM select the coefficients b as a function of the amount of movement M of the previous image I _t-1 for determining the minimum difference sum E _m which the threshold T. Since artificial illumination in the case of flicker as estimated from FIG. 9 (B), the likely brightest portion in the current image B _t of each block is one of the side edge. Minimum difference sum E _m
, It is necessary to move the previous image It _-1 to this brightest part on the current image _Bt , so that the previous image It _-1
Move amount M becomes large. For example the movement amount M shown by the white arrow i of FIG. 9 (B), moves the end of the corresponding block in the previous image I _t-1 to the end portion of the luminance gradient f _t of the current image B _t (x) Things. Focusing on this, based on the actually measured value of the relationship between the movement amount M and the initial / minimum deviation ΔE of the sum of differences E,
An appropriate coefficient b is determined, and when the movement amount M is large, a threshold value T that does not generate a motion vector even for the initial / minimum deviation ΔE of the considerably large sum of differences E is set to the product b
Selected by M. Thus a threshold value determined as a function of the amount of movement M of the current image B _t for obtaining the minimum difference sum E _m may be referred to as a T _b in the following description.

A third approach is a combination of the first and second approach, the minimum difference sum E previous picture I for determining the _{m t-1}
Both the amount of movement M and the initial difference sum E _z of the function f (E _m,
The threshold value T is obtained as M). This function f
(E _m , M) is determined by experiment. The function f (E _{m, M)} threshold defined as, may be referred to as T _ab by collectively following description tripartite of the threshold T _a and T _b.

A threshold T corresponding to noise such as flicker
The procedure for setting is shown in the flowchart of FIG. After obtaining the initial / minimum deviation ΔE of the difference sum E in step 110,
In steps 111, 112, and 113, the magnitude of the movement amount M of the previous image It _-1 for obtaining the minimum total sum _Em is determined by the absolute value | i | or | j | Determine whichever is greater. In step 114, any or all of the three threshold values T _ab are determined by the initial difference sum E _z , the movement amount M, and the function f (E _m , M). step 1
At 15, 116 and 117, the threshold value T _ab thus determined
There when smaller threshold T _c of the constants used for calculating the motion vector threshold T _c of the constant. When the threshold value T _ab is larger than the constant threshold value T _c ,
One of the three threshold values T _ab is appropriately selected and used for calculating the motion vector V. In steps 118, 119 and 120,
The motion vector V is calculated based on whether the initial / minimum deviation ΔE of the difference sum E is greater than a threshold value by the minimum difference sum E _m.
Of the previous image It _-1 to determine the distance i _m , j _m or zero.

The above-mentioned constant threshold value _Tc and the threshold value Tab considering flicker are both the initial value of the difference sum E and the threshold value _Tab.
Since it is used for the minimum deviation ΔE, it is not affected by changes in solar radiation.

The graph of FIG. 10, the block diagram of FIG. 14, and the mask method of FIG. 15, when the detection unnecessary object 50 such as the swaying of the branches and leaves of the tree or the object that repeatedly moves at a fixed position is present in the monitoring area.
This will be described with reference to the flowchart of FIG. Undetected object 50
Since the is difficult to identify only above a threshold T, stores the information about the previously detected unnecessary object 50 during the period Δt the detection necessary target 51 is not present, for each block
Is masked by modifying the threshold value T.

In FIG. 10, the image I of the monitoring area is represented by the block coordinates X and Y in the equation (1), and the image I ₁ during the period Δt during which the detection target 51 does not exist is shown in FIG. Target 51
The image I ₂ when but existing shown in FIG. 10 (B). For simplicity,
Initial _value of the sum of block differences at block coordinate Y ₀
The minimum deviation ΔE is shown in FIG. 10 (C) to FIG. 10 (F) for each block. When the detection-free object 50 is moving, the initial / minimum deviation ΔE of the sum of differences E is equal to the threshold value T as shown in FIG.
Take a value exceeding. Here, the case where the threshold T is a constant _Tc will be described. During the period Δt during which the detection required object 51 does not exist, the size of the positive portion of the difference D between the initial / minimum deviation ΔE of the sum of differences E and the threshold value T, that is, the portion exceeding the threshold value T, is determined for each block. Monitor and further Figure 10
As shown in (D), the maximum value D _m (=
[ΔE−T] _max ) is obtained. The sum of the maximum value D _m of the difference D and the threshold value T is determined as a mask threshold value T _m as shown in FIG. 10E (T _m = D _m + T). Incidentally, when the threshold T _m of a mask for each block initial and minimum deviation of the difference sum corresponding to the maximum value D _m of the difference D Delta] E or to the threshold value T and the constant threshold value T _c It matches the constant threshold value _Tc .

[0034] If the detection of the motion vector using the threshold T _m when the detected required target 51 appeared, was not yet detected subject in need 51 detects this by masking the motion detection unnecessary target 50 What can be detected is clear from FIG. 10 (F). In this case, the masking is performed for the initial / minimum deviation ΔE of the difference sum E, so that it is not affected by the change in the solar radiation.

[0035] FIG. 14 is a block diagram of a mechanism for the mask, the movement amount before the image I _t-1 for moving amount storage means 30 determine the minimum difference sum _{E m M (i m, j} m) Remember
Initial difference sum storage means 31 stores the initial difference sum E _z ,
Minimum difference sum storage unit 32 stores the minimum difference sum E _m,
Threshold excess maximum value storing unit 33 stores the maximum value D _m of the threshold excess D. Detection unnecessary object mask means 34
<Br There Referring to FIG. 10 calculates a masking threshold value T _m in the manner described above, when the detected required target 51 not recognized as a zero motion vector, deemed detection necessary target 51 In this case, the target motion vector V is stored in the moving amount storage means 30.
And output it. The mechanism for performing this mask is
It can be included in the determination means 17a of FIG. However, it may be after the image forming means 18 in FIGS.

[0036] In the flowchart of FIG. 15 showing a process of detecting the maximum value D _m of the threshold excess D, Step 13
The period Δt and the initial time t at which the detection required object 51 does not exist at 0
₀ is set, the initial / minimum deviation ΔE of the sum of differences E is calculated in step 131, the required threshold T is selected in step 132, and the excess D is calculated in step 133 in FIG. It is calculated in the manner described. Obtain the maximum value D _m of the threshold excess D in the period Δt in step 134, 135 and 136. In the flowchart of FIG. 16 illustrating the detection process according to the mask threshold, calculates the initial and minimum deviation ΔE difference sum E in step 140, reads the maximum value D _m of the threshold excess D in step 141, the mask Threshold value _Tm for step 142
Determined in. In steps 143, 144, and 145, detection of the detection-required target 51 and output of a required motion vector are performed.

[0037]

As described in detail above, the method for detecting a motion vector of an image according to the present invention uses the threshold value T for the initial / minimum deviation ΔE of the sum of differences E in luminance between temporally preceding and succeeding images. Therefore, the following remarkable effects are obtained. (A) Stable without erroneous detection or omission of detection without being affected by temporal changes in the solar radiation intensity during the day or evening, weather changes such as clear or cloudy, and changes in the shade and sun due to the shadow of buildings. Motion vector detection can be ensured. (B) Even if there is flicker in a discharge lamp such as a fluorescent lamp or a mercury lamp or other light sources, stable motion vector detection can be performed without erroneous detection or detection omission. (C) Masking the movement of the detection unnecessary object such as the swaying of the branches and leaves of the grove due to the wind, the rotation of the windmill, etc.
Reliable motion vector detection without erroneous detection or omission can be performed. (D) Since the detection is based on the initial / minimum deviation ΔE of the difference sum E of the luminance, it is not necessary to adjust the threshold value for the difference in the solar radiation condition. (E) Since the influence of noise is excluded before the generation of the motion vector, the load on the motion vector image processing can be reduced, and the time required for the recognition processing of the moving object can be shortened.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram of a method for detecting a motion vector of an image according to the present invention.

FIG. 2 is an explanatory diagram of a conventional block image generating device.

FIG. 3 is an explanatory diagram of image frames and blocks.

FIG. 4 is a principle explanatory diagram of block matching.

FIG. 5 is an explanatory diagram of a motion vector image forming unit.

FIG. 6 is an explanatory diagram of an observation area.

FIG. 7 is an explanatory diagram of a motion vector image Iv.

FIG. 8 is a graph illustrating a motion detection process according to the method of the present invention.

FIG. 9 is a graph illustrating a method of removing the influence of flicker.

FIG. 10 is a graph illustrating a mask method for a detection unnecessary object.

FIG. 11 is a block diagram showing the function of the method of the present invention.

FIG. 12 is a basic flowchart of the method of the present invention.

FIG. 13 is a flowchart of a detection method excluding the influence of flicker.

FIG. 14 is a block diagram of a masking method for a detection unnecessary object.

FIG. 15 is a first flowchart of the detection unnecessary target mask method.

FIG. 16 is a second flowchart of the detection unnecessary target mask method.

FIG. 17 is an explanatory diagram of a comparison between the conventional method and the method of the present invention.

FIG. 18 is an explanatory diagram of a threshold directly used for a luminance difference.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Motion vector image formation means 2 Binarization means 3 Identification means 5 Image input means 10a, 10b, 10c, 10d First, second, third, fourth switching means 12a, 12b First, second current image storage means 14 Block matching means 16a, 16b First and second previous image storage means 17 Filter means 17a Reliability determination means 18 Image forming means 30 Moving amount storage means 31 Initial difference sum storage means 32 Minimum difference sum storage means 33 Threshold exceeded Min. Maximum value storage means 34 Detection unnecessary target mask means 50 Detection unnecessary target 51 Detection required target 100 to 145 steps.

──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Yutaka Maeda 3-20-8 Narita Nishi, Suginami-ku, Tokyo Inside Okura Electric Co., Ltd. (72) Susumu Kikuchi 1-3-1, Uchisaiwaicho, Chiyoda-ku, Tokyo (56) References JP-A-4-105485 (JP, A) JP-A-6-46410 (JP, A) JP-A-5-143737 (JP, A) (58) Fields surveyed ( Int.Cl. ⁷ , DB name) G06T 1/00 G06T ^7/ 00-7/60 G08B 13/194-13/196

Claims (5)

(57) [Claims] 1. A contrast current image B _t blocks of n vertical pixels m horizontal pixels in the observation area image I _t continuously generated for a predetermined observation zone, the blocks in the pre-image I _t-1 the portion near that corresponds to the current image B _t moves by one pixel vertically and horizontally, for each pixel movement, the movement amount consisting lateral movement pixel number i and longitudinal movement pixel number j from the original position before movement M ( i, j) and the current image B _t
And the absolute value of the difference in luminance f (x, y) for each pixel between the current image corresponding portion in the previous image It _-1 after moving by the moving amount
_{(| F t (x, y} ) -f t-1 (x + i, y + j) |) of the sum obtained by adding all the pixels of the current image B _t is stored as a difference sum E, stored all minimum difference sum movement amount corresponding to E _m in the difference sum E of (i _m, j _m) of the block motion vectors V
(V _i , V _j ), wherein the moving amount is zero (i =
0, j = 0), the initial difference sum E _z that is the difference sum E
The minimum difference sum constant difference serving difference sum of the initial and minimum deviation ΔE (= E _z -E _m) within each of said blocks between E _m and
Meta predetermined threshold value T or less (Delta] E <T) image motion vector detecting method comprising the motion vector V of the block as zero _{(V i = 0, V j} = 0) when the. 2. The method according to claim 1, wherein each of the
The threshold value T for each block is set to the initial difference sum E of each block.
_A method of detecting a motion vector of an image that is changed according to the magnitude of _z . 3. The method according to claim 1, wherein each of the
Tsu threshold T for each click, said movement amount (i, j) of the size image motion vector detecting method comprising varied depending on at which gives the minimum difference sum E _m of each block. 4. The method according to claim 1, wherein each of said blocks
The threshold T for each click, the motion of the image formed by varied depending on the magnitude of the movement amount giving a magnitude and a minimum difference sum E _m of the initial difference sum E _z of each block (i, j) Vector detection method. 5. The method according to claim 1, wherein a period Δ in which no abnormality is detected in the observation area.
During t, when there are blocks in which the initial / minimum deviation ΔE of the sum of differences E is continuously larger than the threshold value T, the maximum value ΔE of the initial / minimum deviation ΔE of the sum of differences E during the period Δt _A method for detecting a motion vector of an image in which _m is a threshold value T _m of the subsequent block .