JPH10111946A - Image tracking device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image tracking device which can prevent the tracking failure to a non-rigid body or an object whose size varies and also can perform the fast arithmetic processing. SOLUTION: An area division means 403 divides a tracking object area into plural small areas, and an object model enlargement/reduction means 406 performs an enlargement/reduction operation to each divided area. Thus, a model is obtained. Then a partial Hausdorff distance calculation means 407 and a distance normalization 408 calculate the evaluation function value and then the moving destination and the enlargement/reduction ratio of every part of a tracking object against the model. At the same time, the color image data whose resolution is deteriorated compared with a series of color image frames which undergo no edge extraction nor binarization yet are stored in a low resolution image storage means 411. A color histogram calculation means 412 roughly calculates the moving destination of the tracking object based on the low resolution image data. A search area limitation means 413 limits a searching area and performs its processing at a high speed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image tracking device for tracking a specified object in a moving image captured by a video camera or the like.

[0002]

2. Description of the Related Art A method of tracking an object in a moving image is disclosed in, for example, Proceding of IEE ICCV93, pp. 93-101. Hereinafter, a conventional image tracking device will be described with reference to FIG. In FIG.
Is an edge image storage means, 502 is a tracking object input means, 5
03 is an object area storage means, 504 is an object model storage means, 505 is a partial Hausdorff distance calculation means, 506 is an object destination calculation means, and 507 is an object area update means.

[0003] The operation of the image tracking apparatus configured as described above will be described below. First, the user uses the tracking object input unit 502 to store a series of 2D images of the tracking target object stored in the edge image storage unit 501.
Call the first image frame from the value image frame,
A rectangular area surrounding the object to be tracked is designated and stored in the object area storage means 503.

Further, the object is tracked by repeating the following processing with n being a natural number. Object model storage means 5
Reference numeral 04 stores, as the object model M, a set of coordinates of the edge points of the n-th binary image frame I (n) existing in the area stored in the object area storage means 503. part
The Hausdorff distance calculating means 505 calculates a partial Hausdorff distance which is an evaluation function for obtaining a destination in the (n + 1) th image frame I (n + 1) of the object. For each object model M, n + 1 at all positions within the search range
A partial Hausdorff distance hK (M, I (n + 1)) from the image frame I (n + 1) is calculated. Here, the partial Hausdorff distance hK (P, Q) of the sets P and Q is defined by the following equation.

[0005]

(Equation 1)

The object destination calculating means 506 determines the (n + 1) th image frame I (n + n) by using the position giving the minimum value among the partial Hausdorff distances thus obtained.
The destination of the object model M in 1) is output to the object area updating means 507.

[0007] The object area updating means 507 outputs the (n + 1) th image frame I obtained from the object moving destination estimating means 506.
The content of the object area storage means 504 is updated according to the destination of the object model M in (n + 1).

[0008]

However, in the conventional configuration described above, since the tracking target is designated by a single area, the tracking target may be non-rigid or may have a large size in the image of the tracking target. However, there is a problem that the tracking is likely to fail when the data changes. In addition, there is a problem in that when calculating the destination of the tracking target object, the amount of calculation is increased in order to uniformly search in all directions from the current position, and the processing time is increased.

The present invention solves the above-mentioned problems of the prior art, and enables tracking to be successful even when the object to be tracked is non-rigid or its size in an image changes. An object of the present invention is to provide an image tracking device that enables high-speed arithmetic processing.

[0010]

In order to achieve the above object, according to an image tracking apparatus of the present invention, an area dividing means divides an object area to be tracked into a plurality of small areas and enlarges an object model for each of the small areas. For the model in which the reduction means performed the scaling operation, the partial Hausdorff distance calculation means and the distance normalization means
The moving destination and the enlargement / reduction ratio of each part of the object are calculated by calculating the evaluation function value, whereby it is possible to prevent a failure in tracking a non-rigid body or an object whose size changes. Further, the color histogram calculating means roughly determines the destination of the object using the low-resolution image, and the search area limiting means limits the range of the search, so that the processing speed can be increased.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS According to the first aspect of the present invention, there is provided an edge image storing means for storing a series of binary image frames subjected to edge extraction processing, and a first image from the edge image storing means. A tracking object input unit for extracting a frame and specifying an area including a tracking target object; and for an n-th image frame, when n is 1, the area obtained by the tracking object input unit is stored. In the case of a natural number of 2 or more, an object area storing means for storing an area obtained by the object area updating means, and an object model storing means for storing image information in the area stored in the object area storing means as an object model Object model scaling means for scaling the object model stored in the object model storage means, and each object model obtained by the object model scaling means and (n + 1) th Portion of the region of the image frame
A Hausdorff distance calculating means for calculating a Hausdorff distance, and a partial Ha obtained by the partial Hausdorff distance calculating means.
usdorff distance normalizing means for normalizing the distance by a scaling factor, object moving destination calculating means for calculating an area of the (n + 1) th image frame in which the distance calculated by the distance normalizing means takes a minimum value, and the object Object area updating means for outputting the area of the (n + 1) th image frame obtained by the movement destination calculating means to the object area storing means,
Object model scaling means is provided, and by comparing the next image frame with a template in which the object model is scaled in several stages, the parallel movement amount and the scaling ratio of the target object are obtained. Even if it changes, tracking failure can be prevented. At this time, the distance normalization means eliminates a change in the distance due to the scaling operation, and the correct destination can be calculated.

According to a second aspect of the present invention, there is provided an edge image storing means for storing a series of binary image frames subjected to edge extraction processing, and a first image frame is taken out from the edge image storing means. Tracking object input means for designating an area including the tracking target object, area dividing means for dividing the area obtained by the tracking object input means into a plurality of small areas, and n for an n-th image frame. 1
In the case of, the area obtained by the area dividing means is stored, and n
Is a natural number of 2 or more, an object area storing means for storing the area obtained by the object area updating means, and an object model storing for storing image information in the area stored in the object area storing means as an object model Means for calculating a Hausdorff distance between each object model obtained by the object model storage means and a certain area of the (n + 1) th image frame
Hausdorff distance calculation means, object movement destination calculation means for calculating the area of the (n + 1) th image frame in which the distance calculated by the partial Hausdorff distance calculation means takes a minimum value, and n + 1 obtained by the object movement destination calculation means An object area updating means for outputting an area of a second image frame to said object area storing means, provided with an area dividing means, and dividing a tracking target area into a plurality of small areas to perform tracking. Even when the target object is a non-rigid body, it is regarded as a set of a plurality of rigid bodies, and tracking of each rigid body can prevent tracking failure.

According to a third aspect of the present invention, there is provided an edge image storing means for storing a series of binary image frames subjected to an edge extracting process, and extracting a first image frame from the edge image storing means. Tracking object input means for designating an area including the tracking target object, area dividing means for dividing the area obtained by the tracking object input means into a plurality of small areas, and n for an n-th image frame. 1
In the case of, the area obtained by the area dividing means is stored, and n
Is a natural number of 2 or more, an object area storing means for storing the area obtained by the object area updating means, and an object model storing for storing image information in the area stored in the object area storing means as an object model Means, an object model enlargement / reduction means for enlarging / reducing the object model stored in the object model storage means, and each object model obtained by the object model enlargement / reduction means and a certain area of the (n + 1) th image frame. Partial Hausdorff partial Hausdorf to calculate distance
f distance calculating means, distance normalizing means for normalizing the partial Hausdorff distance obtained by the partial Hausdorff distance calculating means by a scaling factor, and the (n + 1) th distance at which the distance calculated by the distance normalizing means takes a minimum value. An object moving destination calculating means for calculating the area of the image frame of the above, and an object area updating means for outputting the (n + 1) th image frame area obtained by the object moving destination calculating means to the object area storing means It is a tracking device, provided with a region dividing means, by dividing the tracking target region into a plurality of small regions, even if the tracking target object is a non-rigid body, consider this as a set of a plurality of rigid bodies, and each rigid body Tracking can prevent tracking failure. In addition, an object model scaling unit is provided, and by comparing the next image frame with a template obtained by scaling the object model in several stages, the translation amount and the scaling ratio of the target object are obtained. In the case of a change, tracking failure can be prevented. At this time, the distance normalization means eliminates a change in the distance due to the scaling operation, and the correct destination can be calculated.

According to a fourth aspect of the present invention, there is provided a low-resolution image storage for storing color image data with reduced resolution for a series of color image frames before performing edge extraction processing and binarization processing. Means for calculating a color histogram in the area stored in the object area storage means for the nth low resolution image frame, and calculating the color histogram from the area stored in the object area storage means for the (n + 1) th low resolution image frame. A color histogram calculating means for calculating a color histogram of an area within a certain range; and a search range limiting means for roughly calculating an object moving destination by comparing the color histograms obtained by the color histogram calculating means. Hausdorff distance calculation means,
The image tracking apparatus according to claim 1, wherein a partial Hausdorff distance is calculated by limiting the search range output from the object movement range limiting unit.
A color histogram calculating means and a search range limiting means, wherein the color histogram calculating means comprises: a color histogram of an object region of an nth color image frame having a reduced spatial resolution stored in the low resolution image storing means;
The color histogram of a certain area is calculated, and the search area limiting means obtains the approximate destination of the object in the area having the color histogram closest to the color histogram of the object area, and restricts it to the surrounding area to calculate the partial Hausdorff distance. Since the means obtains the partial Hausdorff distance, the amount of calculation can be reduced and the processing speed can be increased.

Embodiment 1 Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of the image tracking device according to the first embodiment of the present invention. In FIG. 1, 101 is an edge image storage means, 102 is a tracking object input means, 103 is an object area storage means, 104 is an object model storage means, 105 is an object model enlargement / reduction means, 106 is a partial Hausdorff distance calculation means, 107 is Distance normalizing means, 108 is an object moving destination calculating means, and 109 is an object area updating means.

The operation of the image tracking device configured as described above will be described. Edge image storage means 101
Stores a series of binary image frames obtained by performing an edge extraction process and a binarization process on a moving image in which a tracking target object appears. The user specifies a rectangular area surrounding the object to be tracked from the first binary image frame I (1) by the tracked object input unit 102. Object area storage means 1
03 stores the area obtained by the tracking object input means 102 when n is 1 for the n-th image frame, and stores the object area updating means 10 when n is a natural number of 2 or more.
The area obtained in step 9 is stored.

The object model storage means 104 stores a set of coordinates of the edge points of the n-th binary image frame I (n) stored in the object area storage means 103 as an object model Mi. The object model enlargement / reduction unit 105 enlarges (reduces) the coordinate values of the object model. For example, the enlargement ratio Sx in the x direction and the enlargement ratio Sy in the y direction are independently set to 0.9,
1.0, 1.1}, a total of nine types of object models Mi [Sx, Sy] are generated.

In the partial Hausdorff distance calculating means 106,
A partial Hausdorff distance defined by Expression (1), which is an evaluation function for obtaining a destination and an enlargement / reduction ratio in the (n + 1) th image frame I (n + 1) of the object, is calculated. For each object model Mi [Sx, Sy], a partial Hausdorff distance hK (Mi [Sx, Sy], I (n) to the (n + 1) th image frame I (n + 1) at all positions within the search range.
+1)).

However, since the unit vector of the distance vector between the two points is different due to the enlargement / reduction operation, it is inconvenient as it is. Therefore, the distance normalization means 107 normalizes the distance. The normalization of the distance is performed by dividing the x component and the y component of the distance vector between the two points by Sx and Sy, respectively. Partial Hausdorff distance calculation means 10
In section 6, based on this normalized distance vector,
Calculate usdorff distance.

The object movement destination calculation means 108 calculates the movement of the object model Mi in the (n + 1) th image frame I (n + 1) based on the position giving the minimum value and the enlargement / reduction ratio among the partial Hausdorff distances thus obtained. The data is output to the object area updating unit 109 as the destination and the enlargement / reduction ratio.

The object area updating means 109 outputs the (n + 1) th image frame I obtained from the object moving destination calculating means 108.
The content of the object area storage unit 103 is updated based on the destination and the enlargement / reduction ratio of the object model Mi in (n + 1).

As described above, according to the present embodiment, the object model enlarging / reducing means 105 is provided, and the template obtained by enlarging / reducing the object model in several stages is compared with the next image frame, whereby the object Since the translation amount and the enlargement / reduction ratio are obtained, tracking failure can be prevented even when the size of the target object changes. At this time, the distance normalizing unit 107 can eliminate a change in distance due to the scaling operation, and can calculate a correct destination.

(Embodiment 2) FIG. 2 is a block diagram of an image tracking apparatus according to a second embodiment of the present invention. In FIG. 2, 201 is an edge image storage means, 202 is a tracking object input means, 203 is an area dividing means, 204 is an object area storage means, 205 is an object model storage means, 206 is a partial Hausdorff distance calculation means, and 207 is an object movement. The first calculating means 208 is an object area updating means.

The operation of the image tracking device configured as described above will be described. Edge image storage means 201
Stores a series of binary image frames obtained by performing an edge extraction process and a binarization process on a moving image in which a tracking target object appears. The user specifies a rectangular area surrounding the object to be tracked from the first binary image frame I (1) using the tracked object input unit 202. Region dividing means 203
Divides a rectangular area obtained by the tracking object input means 202 into a plurality of small areas not exceeding a certain size (the number of divisions is m), and each small area Ri (i = 1,
The coordinates of the two vertices (2,..., M) are stored in the object area storage unit 204.
To be stored.

Further, the object is tracked by repeating the following processing with n being a natural number. The object model storage unit 205 stores the n-th binary image frame I existing in the small region Ri stored in the object region storage unit 204.
A set of coordinates of the edge point (n) is stored as the object model Mi.

In the partial Hausdorff distance calculating means 206,
A partial Hausdorff distance defined by Expression (1), which is an evaluation function for obtaining a destination in the (n + 1) th image frame I (n + 1) of the object, is calculated. Each object model Mi
For [x, y], a partial Hausdorff distance hK (Mi [x, y], I (n + 1)) from the (n + 1) th image frame I (n + 1) at all positions within the search range is calculated.

The object movement destination calculating means 207 determines the (n + 1) th image frame I (n + n) with the position giving the minimum value among the partial Hausdorff distances thus obtained.
The destination of the object model Mi in 1) is output to the object area updating means 208.

The object area updating means 208 outputs the (n + 1) th image frame I obtained from the object moving destination calculating means 207.
The contents of the object area storage means 204 are updated according to the destination of the object model Mi in (n + 1).

As described above, according to the present embodiment, the area dividing means 203 is provided to divide the area to be tracked into a plurality of small areas, so that even if the object to be tracked is non-rigid, By considering each rigid body as a set of a plurality of rigid bodies, tracking failure can be prevented.

(Embodiment 3) FIG. 3 is a block diagram of an image tracking apparatus according to a third embodiment of the present invention. 3, reference numeral 301 denotes an edge image storage unit, 302 denotes a tracking object input unit, 303 denotes a region dividing unit, 304 denotes an object region storage unit, 305 denotes an object model storage unit, 306 denotes an object model scaling unit, and 307 denotes a partial Hausdorff. A distance calculating unit 308 is a distance normalizing unit, 309 is an object moving destination calculating unit, and 310 is an object area updating unit.

The operation of the image tracking device configured as described above will be described. Edge image storage means 301
Stores a series of binary image frames obtained by performing an edge extraction process and a binarization process on a moving image in which a tracking target object appears. The user specifies a rectangular area surrounding the object to be tracked from the first binary image frame I (1) by the tracked object input unit 302. Area dividing means 303
Divides the rectangular area obtained by the tracking object input means 302 into a plurality of small areas not exceeding a certain size (the number of divisions is set to m), and each small area Ri (i = 1,
The coordinates of the two vertices (2,..., M) are stored in the object area storage unit 304.
To be stored.

Further, the object is tracked by repeating the following processing with n being a natural number. The object model storage unit 305 stores the n-th binary image frame I existing in the small region Ri stored in the object region storage unit 304.
A set of coordinates of the edge point (n) is stored as the object model Mi. The object model enlargement / reduction unit 306 enlarges (reduces) the coordinate value of the object model around the upper left vertex of each small area. For example, the enlargement ratio Sx in the x direction and the enlargement ratio Sy in the y direction are independently set as {0.9, 1.0, 1.1} to generate a total of nine types of object models Mi [Sx, Sy].

In the partial Hausdorff distance calculating means 307,
A partial Hausdorff distance defined by Expression (1), which is an evaluation function for obtaining a destination and an enlargement / reduction ratio in the (n + 1) th image frame I (n + 1) of the object, is calculated. For each object model Mi [Sx, Sy], a partial Hausdorff distance hK (Mi [Sx, Sy], I (n) to the (n + 1) th image frame I (n + 1) at all positions within the search range.
+1)).

However, since the unit vector of the distance vector between the two points is different due to the enlargement / reduction operation, it is inconvenient as it is. Therefore, the distance normalization means 308 normalizes the distance. The normalization of the distance is performed by dividing the x and y components of the distance vector between the two points by Sx and Sy, respectively. Partial Hausdorff distance calculation means 30
7, a partial Ha based on the normalized distance vector
Calculate usdorff distance.

The object movement destination estimating means 309 calculates the movement of the object model Mi in the (n + 1) th image frame I (n + 1) based on the position giving the minimum value and the enlargement / reduction ratio among the partial Hausdorff distances thus obtained. The data is output to the object area updating unit 310 as the destination and the enlargement / reduction ratio.

The object area updating means 310 calculates the (n + 1) th image frame I obtained from the object moving destination calculating means 309.
The content of the object area storage unit 304 is updated based on the destination and the enlargement / reduction ratio of the object model Mi in (n + 1).

As described above, according to the present embodiment, the area dividing means 303 is provided to divide the area to be tracked into a plurality of small areas so that even if the object to be tracked is non-rigid, By considering each rigid body as a set of a plurality of rigid bodies, tracking failure can be prevented. In addition, an object model scaling unit 306 is provided to compare the template obtained by scaling the object model in several stages with the next image frame, thereby obtaining the translation amount and the scaling ratio of the target object. Even when the size changes, tracking failure can be prevented. At this time, the distance normalizing means 308 eliminates a change in distance due to the scaling operation, and can calculate a correct destination.

(Embodiment 4) FIG. 4 is a block diagram of an image tracking apparatus according to a fourth embodiment of the present invention. 4, reference numeral 401 denotes an edge image storage means, 402 denotes a tracking object input means, 403 denotes an area dividing means, 404 denotes an object area storage means, 405 denotes an object model storage means, 406 denotes an object model enlargement / reduction means, and 407 denotes a partial Hausdorff. Distance calculating means, 408 is distance normalizing means, 409 is an object moving destination calculating means, and 410 is an object area updating means. The configuration up to this point is the same as the configuration from 301 to 310 shown in FIG. Reference numeral 411 denotes a low-resolution image storage unit, 412 denotes a color histogram calculation unit, and 413 denotes a search range limiting unit, which are different from the third embodiment shown in FIG.

The operation of the image tracking apparatus configured as described above will be described focusing on differences from the third embodiment shown in FIG. The low-resolution image storage unit 411 stores a low-resolution image frame having a reduced spatial resolution with respect to a series of color image frames before performing the edge extraction processing and the binarization processing. As in the third embodiment, the divided small area Ri is stored in the object area storage unit 404.
Is stored. Color histogram calculation means 412
Represents each small region Ri stored in the object region storage means 404.
, The color histogram of the nth low-resolution image frame L (n), and n + 1 at all positions within the search range
The color histogram of the low-resolution image frame L (n + 1) is calculated. The search range limiting unit 413 obtains an L (n + 1) region having a color histogram closest to the Ri color histogram in L (n) from the output of the color histogram calculating unit 412. As a result, the destination of the object area can be roughly obtained. part
The Hausdorff distance calculating means 407 is a search range limiting means 4
The partial Hausdorff distance is calculated limited to the vicinity of the rough object movement destination obtained in step S13.

As described above, according to this embodiment, the low-resolution image storing means 411, the color histogram calculating means 412, and the search range limiting means 413 are provided.
Then, the color histogram of the object region of the n-th color image frame whose spatial resolution is stored and the color histogram of the (n + 1) -th certain region are calculated. A region having a close color histogram is obtained as a rough destination of the object, and the partial Hausdorff distance calculating means 407 obtains a partial Hausdorff distance by limiting the region to the approximate destination, thereby reducing the amount of calculation and speeding up the processing. Can be.

This embodiment can be implemented not only in combination with the third embodiment shown in FIG. 3 but also in combination with the first or second embodiment shown in FIG. 1 or FIG. it can.

[0042]

As described above, according to the present invention, the object model enlarging / reducing means is provided, and a template in which the object model is enlarged or reduced in several steps is prepared to obtain not only the translation amount of the object but also the enlarging / reducing ratio. , Preventing the failure of tracking even when the size of the target object changes, and providing a distance normalizing means,
By eliminating the change in the partial Hausdorff distance due to scaling, an excellent image tracking device capable of accurate tracking can be realized.

Further, according to the present invention, even if the object to be tracked is a non-rigid body, this is regarded as a set of a plurality of rigid bodies by dividing the area to be tracked into a plurality of small areas by providing the area to be tracked. By tracking the rigid body, it is possible to realize an excellent image tracking device that can prevent tracking failure.

Also, the present invention provides the area dividing means, the object model enlarging / reducing means and the distance normalizing means, so that the object to be tracked can be non-rigid or the size of the object can be changed. In addition, it is possible to realize an excellent image tracking device that not only can prevent tracking failure, but also enables accurate tracking.

Further, according to the present invention, the color histogram calculating means and the low-resolution image storing means are used to reduce the resolution.
For the nth and (n + 1) th color image frames, the color histogram calculating means calculates a color histogram of a certain region, and determines a region having a color histogram closest to the color histogram of the object region as a rough destination of the object. By obtaining the partial Hausdorff distance by the partial Hausdorff distance calculation means limited to the periphery, it is possible to realize an excellent image tracking device capable of reducing the amount of calculation and increasing the processing speed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an image tracking device according to a first embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of an image tracking device according to a second embodiment of the present invention.

FIG. 3 is a block diagram illustrating a configuration of an image tracking device according to a third embodiment of the present invention.

FIG. 4 is a block diagram showing a configuration of an image tracking device according to a fourth embodiment of the present invention.

FIG. 5 is a block diagram showing a configuration of a conventional image tracking device.

[Explanation of symbols]

101, 201, 301, 401, 501 Edge image storage means 102, 202, 302, 402, 502 Tracking object input means 203, 303, 403 Area division means 103, 204, 304, 404, 503 Object area storage means 104, 205 , 305, 405, 504 Object model storage means 105, 306, 406 Object model scaling means 106, 206, 307, 407, 505 Partial Hausdo
rff distance calculating means 107, 308, 408 distance normalizing means 108, 207, 309, 409, 506 object moving destination calculating means 109, 208, 310, 410, 507 object area updating means

Claims (4)

[Claims] 1. An edge image storage means for storing a series of binary image frames subjected to edge extraction processing, and a first image frame taken out of the edge image storage means to specify a region including a tracking target object. And for the nth image frame, if n is 1, store the area obtained by the tracking object input means, and n is 2
In the case of the above natural numbers, an object area storing means for storing an area obtained by the object area updating means, and an object model storing means for storing image information in the area stored in the object area storing means as an object model. Object model scaling means for scaling the object model stored in the object model storage means, and a part Hausdorff of each object model obtained by the object model scaling means and a certain area of the (n + 1) th image frame A partial Hausdorff distance calculating means for calculating the distance, a distance normalizing means for normalizing the partial Hausdorff distance obtained by the partial Hausdorff distance calculating means by a scaling factor, and a distance calculated by the distance normalizing means being the minimum. Object moving destination calculating means for calculating a region of the (n + 1) th image frame taking a value, and the (n + 1) th image frame obtained by the moving object calculating means. An image tracking apparatus comprising: an object region updating unit that outputs a region of a game to the object region storage unit. 2. An edge image storage means for storing a series of binary image frames subjected to an edge extraction processing, and a first image frame from the edge image storage means for specifying a region including a tracking target object. Tracking object input means, area dividing means for dividing the area obtained by the tracking object input means into a plurality of small areas, and for the n-th image frame, when n is 1, the n is obtained by the area dividing means. Object area storing means for storing the area obtained by the object area updating means when n is a natural number of 2 or more;
Object model storage means for storing the image information in the area stored in the object area storage means as an object model,
Each object model obtained by the object model storage means and n +
A Hausdorff distance calculating means for calculating a Hausdorff distance from a certain region of the first image frame;
Hausdorff object moving destination calculating means for calculating an area of the (n + 1) th image frame in which the distance calculated by the distance calculating means has a minimum value, and n + obtained by the moving object calculating means
An image tracking apparatus comprising: an object area updating unit that outputs an area of a first image frame to the object area storage unit. 3. An edge image storage means for storing a series of binary image frames subjected to edge extraction processing, and a first image frame is taken out from the edge image storage means to specify a region including a tracking target object. Tracking object input means, area dividing means for dividing the area obtained by the tracking object input means into a plurality of small areas, and for the n-th image frame, when n is 1, the n is obtained by the area dividing means. Object area storing means for storing the area obtained by the object area updating means when n is a natural number of 2 or more;
Object model storage means for storing the image information in the area stored in the object area storage means as an object model,
Object model scaling means for scaling the object model stored in the object model storage means, and a partial Hausdorff distance between each object model obtained by the object model scaling means and a region of the (n + 1) th image frame. Hausdorff distance calculating means for calculating
f The distance Hausdorff obtained by the distance calculating means The distance normalizing means for normalizing the distance by the scaling factor, and the area of the (n + 1) th image frame in which the distance calculated by the distance normalizing means takes the minimum value is calculated. An image tracking apparatus comprising: an object moving destination calculating unit; and an object region updating unit that outputs an (n + 1) -th image frame region obtained by the object moving destination calculating unit to the object region storing unit. 4. A low-resolution image storage unit for storing color image data with reduced resolution for a series of color image frames before performing an edge extraction process and a binarization process;
calculating a color histogram in the region stored in the object region storage means for the nth low resolution image frame;
Compare the color histogram calculation means for calculating the color histogram of an area within a certain range from the area stored in the object area storage means with respect to the (n + 1) th low resolution image frame, and the color histogram obtained by the color histogram calculation means Search range limiting means for roughly calculating the object destination by performing
4. The image tracking apparatus according to claim 1, wherein the distance calculating means calculates a partial Hausdorff distance limited to the search range output from the object moving range limiting means.