JP6062483B2 - Digital camera - Google Patents

Description

  The present invention relates to a digital camera.

  2. Description of the Related Art In recent optical camera-equipped devices, an object such as a face image registered in advance is tracked and the position information is output, so that it is used for applications such as autofocus.

  Patent Document 1 (Japanese Patent Laid-Open No. 2008-287648) and Patent Document 2 (Japanese Patent Laid-Open No. 2001-243476) are technologies that enable object tracking processing when the size of an object changes due to a change in camera zoom. Is disclosed.

  In Patent Document 1, in order to track an object whose size changes, matching is performed between a plurality of template images obtained by enlarging / reducing a basic template image and changing the magnification, and captured images.

  In Patent Document 2, a template image corresponding to a zoom magnification is selected from a plurality of template images obtained by enlarging or reducing a basic template image, and matching with a captured image is performed.

JP 2008-287648 A JP 2001-243476 A

  However, Patent Document 1 has a problem that matching with a plurality of template images is required in order to cope with object tracking whose size changes, and processing load increases.

  In Patent Document 2, since matching with a camera image is performed using a single template image corresponding to the zoom magnification, the speed can be increased as compared to Patent Document 1, but template images having a plurality of magnifications are created and stored. It is necessary to keep. In addition, when a captured image is enlarged by zooming, matching is performed using a template image with an increased number of pixels, so matching takes time.

  Therefore, an object of the present invention is to provide a digital camera that can track a target object at high speed and with high accuracy even if the size of the target object changes due to zoom processing.

  An object tracking method according to an embodiment of the present invention is an object tracking method for tracking an object included in a frame image, in which a storage unit stores a template image of an object to be tracked, and an image input unit includes a time series. When the zoom region of the current frame image is different from the zoom factor of the previous frame image, the search region setting unit determines that the object region in the previous frame image and the previous frame image Based on the center coordinates of the zoom, the ratio of the zoom magnification of the previous frame image to the zoom magnification of the current frame image, the search area of the object in the current frame image is set, and the normalization unit sets the current frame image Normalize the image of the search area of the object contained in the And from the image of the normalized search area, characterized by searching for objects area similar to the template image.

  According to an embodiment of the present invention, even if the size of the target object changes due to zoom processing, the target object can be tracked at high speed and with high accuracy.

It is a block diagram which shows the structural example of the digital camera in embodiment of this invention. It is a figure showing the structure of the object tracking device of 1st Embodiment. It is a figure for demonstrating the search area | region of the present frame image when the present frame image has the same zoom magnification as the previous frame image. It is a figure for demonstrating the search area | region of the present frame image when the present frame image differs in zoom magnification from the previous frame image. It is a figure for demonstrating the example of normalization. It is a flowchart showing the operation | movement procedure of the object tracking process of 1st Embodiment. It is a figure showing the structure of the object tracking device of 2nd Embodiment. (A) is a figure showing the i-th frame image. (B) is a diagram showing the (i + 1) -th frame image. (A) is a figure showing the (i + 1) th frame image. (B) is a diagram showing the (i + 2) -th frame image. It is a flowchart showing the operation | movement procedure of the object tracking process of 2nd Embodiment. It is a flowchart showing the procedure of the return process of step S402 of FIG.

Embodiments of the present invention will be described below with reference to the drawings.
[First Embodiment]
(Digital still camera)
FIG. 1 is a block diagram illustrating a configuration example of a digital camera according to an embodiment of the present invention. In the embodiment of the present invention, a face is described as an example of an object to be detected and tracked. However, the method of the embodiment of the present invention is not limited to a face, and the detection and tracking of other objects are also performed. Can be used.

  Referring to FIG. 1, this digital camera includes a semiconductor device, which includes a camera interface (Interface) 1001, a CPU (Central Processing Unit) 1002, an SDRAM (Synchronous Dynamic Random Access Memory) 1003, and the like. ROM (Read Only Memory) 1004, user operation unit I / F 1005, LCD I / F 1007, card I / F 1008, object detection device 1006, and object tracking device 1009, which are connected via a bus 1010 Has been. The object detection device 1006 and the object tracking device 1009 constitute an image processing device 1011.

  The camera I / F 1001 is connected to a camera sensor, receives an image captured by the camera sensor, and writes the image to an SD memory card (not shown) connected via the SDRAM 1003 or the card I / F 1008.

  The CPU 1002 controls the entire system by executing programs stored in the SDRAM 1003 and the ROM 1004. The CPU 1002 sends information representing the position of the face (object) stored in the SDRAM 1003 to the imaging system of the camera through the camera I / F 1001 for focusing. Further, the CPU 1002 sends a graphic (such as a rectangle) representing the position of the face to the LCD I / F 1007 from the information representing the position of the face (object) sent from the object detection device 1006 and the object tracking device 1009.

  The ROM 1004 stores data used for object detection and object tracking.

  The user operation unit I / F 1005 is connected to a shutter button (not shown) and the like, and when the shutter button is pressed, notifies the CPU 1002 to that effect by an interrupt or the like. Upon receiving an interrupt request, the CPU 1002 controls the camera I / F 1001 and stores an image captured by the camera sensor in the SDRAM 1003 or the like. When the user operates a zoom switch (not shown) or the like to change the zoom magnification, the zoom magnification is sent to the CPU 1002 via the user operation unit I / F 1005.

  The LCD I / F 1007 is connected to an LCD panel (not shown) and controls display on the LCD. An image taken by the camera sensor is displayed on the LCD as a preview. The LCD I / F 1007 superimposes and displays a graphic representing the position of the face (object) sent from the CPU 1002 on the captured image.

  A card I / F 1008 is connected to an external recording medium such as an SD memory card, and reads / writes data from / to the recording medium.

  The object detection device 1006 performs face (object) detection processing from an image photographed by the camera sensor and input from the camera I / F 1001, and stores information representing the position of the face (object) in the SDRAM 1003. The object detection apparatus 1006 registers the detected object area as a template.

  The object tracking device 1009 performs tracking processing of the face (object) detected by the object detection device 1006 in the image captured by the camera sensor and input from the camera I / F 1001, and information representing the position of the face (object) Is stored in the SDRAM 1003. The object tracking device 1009 tracks an object by searching a captured image for a region similar to the template obtained by the object detection device. The object tracking device 1009 receives the search range magnification K, zoom magnification, and zoom center coordinates sent from the CPU 1002. The object tracking device 1009 receives a template image and an object region of an object to be tracked from the object detection device 1006.

(Constitution)
FIG. 2 is a diagram illustrating the configuration of the object tracking apparatus according to the first embodiment.

  Referring to FIG. 2, this object tracking apparatus 1009 includes an image input unit 2, a parameter input unit 3, an object region input unit 4, an object region setting unit 6, a template input unit 5, and a template registration unit 7. A template storage unit 8, a search region setting unit 9, a normalization unit 10, and a matching unit 11.

  The image input unit 2 receives a frame image for searching for an object to be tracked from the camera I / F at a predetermined frame rate. When the subject is zoomed (enlarged or reduced) by the user, the image input unit 2 receives the zoomed frame image.

  Parameter input unit 3 receives an input of search range magnification K from CPU 1002. Further, when the subject is zoomed by the user and the zoom magnification of the current frame image is changed from the zoom magnification of the previous frame image, the parameter input unit 3 receives the zoom magnification, the zoom center coordinates, and the like from the CPU 1002. Receive.

  The template input unit 5 receives a template image of an object to be tracked output from the object detection device.

The template storage unit 8 stores a template image of an object to be tracked.
The template registration unit 7 stores the template image input to the template input unit 5 in the template storage unit 8 as an initial template. The template registration unit 7 updates the template image in the template storage unit 8 with the image of the object area specified by the matching unit 11.

  The object area input unit 4 receives information representing an object area to be tracked from the object detection device.

  The object area setting unit 6 sets the object area input to the object area input unit 4 as an initial object area. The object area setting unit 6 sets the object area searched by the matching unit 11 as being similar to the template image as a new object area.

  When the zoom magnification of the current frame image is the same as the zoom magnification of the previous frame image, the search area setting unit 9 includes the object area in the previous frame image, and the search area magnification K is larger than the object area. A region (reference region) that is as large as possible is set as the object region in the current frame image.

  When the zoom magnification of the current frame image is different from the zoom magnification of the previous frame image, the search area setting unit 9 obtains the above-described reference area, and uses this reference area as the zoom center coordinate in the previous frame image. Based on the ratio between the zoom magnification of the previous frame image and the zoom magnification of the current frame image, the zoomed (enlarged or reduced) region is set as the object search region in the current frame image.

  The normalization unit 10 normalizes the image of the object search area in the current frame image to a fixed size to generate a normalized search image.

  When the matching unit 11 searches the normalized search image for a region similar to the template image stored in the template storage unit 8 and finds a portion where the similarity is equal to or greater than a predetermined value, Is specified as the object area of the current frame, and information representing the area is output to the CPU 1002. Here, as the similarity, for example, the reciprocal of the absolute difference between the pixel values of two images or a cross-correlation can be used.

  The matching unit 11 instructs the object detection device to detect the object again when a part having a similarity greater than or equal to a predetermined value cannot be found.

(Example of search area)
FIG. 3 is a diagram for explaining a search area for the current frame image when the current frame image has the same zoom magnification as the previous frame image.

FIG. 3A shows a frame image at the previous time point.
In the frame image 201, the upper left coordinate is O (0, 0), the upper right coordinate is A (DX, 0), the lower left coordinate is B (0, DY), and the lower right coordinate is C (DX, DY). To do.

The position of the object area 202 detected in the previous frame is as follows.
The center coordinate is M (X1, Y1), the upper left coordinate is D (X1-ax, Y1-ay), the upper right coordinate is E (X1 + ax, Y1-ay), and the lower left coordinate is F (X1-ax, Y1 + ay). The lower right coordinate is G (X1 + ax, Y1 + ay).

FIG. 3B shows a current frame image.
In the frame image 203, as in the previous time point, the upper left coordinate is O (0, 0), the upper right coordinate is A (DX, 0), the lower left coordinate is B (0, DY), and the lower right coordinate is C. (DX, DY). The position of the search area 204 in the current frame image is as follows.

  The central coordinate is M (X1, Y1), the upper left coordinate is H (X1-ax ', Y1-ay'), the upper right coordinate is I (X1 + ax ', Y1-ay'), and the lower left coordinate is J ( X1-ax ′, Y1 + ay ′), and the lower right coordinate is K (X1 + ax ′, Y1 + ay ′). Here, ax ′ = K × ax and ay ′ = K × ay. K is a search range magnification.

  Therefore, if the current frame image has the same zoom magnification as the previous frame image, the current frame search area is set to a rectangular area (reference area) 204 with H, I, J, and K as vertices. Will be. The rectangular area 204 includes the object area 202 in the previous frame image, and is an area K times larger than the object area.

  FIG. 4 is a diagram for explaining a search area for the current frame image when the current frame image has a zoom magnification different from that of the previous frame image.

FIG. 4A shows a frame image at the previous time point.
In the frame image 201, the upper left coordinate is O (0, 0), the upper right coordinate is A (DX, 0), the lower left coordinate is B (0, DY), and the lower right coordinate is C (DX, DY). To do.

The position of the object area 202 detected in the previous frame is as follows.
The center coordinate is M (X1, Y1), the upper left coordinate is D (X1-ax, Y1-ay), the upper right coordinate is E (X1 + ax, Y1-ay), and the lower left coordinate is F (X1-ax, Y1 + ay). The lower right coordinate is G (X1 + ax, Y1 + ay).

  Here, as a result of zooming the camera with the position L (X0, Y0) as the center coordinate of zoom and the ratio of the zoom magnification of the previous frame image to the zoom magnification of the current frame image as ZR, It is assumed that the rectangular area 205 having P, Q, R, and S as vertices in the frame image is enlarged to be the entire area of the current frame image. Here, the position of P is (X0−EX / 2, Y0−EY / 2), the position of Q is (X0 + EX / 2, Y0−EY / 2), and the position of R is (X0−EX / 2, Y0 + EY / 2) The position of S is (X0 + EX / 2, Y0 + EY / 2). The ratio of zoom magnification ZR = DX / EX = DY / EY. For example, when the zoom magnification of the camera at the previous time point is 2 and the zoom magnification of the current camera is 4, the zoom magnification ratio ZR is 2 (= 4/2). When the zoom magnification is increased, that is, when the zoom magnification is increased, the zoom magnification ratio ZR is larger than 1. When the zoom magnification is decreased, that is, when the zoom magnification is reduced, the zoom magnification ratio ZR is smaller than 1.

FIG. 4B shows a current frame image.
In the frame image 206, as in the previous time point, the upper left coordinate is O (0, 0), the upper right coordinate is A (DX, 0), the lower left coordinate is B (0, DY), and the lower right coordinate is C. (DX, DY).

  Further, the zoom center L at the previous time point moves to a position L ′ representing the center of the camera image at the center of the frame image 206.

  The object area 202 detected in the previous frame is moved and zoomed to a rectangular area 207 whose center is M ′ and whose vertices are D ′, E ′, F ′, and G ′. Here, the position of M ′ is (XP, YP). However, XP = DX / 2 + ZR × (X1−X0) and YP = DY / 2 + ZR × (Y1−Y0). The position of D ′ is (XP−ax × ZR, YP−ay × ZR), the position of E ′ is (XP + ax × ZR, YP−ay × ZR), and the position of F ′ is (XP−ax × ZR, YP + ay ×). The position of ZR) and G ′ is (XP + ax × ZR, YP + ay × ZR).

  The position of the search area 208 is set to a rectangular area 208 whose center is M ′ and whose vertices are H ′, I ′, J ′, and K ′.

  Here, the position of M ′ is (XP, YP), the position of H ′ is (XP−ax ′ × ZR, YP−ay ′ × ZR), and the position of I ′ is (XP + ax ′ × ZR, YP−ay ′ × ZR), the position of J ′ is (XP−ax ′ × ZR, YP + ay ′ × ZR), and the position of K ′ is (XP + ax ′ × ZR, YP + ay ′ × ZR). Here, ax ′ = K × ax and ay ′ = K × ay. K is a search range magnification.

  Therefore, when the current frame image has a zoom magnification different from that of the previous frame image, the current frame search area is set to a rectangular area 208 having vertices at H ′, I ′, J ′, and K ′. Will be.

  This rectangular area 208 includes an area 207 obtained by zooming the object area 202 in the previous frame image, and is an area K times larger than the object area 207. Further, the rectangular area 208 is obtained by changing the current frame image search area (the area 204 in FIG. 3B) when the current frame image has the same zoom magnification as the previous frame image in the previous frame image. This is a zoomed (enlarged or reduced) area based on the zoom center coordinate L and the zoom magnification ratio ZR.

(Normalization example)
FIG. 5 is a diagram for explaining an example of normalization.

  In FIG. 5, the normalization unit 10 extracts an image of the search area 303 from the frame image. As shown in FIG. 4B, when the current frame image has a zoom magnification different from that of the previous frame image, the number of pixels in the horizontal direction of the extracted image is 2 × ax ′ × ZR, The number of pixels is 2 × ay ′ × ZR.

  The normalization unit 10 generates a normalized search image 304 by enlarging or reducing the extracted image to a fixed size. As shown in FIG. 5, in the fixed size, the number of pixels in the horizontal direction is XS, and the number of pixels in the vertical direction is YS. The normalizing unit 10 performs a process such as interpolation when enlarging, and performs a process such as thinning out when reducing.

(Operation)
FIG. 6 is a flowchart showing the operation procedure of the object tracking process of the first embodiment.

  Referring to FIG. 6, first, object region input unit 4 receives information representing an object region to be tracked from an object detection device. The object area setting unit 6 sets the object area output from the object detection apparatus as an initial object area (step S101).

  Next, the template registration unit 7 stores the image of the target object output from the object detection device in the template storage unit 8 as an initial template (step S102).

  Next, the image input unit 2 receives the current frame image from the camera I / F (step S103).

  Next, when the parameter input unit 3 receives the zoom magnification and the zoom center coordinates from the CPU 1002 (YES in step S104), the search area setting unit 9 sets the current frame image zoom magnification to the previous time point. Since it is different from the zoom magnification of the frame image, an object area in the previous frame image is obtained, and an area (reference area) larger than the object area by a predetermined multiple is obtained, and the reference area is further determined in the previous frame image. Based on the zoom center coordinates and the ratio of the zoom magnification of the previous frame image to the zoom magnification of the current frame image, the zoomed (enlarged or reduced) area is set as the object search area in the current frame image ( Step S105).

  On the other hand, when the parameter input unit 3 does not receive the zoom magnification and the zoom center coordinates from the CPU 1002 (NO in step S104), the search area setting unit 9 determines that the current frame image has a zoom magnification of the previous time point. Since it is the same as the zoom magnification of the image, a reference area that includes the object area in the previous frame image and is larger than the object area by a predetermined multiple is set as the object area in the current frame image (step S106).

  Next, the normalization unit 10 normalizes the image of the object search area in the current frame image to a fixed size to generate a normalized search image (step S107).

  Next, the matching unit 11 searches the normalized search image for a region similar to the template image stored in the template storage unit 8 and searches for a portion where the similarity is equal to or greater than a predetermined value.

  When the matching unit 11 finds a part where the similarity is equal to or greater than a predetermined value, that is, when the matching is successful (YES in step S108), the matching unit 11 sets the found part as the object area of the current frame. The information specifying the area is output to the CPU 1002 (step S109).

  The object area setting unit 6 sets the object area specified by the matching unit 11 as a new object area (step S110).

  The template registration unit 7 updates the template image in the template storage unit 8 with the image of the object area specified by the matching unit 11 (step S111).

  Thereafter, the process returns to step S103, and the processing from the input of the next frame image is repeated.

  If the matching unit 11 has not found a portion where the degree of similarity is equal to or greater than the predetermined value, that is, if the matching has failed (NO in step S108), the matching unit 11 returns the object to the object detection device again. The detection is instructed (step S112).

(effect)
As described above, in the present embodiment, a search area that takes into account changes in the size and position of the target object even if the magnification of the frame image input in the zoom process changes using the zoom magnification and the center coordinates of the zoom. Can be set. This eliminates the need for matching with a plurality of template images.

  In the present embodiment, since the image of the search area is reduced to a normalized image of a certain size and template matching is performed, processing is performed by enlarging the template image and performing matching processing as in the past. It can prevent the time from becoming longer.

[Second Embodiment]
While tracking the target object, the object to be tracked may be out of the frame image due to enlargement. Conventionally, since the target object once out of the frame does not exist in the frame image, the position information is not saved. After that, even when the target object is framed into the frame image again, there is no information on the reference position for starting the tracking process (that is, the object detection area at the center of the search area), so the object can be tracked. Can not. Therefore, conventionally, when an object to be tracked is out of frame, there is no choice but to stop the tracking process and perform the object detection process again.

FIG. 7 is a diagram illustrating the configuration of the object tracking device according to the second embodiment.
The object tracking device 51 of FIG. 7 is different from the object tracking device 1009 of FIG. 2 as follows.

  The holding unit 28 holds the set object search area when a part or all of the object search area set by the search area setting unit 29 is not included in the current frame image.

  FIG. 8A shows the i-th frame image. FIG. 8B shows the (i + 1) th frame image.

  In the i-th frame image 601, it is assumed that an enlarged image with the zoom center as L <b> 1 and the zoom magnification ratio as ZR <b> 1 is the (i + 1) -th frame image 603. Here, ZR1> 1.

  As shown in FIG. 8B, when the search area 605 centering on M ′ and having vertices H ′, I ′, J ′, K ′ protrudes outside the frame image 603, the holding unit 28 , M ′, H ′, I ′, J ′, K ′. The position of M ′ is (XP, YP), the position of H ′ is (XP-ax ′ × ZR1, YP-ay ′ × ZR1), and the position of I ′ is (XP + ax ′ × ZR1, YP-ay). ′ × ZR1), the position of J ′ is (XP−ax ′ × ZR1, YP + ay ′ × ZR1), and the position of K ′ is (XP + ax ′ × ZR1, YP + ay ′ × ZR1). However, XP = DX / 2 + ZR1 × (X1−X0) and YP = DY / 2 + ZR1 × (Y1−Y0).

  If the zoom magnification of the current frame image is different from the zoom magnification of the previous frame image, the search area setting unit 29 may hold the object search area in the previous frame image in the holding unit 28. Zoom (enlarge or reduce) based on the search area of the held object, the center coordinates of the zoom in the previous frame image, and the ratio of the zoom factor of the previous frame image to the zoom factor of the current frame image ) Is set as an object search area in the current frame image.

  FIG. 9A shows the (i + 1) th frame image. FIG. 9B shows the (i + 2) th frame image.

  In the (i + 1) th frame image 603, it is assumed that an image reduced by setting the zoom center to L2 and the zoom magnification ratio to ZR2 is the (i + 2) th frame image 610. Here, ZR2 <1.

  The search area setting unit 29 compares the positions of M ′, H ′, I ′, J ′, and K ′ held by the holding unit 28 with the zoom center coordinates L2 (X0 ′, Y0 ′) and the zoom magnification. Using ZR2, a search area 612 of an object having M ″ as the center and H ″, I ″, J ″, and K ″ as vertices is obtained.

  The position of M ″ is (XP ′, YP ′), the position of H ″ is (XP′−ax ′ × ZR1 × ZR2, YP′−ay ′ × ZR1 × ZR2), and the position of I ″ is ( XP ′ + ax ′ × ZR1 × ZR2, YP′−ay ′ × ZR1 × ZR2), and the position of J ″ is (XP′−ax ′ × ZR1 × ZR2, YP ′ + ay ′ × ZR1 × ZR2), The position of K ″ is (XP ′ + ax ′ × ZR1 × ZR2, YP ′ + ay ′ × ZR1 × ZR2), where XP ′ = DX / 2 + ZR2 × (XP−X0 ′), YP ′ = DY / 2 + ZR2 × (YP-Y0 ').

(Operation)
FIG. 10 is a flowchart illustrating an operation procedure of object tracking processing according to the second embodiment.

  Referring to FIG. 10, first, object region input unit 4 receives information representing an object region to be tracked from the object detection device. The object area setting unit 6 sets the object area output from the object detection apparatus as an initial object area (step S101).

  Next, the template registration unit 7 stores the image of the target object output from the object detection device in the template storage unit 8 as an initial template (step S102).

  Next, the image input unit 2 receives the current frame image from the camera I / F (step S103).

  Next, when the parameter input unit 3 receives the zoom magnification and the zoom center coordinates from the CPU 1002 (YES in step S104), the search area setting unit 29 sets the current frame image zoom magnification to the previous time point. Since it is different from the zoom magnification of the frame image, an object area in the previous frame image is obtained, and an area (reference area) larger than the object area by a predetermined multiple is obtained, and the reference area is further determined in the previous frame image. Based on the zoom center coordinates and the ratio of the zoom magnification of the previous frame image to the zoom magnification of the current frame image, the zoomed (enlarged or reduced) area is set as the object search area in the current frame image ( Step S105).

  When a part or all of the set search area is outside the current frame image (YES in step S401), a return process described later is performed (step S402).

  On the other hand, when the parameter input unit 3 does not receive the zoom magnification and the zoom center coordinates from the CPU 1002 (NO in step S104), the search area setting unit 29 determines that the current frame image has a zoom magnification of the previous time point. Since it is the same as the zoom magnification of the image, a reference area that includes the object area in the previous frame image and is larger than the object area by a predetermined multiple is set as the object area in the current frame image (step S106).

  Next, the normalization unit 10 normalizes the image of the object search area in the current frame image to a fixed size to generate a normalized search image (step S107).

  Next, the matching unit 11 searches the normalized search image for a region similar to the template image stored in the template storage unit 8 and searches for a portion where the similarity is equal to or greater than a predetermined value.

  When the matching unit 11 finds a part where the similarity is equal to or greater than a predetermined value, that is, when the matching is successful (YES in step S108), the matching unit 11 sets the found part as the object area of the current frame. The information specifying the area is output to the CPU 1002 (step S109).

  The object area setting unit 6 sets the object area specified by the matching unit 11 as a new object area (step S110).

  The template registration unit 7 updates the template image in the template storage unit 8 with the image of the object area specified by the matching unit 11 (step S111).

  Thereafter, the process returns to step S103, and the processing from the input of the next frame image is repeated.

  If the matching unit 11 has not found a portion where the degree of similarity is equal to or greater than the predetermined value, that is, if the matching has failed (NO in step S108), the matching unit 11 returns the object to the object detection device again. The detection is instructed (step S112).

(Return processing)
FIG. 11 is a flowchart showing the procedure of the return process in step S402 of FIG.

  First, the holding unit 28 holds information representing the set search area (step S501).

  The image input unit 2 receives the next frame image from the camera I / F (step S502).

  When the parameter input unit 3 receives the zoom magnification and the zoom center coordinates from the CPU 1002 (YES in step S503), the search area setting unit 29 determines that the current frame image zoom magnification is that of the previous frame image. Since it is different from the zoom magnification, the object search area in the frame image at the previous time point stored, the center coordinates of the zoom in the frame image at the previous time point, the zoom magnification of the frame image at the previous time point, and the zoom magnification of the current frame image Based on this ratio, the zoomed (enlarged or reduced) area is set as the object search area in the current frame image (step S504).

  If a part or all of the set search area is outside the current frame image (YES in step S505), the processing from step S501 is repeated.

(effect)
As described above, according to the second embodiment, even if the object to be tracked is out of the frame image due to the enlargement, the search area of the object out of the frame is held and the search in the next frame is performed. Since it is used for setting an area, it is possible to continue tracking an object.

  If YES in step S505 in FIG. 11 of the present embodiment, the process always returns to step S501. However, if the return process in FIG. 11 continues for a predetermined time or more, or the loop of steps S501 to S505 continues. When continuing for a predetermined number of times or more, the return process may be terminated and the object detection unit may be caused to perform the object detection process again.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

  2 image input unit, 3 parameter input unit, 4 object region input unit, 5 template input unit, 6 object region setting unit, 7 template registration unit, 8 template storage unit, 9, 29 search region setting unit, 10 normalization unit, DESCRIPTION OF SYMBOLS 11 Matching part, 28 Holding part, 1000 Digital camera, 1001 Camera I / F, 1002 CPU, 1003 SDRAM, 1004 ROM, 1005 User operation part I / F, 1006 Object detection apparatus, 1007 LCD I / F, 1008 Card I / F 51,1009 Object tracking device, 1011 Image processing device.

Claims (5)

An image acquisition unit for acquiring frame images in time series;
An operation unit in which a user gives an operation instruction;
An image processing unit for detecting and tracking an object from the frame image;
A control unit that controls the image processing unit in response to an operation instruction from a user ,
The image processing unit
An object detection unit that detects a tracking target object from the first frame image and sets a first search region as a rectangular region including the tracking target object;
An object search unit that sets a second search region for tracking the tracking target object from a group of frame images subsequent to the first frame image;
The second search area is set based on the first search area, the zoom center coordinates of the first frame image, and the ratio between the zoom magnification of the first frame image and each zoom magnification of the frame image group. A digital camera characterized by being made. The object search unit further includes:
The digital camera according to claim 1, further comprising a normalization unit that enlarges or reduces the second search area to a fixed size. When the user has not instructed zooming from the operation unit,
Prior Symbol object searching unit, searches according to area ratio, according to claim 1 or 2 digital camera according to, characterized in that by enlarging the first search area and the second search area. When the user gives a zoom instruction from the operation unit,
The object search unit according to FIG. Over arm magnification and's over arm center coordinates, the first to increase or decrease the search area, further characterized in that enlarged in accordance with the search area ratio and the second search area The digital camera according to claim 3. The second case of the the search area does not contain some or all of the tracked object, according to claim 3 or 4 digital camera, wherein the to continue the search from the frame image group.

Images