JP2010267257A - Apparatus, method and program for processing image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing apparatus, an image processing method and an image processing program, which can obtain approximative front image which can be used for identification processing of a target object such as a vehicle and a figure photographed at various angles, by simple processing which can be achieved even with a semiconductor without depending on a computer with high throughput and which can be incorporated also in a camera. <P>SOLUTION: The image processing apparatus includes: an target object extraction part 101 which detects the target object such as the vehicle and the figure from an image photographed with the camera, etc. installed on a traffic passage, etc. to extract a region of the target object; an image correction part 102 which corrects distortion of the image, due to an installed position of the camera and inclination of the target object; a feature image generation part 103 which generates a feature image to the image corrected by the image correction part 102; a center position correction part 104 which detects a center position where correlation of right and left at each height becomes the highest to the target object region in the feature image generated by the feature image generation part 103, and vertically aligns the center position; and a collation region extraction part 105 which determines a region usable for the identification processing in the image obtained by the center position correction part 104. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  INDUSTRIAL APPLICABILITY The present invention is used for a vehicle identification system, an entrance / exit system using face authentication, a search system for selecting an image of a specific object from a database storing images of vehicles and people taken from various angles, and the like. The present invention relates to a suitable image processing apparatus, an image processing method, and an image processing program.

  In recent years, with the widespread use of cameras, a technique for identifying a vehicle or a person photographed with a camera installed on a traffic road or the like has been put into practical use. In the identification technique, a target face area or vehicle area is extracted from an image, and the image of the extracted area is collated with an image registered in a database in advance to determine similarity. Then, based on the determined similarity, a search is performed when selecting a similar image in the registered image, a face authentication is performed when a person is specified from the selected registered face image, and a vehicle type is specified from the selected vehicle image. It is vehicle type identification.

  In the identification technology, when the image of the object is taken at an angle different from the registered image, the identification accuracy is significantly lowered. As a method for solving this problem, for example, in the case of a face image, in Patent Document 1, a plurality of cameras are used, the same face is photographed from a plurality of directions, and a front face is selected from the photographed face images. Select the orientation image and use it for verification. In Patent Document 2, a registered face image and a reference model of a wire frame are used to synthesize and register a pseudo registered face image when viewed from the shooting direction of the camera. In Patent Document 3, face images taken from a plurality of directions are registered in advance to authenticate images other than the front face.

  Further, as a method applicable to a vehicle, in Patent Document 4, a plurality of wire frame models are prepared in advance, and an input image is mapped on each wire frame model assuming various shooting directions. Of the images obtained when the model is directed to the front, the image having the highest left-right symmetry is used as the front image for collation.

JP 2008-304979 A JP 2003-263639 A JP 2002-092601 A JP 2008-217220 A

  However, in the method of photographing a single object at the same time with a plurality of cameras described in Patent Document 1, a dedicated camera is required, and implementation with a normal camera is difficult. In the method of preregistering images taken from a plurality of directions described in Patent Document 2, a plurality of cameras are provided when taking a registered image, as in the case of a method of photographing a single object simultaneously with a plurality of cameras. A device is necessary, or a great amount of work is required by photographing the same object many times from different directions with a single camera. In addition, the number of registered images is several times that required when only the front image is used. This not only requires a large storage medium for the identification device, but also increases the verification processing time.

  In the method using the reference model of the wire frame described in Patent Document 3, from a box type such as a bus or a wagon car, to a shape having irregularities such as a sedan type or a streamline type such as a sports car. In the case of vehicles having various shapes, the reference model cannot be determined and cannot be applied. In the method described in Patent Document 4, the input image is mapped on the assumption of various shooting directions for each wire frame model. Therefore, compared to the method described in Patent Document 3, the processing of the number of models × shooting direction is performed. A computer with high processing capability is essential.

  The present invention has been made in view of such circumstances, and it is possible to realize a vehicle or a person photographed from various angles by a simple process that can be implemented in a semiconductor without being based on a computer with high processing capability and can be incorporated into a camera. An object of the present invention is to provide an image processing apparatus, an image processing method, and an image processing program capable of obtaining an approximate front image that can be used for identification processing of an object such as the above.

  The image processing apparatus of the present invention detects an object such as a vehicle or a person from an image taken by a camera and extracts an object area, and depends on an installation position of the camera and an inclination of the object. An image correction unit that corrects image distortion, and a center position where the left-right correlation at each height is the highest for each object region in the image corrected by the image correction unit is detected. A center position correcting unit that aligns vertically, and a collation region extracting unit that determines a region that can be used for identification processing in an image obtained by the center position correcting unit.

  According to this configuration, an object area such as a vehicle or a person is detected from an image captured by the camera to extract an object area, and image distortion due to the camera installation position and the object inclination is corrected. Then, the center position where the left-right correlation at each height is the highest for the object area in the corrected image is detected, the detected center position is aligned vertically, and the center position is aligned vertically. The area that can be used for identification processing is determined, so it can be realized in semiconductors without relying on computers with high processing capabilities, and it can be incorporated into cameras, such as vehicles and people taken from various angles. An approximate front image that can be used for the object identification process can be obtained.

  The image processing apparatus of the present invention detects an object such as a vehicle or a person from an image taken by a camera and extracts an object area, and depends on an installation position of the camera and an inclination of the object. An image correction unit that corrects image distortion, a feature image generation unit that generates a feature image for the image corrected by the image correction unit, and an object region in the feature image generated by the feature image generation unit In contrast, the center position where the left-right correlation at each height is the highest is detected, and the center position correction unit that aligns the center position vertically, and used for the identification processing in the image obtained by the center position correction unit And a collation area extraction unit that determines a possible area.

  According to this configuration, an object area such as a vehicle or a person is detected from an image captured by the camera to extract an object area, and image distortion due to the camera installation position and the object inclination is corrected. Then, a feature image is generated for the corrected image, and a center position where the left-right correlation at each height is the highest is detected for the object region in the generated feature image. Vehicles taken from various angles with simple processing that can be implemented in semiconductors and can be incorporated into cameras, because it determines areas that can be used for identification processing in aligned images, without relying on computers with high processing capabilities. It is possible to obtain an approximate front image that can be used for identifying an object such as a person or a person.

  In the above configuration, the center position correction unit detects and detects the center position where the left-right correlation at each height is the highest for the object region in the feature image generated by the feature image generation unit. After smoothing the center position, align the center position vertically.

  In the above configuration, the center position correction unit detects the center position where the left-right correlation at each height is the highest for the object region in the feature image generated by the feature image generation unit, The center position is determined for each area in each small area divided horizontally, and the center position of each small area is aligned vertically.

  In the above configuration, when the collation region extraction unit changes the region width of the left and right regions sandwiching the center position within a predetermined range with respect to the object region in the center position correction image generated by the center position correction unit. The change in the correlation value of the left and right regions of the region is examined, and the region width where the correlation value rapidly decreases is set as the width of the collation range.

  In the above configuration, when the collation region extraction unit changes the region width of the left and right regions sandwiching the center position within a predetermined range with respect to the object region in the center position correction image generated by the center position correction unit. When the change in the correlation value of the left and right areas of the image is examined, and the value in the area width with the highest correlation is smaller than the predetermined value, an image in which either the left or right half of the matching image is replaced with the other inverted copy is generated. Is a collation image.

  The image processing method of the present invention includes an object extraction step for detecting an object such as a vehicle or a person from an image photographed by a camera and extracting an object area, an installation position of the camera, and an inclination of the object An image correction step for correcting image distortion due to the image, a feature image generation step for generating a feature image for the image corrected in the image correction step, and an object in the feature image generated in the feature image generation step The center position where the left-right correlation at each height is the highest for each region is detected, and the center position correction step for vertically aligning the center position is used for identification processing in the image obtained by the center position correction step. And a collation region extraction step for determining a possible region.

  According to this method, an object region such as a vehicle or a person is detected from an image photographed by a camera to extract an object area, and image distortion due to the camera installation position and the object inclination is corrected. Then, a feature image is generated for the corrected image, and a center position where the left-right correlation at each height is the highest is detected for the object region in the generated feature image. Vehicles taken from various angles with simple processing that can be implemented in semiconductors and can be incorporated into cameras, because it determines areas that can be used for identification processing in aligned images, without relying on computers with high processing capabilities. It is possible to obtain an approximate front image that can be used for identifying an object such as a person or a person.

  An image processing apparatus according to the present invention includes an image input unit that inputs an image of a processing target, a display unit that displays an image input by the image input unit and displays an instruction of an operator, and an instruction of the operator An instruction unit for instructing a symmetric position and a center position of a processing object in an image input by the image input unit, and an image input by the image input unit for a symmetric position and a center position instructed by the instruction unit. An image deformation unit that deforms the image using the image deformation unit, and an image output unit that outputs an image deformed by the image deformation unit.

  According to this configuration, the operator can freely set the symmetric position and the center position of the processing object in the input image, and the input image can be set using the symmetric position and the center position of the processing object set by the operator. The image deformation is performed.

  An image processing apparatus of the present invention includes an image input unit that inputs an image of a processing object, a display unit that displays an image input by the image input unit and displays an instruction of an operator, A condition recording unit for recording a ready-made value, an instruction unit for instructing a symmetrical position, a center position, and a size index of the processing object in the image input by the image input unit, and an image input by the image input unit The image is deformed using the symmetrical position and the center position indicated by the instruction unit, and further, the image is deformed using the ready-made value recorded in the condition recording unit and the size index specified by the instruction unit. And an image output unit that outputs the image deformed by the image deforming unit.

  According to this configuration, the operator can freely set the symmetric position, the center position, and the size index of the processing object in the input image, and the symmetric position, the center position, and the size of the processing object set by the operator. Image deformation of the input image is performed using the index.

  An image processing apparatus of the present invention includes an image input unit that inputs an image of a processing object, a display unit that displays an image input by the image input unit and displays an instruction of an operator, A condition recording unit for recording a ready-made value, an instruction unit for instructing a symmetric position and a central position, a size index, and an unused portion of the processing object in the image input by the image input unit, and the image input unit The input image is deformed using the symmetric position and the center position specified by the instruction unit, and the ready-made value recorded in the condition recording unit and the instruction unit are specified for the obtained image. An image deforming unit that deforms an image using the size index that has been generated and further supplements a non-use portion instructed by the instruction unit with an image of a symmetric portion, and an image output that outputs an image deformed by the image deforming unit And a section.

  According to this configuration, the operator can freely set the symmetric position and center position of the processing object in the input image, the size index, and the unused portion, and the symmetric position of the processing object set by the operator Image deformation of the input image is performed using the center position, the size index, and the unused portion.

  In the above configuration, the image output unit outputs the unused portion information instructed by the instruction unit together with the pre-replenishment image.

  In the above configuration, when there are a plurality of pairs of the symmetric position and the central position designated by the instruction unit, the image transformation unit deforms the image for each area divided by a line segment connecting the symmetric position and the central position.

  According to an image processing method of the present invention, an image input step for inputting an image of an object to be processed, a display step for displaying an image input in the image input step and displaying an instruction of the operator, and an instruction of the operator An instruction step for instructing the symmetric position and center position of the processing object in the image input in the image input step, and the symmetric position and center position instructed in the instruction step for the image input in the image input step. An image transformation step for transforming the image using the image transformation step, and an image output step for outputting the image transformed in the image transformation step.

  According to this method, the operator can freely set the symmetric position and the center position of the processing object in the input image, and the input image is used by using the symmetric position and the center position of the processing object set by the operator. The image deformation is performed.

  The image processing method of the present invention includes an image input step for inputting an image of a processing object, a display step for displaying an image input in the image input step and an instruction of an operator, A condition recording step for recording a ready-made value, an instruction step for instructing a symmetrical position, a central position and a size index of the processing object in the image input in the image input step, and an image input in the image input step The image deformation step of deforming the image using the symmetrical position and the central position instructed in the instruction step, and further deforming the image by using the ready value recorded in the condition recording step and the size index instructed in the instruction step And an image output step for outputting the image deformed in the image deformation step.

  According to this method, the operator can freely set the symmetric position, the center position, and the size index of the processing object in the input image, and the symmetric position, the center position, and the size of the processing object set by the operator. Image deformation of the input image is performed using the index.

  The image processing method of the present invention includes an image input step for inputting an image of a processing object, a display step for displaying an image input in the image input step and an instruction of an operator, A condition recording step for recording a ready-made value, an instruction step for instructing a symmetric position and a center position, a size index, and an unused portion of the processing object in the image input in the image input step, and an image input step. The input image is deformed using the symmetrical position and the center position specified in the instruction step, and the image obtained by this is instructed in the ready value recorded in the condition recording step and in the instruction step. An image deformation step of deforming the image using the size index, and supplementing the unused portion instructed in the instruction step with an image of a symmetric portion; An image output step of outputting the image deformation by the image deformation step, with a.

  According to this method, the operator can freely set the symmetric position and center position of the processing object in the input image, the size index, and the unused portion, and the symmetric position of the processing object set by the operator Image deformation of the input image is performed using the center position, the size index, and the unused portion.

  In the above method, in the image output step, the unused portion information instructed in the instruction step is output together with the pre-replenishment image.

  In the above method, in the image deformation step, when there are a plurality of pairs of the symmetric position and the central position indicated in the instruction step, the image is deformed for each region divided by a line segment connecting the symmetric position and the central position.

  The image processing program of the present invention is for causing a computer to function as the image processing apparatus.

  According to this program, the operator can freely set the symmetry position and the center position of the processing object in the input image, the size index, and the unused portion, and the processing object set by the operator is symmetric. Image deformation of the input image is performed using the position and the center position, the size index, or the unused portion.

  The recording medium of the present invention stores the image processing program.

  According to this recording medium, the image processing program being stored can be read and executed by various computers.

  The present invention, when identifying an object such as a vehicle or a person photographed with a camera, can be realized with a semiconductor without using a computer with high processing capability, and can be incorporated into a camera at various angles. Thus, it is possible to obtain an approximate front image that can be used for identification processing of an object such as a vehicle or a person photographed. In particular, the symmetry position and the center position, the size index, or the unused portion can be set automatically or manually for the image of the processing object, and the convenience can be improved. Further, the accuracy of the corrected image can be improved by setting the symmetry position and the center position, the size index, or the unused portion with a simple operation on the image of the processing object.

1 is a block diagram showing a functional configuration of an image processing apparatus according to Embodiment 1 of the present invention. The flowchart which shows the process of the target object extraction part of the image processing apparatus which concerns on Embodiment 1 of this invention. In the processing in the object extraction unit of the image processing apparatus according to the first embodiment of the present invention, (a) is a diagram showing an input image, and (b) is a diagram showing an extracted target region. 7 is a flowchart showing processing of the image correction unit of the image processing apparatus according to the first embodiment of the present invention. The figure for demonstrating the process in the image correction part of the image processing apparatus which concerns on Embodiment 1 of this invention. In the processing in the central position correction unit of the image processing apparatus according to the first embodiment of the present invention, (a) is a diagram for explaining the central position detection processing, and (b) is a diagram for explaining the connection of processing regions. FIG. 9C is a diagram for explaining the center position correction. The flowchart which shows the process of the center position correction | amendment part of the image processing apparatus which concerns on Embodiment 1 of this invention. The flowchart which shows in detail the center position calculation step within the attention rectangle of the image processing apparatus which concerns on Embodiment 1 of this invention. In the processing in the collation region extraction unit of the image processing apparatus according to Embodiment 1 of the present invention, (a) is a diagram for explaining collation range extraction processing, and (b) is a diagram for explaining correlation evaluation values. (C) is a figure for demonstrating the collation range. In the processing in the collation region extraction unit of the image processing apparatus according to the first embodiment of the present invention, (a) is a diagram for explaining a left half-inversion copy image, and (b) is a right half-inversion copy image. Illustration for The flowchart which shows the process of the collation area | region extraction part of the image processing apparatus which concerns on Embodiment 1 of this invention. FIG. 3 is a block diagram showing a functional configuration of an image processing apparatus according to Embodiment 2 of the present invention. The flowchart which shows the process of the target object extraction part of the image processing apparatus which concerns on Embodiment 2 of this invention. In the processing in the object extraction unit of the image processing apparatus according to the second embodiment of the present invention, (a) shows an input image, and (b) shows an extracted object region. In the processing in the image correction unit of the image processing apparatus according to the second embodiment of the present invention, (a) is a diagram for explaining an object region, (b) is a diagram for explaining inclination correction, (c) ) Is a diagram for explaining reverse perspective transformation 7 is a flowchart showing processing of the image correction unit of the image processing apparatus according to Embodiment 2 of the present invention. The flowchart which shows the process of the center position correction | amendment part of the image processing apparatus which concerns on Embodiment 2 of this invention. In the processing of the central position correction unit of the image processing apparatus according to the second embodiment of the present invention, (a) is a diagram for explaining central position calculation, and (b) is a diagram for explaining central position correction. In the processing of the central position correction unit of the image processing apparatus according to the second embodiment of the present invention, (a) is a diagram for explaining a left half-side reversed copy image, and (b) is a right half-side reversed copy image. Figure of FIG. 3 is a block diagram showing a schematic configuration of an image processing apparatus according to Embodiment 3 of the present invention. Flowchart for explaining image processing of the image processing apparatus according to the third embodiment of the present invention. The figure which shows an example of an input image in the image processing apparatus which concerns on Embodiment 3 of this invention. The figure which shows the example which the operator specified two symmetrical points in the input image of FIG. The figure which shows the line segment which connects two symmetrical points designated by the operator in the image of FIG. The figure which shows the example which the operator specified the middle point in the line segment which connects two symmetrical points which the operator specified in the image of FIG. FIG. 24 is a diagram showing an example in which the operator has specified a number of two symmetrical points required in the image of FIG. The figure which shows the example which divided | segmented the image of FIG. 26 into several trapezoid area | regions In the image of FIG. 27, a diagram showing an example in which each trapezoidal region is transformed into a rectangle and the horizontal scale is unified. The figure which shows the example which matches a rectangle in the center position in the image of FIG. The figure which shows the image after the process in FIG. 29 is complete | finished The figure which shows the example of the size correction with respect to the image of FIG. The figure which shows the image after the size correction process in FIG. 31 is complete | finished The figure which shows the other example of the size correction with respect to the image of FIG. FIG. 32 is a diagram showing an example in which a line-symmetric frame is displayed at the center position on the image of the size correction processing of FIG. 32 and the image is transformed into a size-corrected image by adjusting the vertical and horizontal scales. The figure which shows the example which displayed the partial region frame with respect to the image of FIG. The figure which shows the example which designated the partial area | region which an operator does not use with respect to the image of FIG. 36 is a diagram showing an example in which the image of the designated partial area is transferred and copied from the opposite side in the image of FIG. The figure which shows the image after transcription | transfer copying in FIG. Flowchart for explaining image processing in an application example of the image processing apparatus according to the third embodiment of the present invention.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments for carrying out the invention will be described in detail with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram showing a functional configuration of the image processing apparatus according to Embodiment 1 of the present invention. The image processing apparatus according to the present embodiment can generate an approximate front image for vehicle identification. As shown in FIG. 1, the object extraction unit 101, the image correction unit 102, and the feature image generation unit 103, a center position correction unit 104, and a collation region extraction unit 105.

  The target object extraction unit 101 detects a target object such as a vehicle or a person from an image taken by a camera (not shown) installed on a traffic path, and extracts a target area. FIG. 2 is a flowchart showing processing of the object extraction unit 101. In FIG. 2, first, a license plate in the image is detected in step S101, and it is determined whether or not a license plate is detected in step S102. If it is determined that there is a vehicle when the license plate is detected, the coordinates of the four corners of the license plate are detected in step S103. Next, in step S104, an object region is determined based on the detected license plate position and size.

  In FIG. 3, (a) shows the input image, and (b) shows the detected license plate four corner coordinates P and the object region R1. In the license plate detection step of step S101 in FIG. 2, for example, a known method (for example, see Japanese Patent Application Laid-Open No. 2008-217347) such as detecting a straight line of a license plate frame from an image using Hough transform or the like is used. Detect the license plate. The object extraction unit 101 can use not only a method of detecting a license plate but also a pattern matching method using a vehicle-shaped template. Moreover, when a continuous image is input, a method of detecting a moving object from a change between successive images and setting it as a target can be used.

  The image correction unit 102 corrects image distortion due to the camera installation position and the tilt of the object. That is, the image correction unit 102 corrects an image based on the installation position of the camera for the image that is determined by the object extraction unit 101 to show the vehicle. FIG. 4 is a flowchart showing the processing of the image correction unit 102. In FIG. 4, the correction parameter is calculated in step S201. In this case, if the camera is fixed, a correction parameter for reverse perspective transformation can be determined from the distance, angle (deflection angle and depression angle) between the camera and a reference point on the object, and the rotation angle of the camera itself. Next, in step S202, the image is subjected to inverse perspective transformation using the parameters obtained in step S201. The reverse perspective transformation is a transformation in which a vertical plane directed in an oblique direction is directed to the viewing direction. In this case, conversion is performed so that the surface facing the front of the vehicle is directed to the line of sight. At that time, the points in the same depth plane are corrected to an image that faces the front correctly, but the positions of the planes with different depths are shifted to the left and right on the image, so that the entire vehicle image becomes a more distorted image. . An example is shown in FIG.

  Even if the camera is movable or the installation angle is unknown, the correction parameter can be determined based on the four corner coordinates of the license plate detected by the object extraction unit 101. If the vertical axis of the vehicle does not coincide with the vertical axis of the image due to the inclination of the vehicle itself or the like even after correction based on the camera position, the two are matched by inclination correction (step S203). At this time, the left and right sides of the license plate can be used as the direction of the vertical axis of the vehicle.

  With the above correction, the left-right direction on the front of the vehicle matches the horizontal direction of the image, and the vertical (up-down) direction matches the vertical direction of the image.

  Next, the feature image generation unit 103 generates a feature image to be used in the subsequent processing for the image corrected by the image correction unit 102. For example, an edge image or the like subjected to gradation processing, binary image formation, or filter processing such as Hough transform or Sobel filter is generated. Note that the pixel value of the original image may be used as it is without performing the feature image generation process.

  Next, the center position correction unit 104 detects the center position where the left-right correlation at each height is the highest for the object region in the feature image generated by the feature image generation unit 103, and the center position Are aligned vertically. The processing of the center position correction unit 104 will be described with reference to FIG. 6A shows the object region R ″ 1 in the feature image obtained by the feature image generation unit 103. Here, the horizontal direction of the image is the x direction, and the vertical direction is the y direction. The coordinates of the four corners of the region R ″ 1 are (x1, y1), (x2, y1), (x2, y2), (x1, y2). If the pixel value at the coordinates (x, y) is G (x, y), the difference between the left and right at the coordinates (x, y) can be obtained by equation (1). Here, K is a constant determined based on the size of the license plate.

K
D (x, y) = Σ | G (x−k, y) −G (x + k, y) | (1)
k = 1

  For each y coordinate in the rectangular area R ″ 1, the left and right differences D (x, y) are obtained for all x in the range from x1 + K to x2−K, and the x having the smallest value is determined in the height y coordinate. (The smaller the difference is, the higher the correlation value is.) A curve connecting the central positions (x0 (y), y) obtained by the respective y coordinates is shown in FIG. L1 shown in FIG.

  When the center positions at all y coordinates in the rectangular area are obtained, the same processing is performed on the rectangular area R ″ 2 adjacent to the upper and lower sides, and the line L1 connecting the center positions is extended. However, the second rectangular area R ″ 2 is arranged so that the median value is at the center of the rectangular area at the connection y coordinate (y1 for the lower side, y2 for the upper side) with the rectangular area R ″ 1. At this time, when the region is outside the image, the portion outside the region is not processed.

  FIG. 6C is a diagram in which the center positions obtained as described above are aligned vertically. In the actual image, since the center position does not always continue smoothly, if the obtained center position is smoothed before being aligned, a more natural corrected image can be obtained. The smoothing is performed by a known method such as using a moving average. Alternatively, instead of smoothing, the image may be divided into small areas in the horizontal direction, and within the small area, the same center position x coordinate (for example, the section average value) may be taken for all y coordinates. In this case, the boundary to be divided into small areas is a horizontal edge position obtained by color change in a color image, change in gray value in a gray image, or various filter processes such as Hough transform and Sobel filter. Can be used.

  FIG. 7 is a flowchart showing the processing of the center position correction unit 104. In FIG. 7, first, in step S301, a central position x coordinate is obtained for each y coordinate in the object region R ″ 1 in the feature image obtained by the feature image generation unit 103. Next, in step S302, R ″. The position of the rectangular area R ″ 2 adjacent to the upper side is determined by the algorithm. Next, in step S303, it is determined whether or not the determined rectangular area is included in the image. In step S304, the center position x coordinate is obtained for each y coordinate in the target rectangle, and if the determined rectangular area is not included in the image in step S303 (that is, the upper process is completed). In steps S305 to S307, the same processing as in steps S302 to S304 is performed on the lower side of the original object region R ″ 1. In step S306, when the lower target rectangular area is no longer in the image (that is, when the lower processing is completed), the center position x coordinate obtained for each y coordinate is smoothed in step S308. Next, in step S309, an average value X0 of the center position x coordinate after smoothing for each y coordinate is obtained. In step S310, the pixels in the horizontal direction of each y coordinate are shifted left and right so that the center position for each y coordinate matches the average value X0 obtained in step S309.

  FIG. 8 is a flowchart showing the detailed processing of the center position calculation step (steps S301, S304, S307) in the target rectangle. In FIG. 8, first, in step S311, y = y1 is set. Next, in step S312, it is determined whether y1 ≦ y and y ≦ y2, and if not applicable, this processing is terminated. If applicable, x = x1 + K is set in step S313. After performing the process of step S313, in step S314, it is determined whether x1 + K ≦ x and x ≦ x2 + K. If not, x0 is stored as the center position in y in step S319, and then is set as y ++. The process returns to step S312. On the other hand, in step S314, if applicable, in step S315, the left-right difference D (x, y) when (x, y) is the center point is calculated. In step S316, it is determined whether D (x, y) <Dmin. If not, Dmin = D (x, y) and x0 = x are set in step S317. If it is determined in step S316 that it is not applicable, or after performing the process of step S317, the process returns to step S314 as x ++.

  In this example, the sum of the difference between the left and right values is used as the index of the right and left target levels. However, the sum of square errors may be used, and substitution with similarity and correlation values is also possible. is there. However, when the evaluation is performed using the similarity or the correlation value, the magnitude relationship is reversed from the case where the sum of the differences between the left and right values is used.

  Here, an example in which the center position x coordinate is smoothed for each y coordinate has been described, but the center position may be determined for each region such as a windshield, a bonnet, a grill, and a bumper. Each area is divided at a position where the color and brightness change greatly.

  Next, the process of the collation area extraction unit 105 will be described with reference to FIG. The collation region extraction unit 105 determines a region that can be used for identification processing in the image obtained by the center position correction unit 104. FIG. 9A shows an object region in a feature image that has been subjected to center position correction. As in FIG. 6A, the horizontal direction of the image is the x direction, the vertical direction is the y direction, and the pixel value at the coordinates (x, y) is G (x, y). When the x coordinate of the central position is x0, the left and right difference in the coordinate (x0, y) is obtained in the range from (x0−t, y) to (x0 + t, y), which is obtained by the following equation (2). . t takes a range from a predetermined constant tmin to tmax = Min (x0−x1, x2−x0).

t
D ′ (y, t) = Σ | G (x0−k, y) −G (x0 + k, y) | (2)
k = 1

  The left-right difference D ′ (y, t) in the range t obtained for each y-coordinate is added in the range from y1 to y2, and the correlation evaluation value S (t) in the left-right region is obtained. The equation (3) for obtaining the correlation evaluation value S (t) is shown below.

y2
S (t) = ΣD ′ (y, t) / t (3)
y = y1

  The higher the correlation, the smaller this value. As shown in (b) of FIG. 9, the correlation evaluation value S (t) increases with t0 as a boundary. Based on this t0, a range from x0-t0 to x0 + t0 is set as a collation region (<R>) ((c) in FIG. 9). When the minimum value Smin of S (t) is smaller than the predetermined value Th, a left half image in the range of x0−t0 ≦ x ≦ x0 is used instead of the right half image in the range of x0 ≦ x ≦ x0 + t0. Instead of the left half-side inverted copy image (FIG. 10A) that has been flipped horizontally and the left half-side image in the range of x0−t0 ≦ x ≦ x0, the right half-side image in the range of x0 ≦ x ≦ x0 + t0 Two images of the right half-side inverted copy image (FIG. 10B) that are inverted and fitted are generated, and used as a collation image instead of the center position correction image generated by the center position correction unit 104. As the collation range, a range of x0−t0 ≦ x ≦ x0 + t0 is used as in the case of the center position corrected image.

  FIG. 11 is a flowchart showing the processing of the collation area extraction unit 105. In FIG. 11, t = tmin and S (t) = 0 are set in step S411, and it is determined in step S412 whether tmin ≦ t and t ≦ tmax. If yes, in step S413, y = y1, and in step S414, determine whether y1 ≦ y and y ≦ y2, and if so, in step S415, set (x0, y) as the center point. The difference D ′ (y, t) between the left and right range t is calculated. In step S416, S (t) + = D '(y, t), y ++ is set, and the process returns to step S414. If it is determined in step S414 that y1 ≦ y and y ≦ y2 do not apply, the correlation evaluation value S (t) is stored in step S417, and thereafter, y ++ is returned to step S412. On the other hand, if it is determined in step S412, that tmin ≦ t and t ≦ tmax are not satisfied, S (t) is analyzed in step S419, t = t0 where the value increases rapidly is determined, and x0−t0 ≦ x ≦ x0 + t0. Is the collation range. In step S420, it is determined whether Smin> Th. If not, the process is terminated. If Smin> Th, the left side reverse copy image and the right side reverse copy image are generated in step S421. To do. After finishing the process of step S421, this process is complete | finished.

  Thus, in step S411 to step S418, the correlation evaluation value S (t) for t of tmin ≦ t ≦ tmax is obtained, and in step S419, the t value t0 where the correlation evaluation value S (t) increases rapidly is obtained. (A large difference means that the correlation is broken.) Then, based on the value, a range of x0−t0 ≦ x ≦ x0 + t0 of the image subjected to the center position correction is set as a collation range. Here, as in the central position correction unit 104, the sum of the differences between the left and right values has been described as the index of the left and right target degrees, but the sum of the square errors may be used. Also, substitution with similarity and correlation values is possible. However, when the evaluation is performed using the similarity or the correlation value, the magnitude relationship is reversed from the case where the sum of the differences between the left and right values is used.

  In this way, when identifying an object such as a vehicle photographed with a camera installed on a traffic path, it is a simple process that can be realized even with a semiconductor that can be incorporated in the camera without depending on a computer with high processing capacity, It is possible to obtain an approximate front image of an object having symmetrical characteristics. Thus, it is possible to perform identification processing of an object such as a person photographed from various angles.

(Embodiment 2)
The image processing apparatus according to the second embodiment of the present invention can generate an approximate front image for human face authentication, and as shown in FIG. 12, an object extraction unit 1011 and an image correction unit 1021. A feature image generation unit 1031, a center position correction unit 1041, and a collation region extraction unit 1051.

  FIG. 13 is a flowchart showing the processing of the object extraction unit 1011. In FIG. 13, first, in step S501, a human face in the image is detected, and then in step S502, it is determined whether or not a face is detected. If a face is detected, it is determined that there is an object. In step S503, eye opening position coordinates are detected. In step S504, an object region is determined based on the detected eye opening position.

  In FIG. 14, (a) shows the input image, and (b) shows the extracted eye opening position coordinates P and the object region R1. The face detection step in step S501 is executed using a known face detection method such as a pattern matching method using a template (see, for example, JP-A-2004-318204). In addition, when continuous images are input, a method of detecting a moving object from a change between successive images and setting it as a target can be used.

  The image correction unit 1021 corrects the image based on the camera position for the image determined by the object extraction unit 1011 as having a face. FIG. 15 is a diagram illustrating processing in the image correction unit 102. FIG. 16 is a flowchart showing the processing of the image correction unit 102. If the object is a face, first, in step S601, tilt correction is performed so that the vertical axis of the face matches the vertical axis of the image. At this time, the perpendicular line P3H drawn from the mouth position (P3) to the straight line (P1P2) connecting both eye positions P1 and P2 is taken as the vertical axis of the face ((a) in FIG. 15), and this perpendicular line P3H is in the vertical direction of the image. The rotation correction is performed ((b) of FIG. 15). Next, in step S602, correction parameters for reverse perspective transformation are determined based on the ratio of the distance between the left and right eye position coordinates and the foot H of the perpendicular (P1H: P2H). In step S603, the image is subjected to inverse perspective transformation using the parameters obtained in step S602. With the above correction, the left and right direction of the front face of the face coincides with the horizontal direction of the image, and the vertical (up and down) direction coincides with the vertical direction of the image ((c) of FIG. 15).

  Next, the feature image generation unit 1031 generates a feature image to be used in subsequent processing for the image that has been subjected to image correction by the image correction unit 1021. For example, an edge image or the like subjected to gradation processing, binary image formation, or filter processing such as Hough transform or Sobel filter is generated. Alternatively, the pixel value of the original image may be used as it is without performing the feature image generation process.

  FIG. 17 is a flowchart showing processing in the center position correction unit 1041. 17, in step S701, the center position x coordinate is obtained for each y coordinate in the object region R ″ 1 in the feature image obtained by the feature image generation unit 1031 ((a) in FIG. 18). In step S702, the center position x coordinate obtained for each y coordinate is smoothed, and in step S703, the average value X0 of the center position x coordinate for each smoothed y coordinate is obtained. In step S704, the pixels in the horizontal direction of each y coordinate are shifted left and right so that the center position for each y coordinate matches the average value X0 obtained in step S703 (FIG. 18B). The processing in step S701 is the same as the step of calculating the center position in the target rectangle (steps S301, S304, S307) when the vehicle is the target. Treatment in 51 may also be the same as if the target vehicle.

  In this way, when identifying an object such as a vehicle photographed with a camera installed on a traffic path, it is a simple process that can be realized even with a semiconductor that can be incorporated in the camera without depending on a computer with high processing capacity, It is possible to obtain an approximate front image of an object having symmetrical characteristics. This makes it possible to perform identification processing of objects such as vehicles and people photographed from various angles.

  The object extraction units 101 and 1011, the image correction units 102 and 1021, the feature image generation units 103 and 1031, the center position correction units 104 and 1041, and the collation region extraction units 105 and 1051 described in the first and second embodiments. Each of these functional blocks is typically realized as an LSI which is an integrated circuit. These may be individually made into one chip, or may be made into one chip, such as a part or all of them.

  The method of circuit integration is not limited to LSI, and may be realized by a dedicated circuit or a general-purpose processor.

  Also, an FPGA (Field Programmable Gate Array) that can be programmed after manufacturing the LSI or a reconfigurable processor that can reconfigure the connection and setting of circuit cells inside the LSI may be used.

  Furthermore, if integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology or a derivative other technology, it is naturally also possible to carry out function block integration using this technology.

(Embodiment 3)
FIG. 20 is a block diagram showing a schematic configuration of an image processing apparatus according to Embodiment 3 of the present invention. 20, the image processing apparatus according to the present embodiment includes an image input unit 110, a display unit 111, an image output unit 112, a condition recording unit 113, an image deformation unit 114, a symmetrical position instruction unit 115, A size instruction unit 116 and a use part instruction unit 117 are provided.

  The image input unit 110 inputs an image of a processing target photographed by a camera (not shown). The display unit 111 displays an image of the processing object, various instructions, and a processing image of the processing object according to the various instructions. Here, the various instructions are a target position instruction, a size index instruction, and a usage part instruction. The image output unit 112 outputs a processed image of the processing object processed by the image deformation unit 114. The condition recording unit 113 records a ready-made value such as the size of the processing object. The symmetrical position instruction unit 115 instructs the image deforming unit 114 about the symmetrical position and the center position of the object in the image. The operator gives instructions for the symmetrical position and the central position. That is, the operator operates the target position instruction unit 115 while viewing the image of the processing object displayed on the display unit 111 to specify the symmetrical position and the center position of the processing object image. By this designation, the symmetric position instruction unit 115 instructs the image deformation unit 114 about the symmetric position and the center position of the target specified by the operator.

  The size instruction unit 116 instructs the image transformation unit 114 to specify the size index of the object in the image. The operator instructs the size index. That is, the operator designates the size index of the image of the processing object by operating the size instruction unit 116 while viewing the image of the processing object displayed on the display unit 111. By this designation, the size instruction unit 116 instructs the image deformation unit 114 to specify the size index of the image of the processing target specified by the operator.

  The use part instruction unit 117 instructs the image deforming unit 114 to designate a non-use part of the processing target in the image. The operator instructs the non-use part. That is, the operator operates the utilization part instructing unit 117 while viewing the image of the processing object displayed on the display unit 111 and designates an unused part of the processing object. With this designation, the use part instruction unit 117 instructs the image deformation unit 114 to designate a non-use part of the target designated by the operator.

  The image deformation unit 114 performs image deformation (“first image deformation”) on the image input by the image input unit 110 using the symmetric position and the center position specified by the symmetric position instruction unit 115. In addition, after the first image deformation, the image deformation (“second image deformation”) is performed using the existing size value recorded in the condition recording unit 113 and the size index instructed by the size instruction unit 116. To do. In addition, after the second image deformation, the non-use part designated by the use part instructing unit 117 is supplemented with an image of a symmetric part (“third image deformation”). Then, the image obtained by the third image deformation is output to the image output unit 112.

Next, the operation of the image processing apparatus according to this embodiment will be described.
FIG. 21 is a flowchart for explaining the image processing of the image processing apparatus according to the present embodiment. In FIG. 21, first, in step S <b> 1001, the image input unit 110 inputs an image to be processed and outputs it to the image deformation unit 114. The processing target image output to the image transformation unit 114 is displayed on the display unit 111.

  When the operator performs an operation of designating the symmetric position and the center position of the processing target object while the processing target image is displayed on the display unit 111, in step S1002, the symmetric position instruction unit 115 performs processing in the image. The image deforming unit 114 is instructed about the symmetrical position and the center position of the object. When receiving an instruction from the symmetric position instruction unit 115, the image deforming unit 114 deforms the processing object into a front image using the symmetric position and the center position of the processing object in the image in step S1003 (first image). Image deformation processing). The image of the processing object transformed into the front image is displayed on the display unit 111.

  After the object to be processed is deformed to the front image, in step S1004, the image deforming unit 114 determines whether or not there is a symmetrical position instruction, and if it continues to determine that there is a symmetrical position instruction, the process returns to step S1002 and continues to the symmetrical position. An instruction from the instruction unit 115 is received. On the other hand, if it is determined that there is no symmetrical position instruction, the next instruction is waited. At this time, a front image is displayed on the display unit 111.

  After the first image deformation process is performed, when the operator performs an operation to specify the size index of the processing object while the front image is displayed on the display unit 111, the size instruction unit 116 is displayed in step S1005. Instructs the image transformation unit 114 of the size index designated by the operator. Upon receiving an instruction from the size instruction unit 116, the image transformation unit 114 uses the instructed size index and the size of the target object using the size already recorded in the condition recording unit 113 in step S1006. Is corrected (second image deformation process). The image of the processing object whose size has been corrected is displayed on the display unit 111.

  Next, in step S1007, the image transformation unit 114 determines whether or not there is a size index instruction. If it is determined that there is a size index instruction, the process returns to step S1005 and continuously receives an instruction from the size instruction unit 116. On the other hand, if it is determined that there is no size index instruction, the next instruction is awaited. At this time, the front image whose size has been corrected is displayed on the display unit 111.

  After the second image transformation process is performed, if the operator performs an operation for instructing a non-use part of the processing object while the front image is displayed on the display unit 111, a use part instruction is provided in step S1008. The unit 117 instructs the image deforming unit 114 to designate a non-use portion designated by the operator. Upon receiving an instruction from the use part instruction unit 117, the image deformation unit 114 supplements the non-use part of the instructed image with a symmetrical part image (third image deformation process) in step S1009. The image of the processing object supplemented with the unused portion is displayed on the display unit 111.

  Next, in step S1010, the image deforming unit 114 determines whether or not there is an instruction for refilling the unused part. Accepts instructions from. On the other hand, if it is determined that there is no instruction for refilling the unused portion, the next instruction is awaited. At this time, the display unit 111 displays a front image supplemented with unused portions.

  As described above, the image deforming unit 114 performs the first image deforming process based on the symmetrical position instruction, the second image deforming process based on the size index instruction, and the third image deforming process based on the non-use part instruction, and then in step S1011. The image after each deformation process is output, and further, the deformation process condition is output in step S1012, and the process ends. Here, the deformation conditions are the symmetric position and center position of the object designated in step S1002, the size index designated in step S1005, and the unused condition designation condition in the image designated in step S1008. The image output unit 112 outputs an input image, a final deformed image, and deformation conditions. Steps S1001 to S1012 may be described in a program that can be executed by a computer.

Next, the operation of the image processing apparatus according to the present embodiment will be described with a specific example.
FIG. 22 is a diagram illustrating an example of an input image. The input image shown in the figure includes _a front image Ia of the vehicle. 23 to 26 are diagrams showing each process in the first image deformation processing. The first image deformation process is performed by the symmetrical position instruction unit 115 and the image deformation unit 114. First the operator performs an operation of instructing the symmetrical position P _1, P ₂ with respect to the front image I _a vehicle. When this operation is performed, as shown in FIG. 23, the image deformation unit 114 displays the symmetric positions P ₁ and P ₂ on the display unit 111 in accordance with an instruction from the symmetric position instruction unit 115. Further, as shown in FIG. 24, a line segment L ₁ connecting the symmetrical positions P ₁ and P ₂ is displayed.

Then, when an operation by the operator to indicate the center position P _c1 on the line segment L _1, as shown in FIG. 25, the image deformation unit 114 on the line L ₁ in accordance with an instruction from the symmetry position indicator 115 The center position _Pc1 is displayed. When the operator performs the same operation (instruction for symmetry position and instruction for center position) for a required number of times, a result as shown in FIG. 26 is obtained. If the position indicated by the operator is out of position, the position can be corrected or deleted.

Next, the second image deformation process will be described. The second image deformation process is performed by the size instruction unit 116 and the image deformation unit 114. When the operator presses a “conversion” button (not shown) in the state where the first image deformation process is completed, as shown in FIG. 27, the image deformation unit 114 converts the front image _Ia into a plurality of trapezoidal regions ( An example is divided into symbols S ₁ and S ₂ . Next, as shown in FIG. 28, the image transformation unit 114 transforms each trapezoidal region into a rectangular region, and further adjusts the left and right scales so that the symmetrical position is the same length from the central position (an example is denoted by S _1). ', _{S 2'} is shown by). Then, as shown in FIG. 29, it deforms the front image I _b together the center position of each rectangular area. Figure 30 shows the front image _{I b.} The operator confirms the front image _Ib , and if the correction has not been completed, the operation of the symmetrical position instruction is performed again.

Then, when an operation by the operator to instruct the plate position is the size indicator front image I _b, the size indication 116, as shown in FIG. 31, the display plate position P ₅₀ to P ₅₃ that is instructed Displayed on the unit 111. When the operator presses a “conversion” button (not shown), the image deforming unit 114 refers to the plate size recorded in the condition recording unit 113 and adjusts the vertical and horizontal scales to the size-corrected image I _c1 shown in FIG. And is displayed on the display unit 111.

In addition to specifying the four corners of the plate and adjusting the size with the nominal plate value, the size correction method specifies the vehicle width (W) and vehicle height (H) as shown in FIG. It is also possible to adjust the size. Size instruction unit 116 displays a line-symmetrical frame FR ₁ at a central location in the front image I _b displayed on the display unit 111 as shown in FIG. 34. The operator moves the width and vertical position of the frame FR ₁ and performs an operation for instructing the vehicle width / height as the size index. After the operation for instructing the vehicle width / vehicle height is performed, when the operator presses a “convert” button (not shown), the image transformation unit 114 displays the catalog of the vehicle width / vehicle height recorded in the condition recording unit 113. By referring to the values, the scales of the vertical and horizontal directions are _matched and transformed into a size corrected image _Ic2 shown in FIG. The operator confirms the size-corrected image I _c1 or I _c2 and _{repeats the} size instruction operation if correction is not completed.

Next, the third image deformation process will be described. The third image deformation process is performed by the use part instruction unit 117 and the image deformation unit 114. As shown in FIG. 35, the image transformation unit 114 according to instruction from a partial instruction unit 117 displays a partial area frame FR ₂ in size correction image _{I c2.} When the operator looks at this display and performs an operation for designating the non-use part G as shown in FIG. 36, the image transformation unit 114 follows the instruction from the use part instruction unit 117 and instructs the non-use part G. The partial area frame corresponding to is not displayed. When the operator presses a “convert” button (not shown) in this state, as shown in FIG. 37, the image transformation unit 114 supplements the unused portion G with the image of the symmetric portion, and as shown in FIG. The non-use partial supplement image _Id is transformed and displayed. The operator confirms the non-use partial replenishment image I _d, redo operations use partial instructions if correction incomplete.

  In this way, when an object such as a vehicle photographed by a camera is identified, it has an interface that allows a human to freely set the center of symmetry and both ends, thereby providing a degree of freedom in use. Rise. Of course, the processing after making these settings is performed in the same manner as the image processing apparatuses of the first and second embodiments.

  In the image processing apparatus according to the present embodiment, the symmetrical position and the center position of the object in the image are indicated. However, if the image has already been front-corrected, only the center instruction may be used.

  Further, in the image processing apparatus according to the present embodiment, if the processing results of the respective instruction units of the symmetric position instruction unit 115, the size instruction unit 116, and the usage portion instruction unit 117 are not expected, the instructions may be corrected and input. Absent.

  In the image processing apparatus according to the present embodiment, the image deforming unit 114 supplements the non-use part designated by the use part instructing part 117 with the image of the symmetric part, and outputs the result to the image output part 112. Without supplementing the image, the unused part information instructed by the utilization part instructing unit 117 is output together with the pre-replenishment image, converted into the feature amount used for the process in the subsequent recognition process, and the like. You may make it supplement with a feature-value.

  The image processing apparatus according to the present embodiment performs processing for each instruction. However, all the instructions may be given at once and processed in a batch. The operation in this case is as shown in the flowchart of FIG. As shown in FIG. 39, after the image input unit 110 inputs an image in step S1101, the symmetric position instruction unit 115 instructs the image deformation unit 114 about the symmetric position and the center position of the processing object in the image in step S1102. In step S1103, the size instruction unit 116 instructs the image transformation unit 114 on the size index designated by the operator. In step S1104, the use part instruction unit 117 instructs the image deformation unit 114 to designate a non-use part specified by the operator.

  After the use part instruction unit 117 instructs the image deformation unit 114, in step S1105, the image deformation unit 114 deforms the processing object into a front image using the symmetrical position and the center position of the processing object in the image (first image). And the deformed front image is displayed on the display unit 111. Next, the image transformation unit 114 corrects the size of the object using the size index instructed from the size instruction unit 116 and the ready-made value of the size recorded in advance in the condition recording unit 113 (second image). Deformation process), and the image whose size has been corrected is displayed on the display unit 111. Next, the image transformation unit 114 supplements the unused portion of the image instructed from the usage part instruction unit 117 with the symmetrical portion image (third image transformation process), and displays the image supplemented with the unused portion as the display unit 111. To display.

  Next, in step S1106, the image transformation unit 114 continues to determine whether there is an instruction for any one of the designation of the symmetric position, the size designation, or the designation of the use portion, and if any one is designated, Returning to step S1102, the processing instructed is repeated. If there is no instruction from any of the designation of the symmetric position, the designation of the size, and the designation of the used part, the processed deformed image is output in step S1107, and the deformation condition is further output in step S1108. The processing order of the deformation condition instruction and the deformation and display of the input image may be switched. Steps S1101 to S1108 may be described in a program that can be executed by a computer.

  The functional blocks of the image input unit 110, the image output unit 112, the condition recording unit 113, the image transformation unit 114, the symmetric position instruction unit 115, the size instruction unit 116, and the use part instruction unit 117 described in the third embodiment are as follows. Typically, it is realized as an LSI which is an integrated circuit. These may be individually made into one chip, or may be made into one chip, such as a part or all of them.

  The method of circuit integration is not limited to LSI, and may be realized by a dedicated circuit or a general-purpose processor.

  Further, an FPGA that can be programmed after the LSI is manufactured, or a reconfigurable processor that can reconfigure the connection and setting of circuit cells inside the LSI may be used.

  Furthermore, if integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology or a derivative other technology, it is naturally also possible to carry out function block integration using this technology.

  As described above, the embodiment in which the present invention is applied to the face of a vehicle and a person has been described. However, the image processing apparatus of the present invention is not limited to the above-described embodiments, and the orientation of an object is specified by some measurement method. If possible, the present invention can be applied to any symmetrical object.

  The image processing apparatus according to the present invention can obtain an approximate front image of an object such as a vehicle or a person from an image taken by a monitoring camera or the like installed on a traffic path or the like with a simple process. It has an effect and can be applied to an entrance / exit system using a vehicle identification system or face authentication.

  Further, the present invention can be applied to uses such as a search system for selecting an image of a specific object from a database in which images of vehicles and people taken from various angles are stored.

101, 1011 Object extraction unit 102, 1021 Image correction unit 103, 1031 Feature image generation unit 104, 1041 Center position correction unit 105, 1051 Collation area extraction unit 110 Image input unit 111 Display unit 112 Image output unit 113 Condition recording unit 114 Image deformation unit 115 Symmetric position instruction unit 116 Size instruction unit 117 Usage part instruction unit

Claims (19)

An object extraction unit that detects an object such as a vehicle or a person from an image captured by a camera and extracts an object area;
An image correction unit that corrects image distortion due to the installation position of the camera and the inclination of the object;
A center position correcting unit that detects the center position where the left-right correlation at each height is highest for the object region in the image corrected by the image correcting unit, and aligns the center position vertically;
A collation region extraction unit that determines a region that can be used for identification processing in the image obtained by the central position correction unit;
An image processing apparatus. An object extraction unit for detecting an object such as a vehicle or a person from an image taken by a camera and extracting an object area;
An image correction unit that corrects image distortion due to the installation position of the camera and the inclination of the object;
A feature image generation unit that generates a feature image for the image corrected by the image correction unit;
A center position correction unit that detects a center position where the left-right correlation at each height is highest for the object region in the feature image generated by the feature image generation unit, and aligns the center position vertically;
A collation region extraction unit that determines a region that can be used for identification processing in the image obtained by the central position correction unit;
An image processing apparatus.   The center position correction unit detects a center position where the left-right correlation at each height is the highest for the object region in the feature image generated by the feature image generation unit, and smoothes the detected center position. The image processing apparatus according to claim 1, wherein after the conversion, the center position is aligned vertically.   The center position correction unit detects the center position where the left-right correlation at each height is the highest for the object region in the feature image generated by the feature image generation unit, and then horizontally shifts the image in the horizontal direction. The image processing apparatus according to claim 1, wherein a central position is determined for each divided small area, and the central positions of the small areas are aligned vertically.   The collation region extraction unit is configured to change the left and right regions when the region width of the left and right regions sandwiching the center position is varied within a predetermined range with respect to the object region in the center position correction image generated by the center position correction unit. The image processing apparatus according to claim 1, wherein a change in the correlation value is examined, and an area width where the correlation value rapidly decreases is set as a width of the collation range.   The collation region extraction unit is configured to change the left and right regions when the region width of the left and right regions sandwiching the center position is varied within a predetermined range with respect to the object region in the center position correction image generated by the center position correction unit. When the change in the correlation value is examined, and the value in the region width having the highest correlation is smaller than the predetermined value, an image in which either the left or right half of the collation image is replaced with the other inverted copy is generated, and this is referred to as the collation image. The image processing apparatus according to any one of claims 1 to 4. An object extraction step of detecting an object such as a vehicle or a person from an image taken by a camera and extracting an object area;
An image correction step for correcting image distortion due to the installation position of the camera and the inclination of the object;
A feature image generation step for generating a feature image for the image corrected in the image correction step;
A center position correcting step for detecting a center position where the left-right correlation at each height is highest for the object region in the feature image generated in the feature image generating step, and aligning the center position vertically,
A collation region extraction step for determining a region that can be used for identification processing in the image obtained by the center position correction step;
An image processing method comprising: An image input unit for inputting an image of the processing object;
A display unit for displaying an image input by the image input unit and displaying an instruction of an operator;
An instruction unit for instructing a symmetric position and a center position of the processing object in the image input by the image input unit according to an operator's instruction;
An image deformation unit that deforms an image input by the image input unit using a symmetric position and a center position instructed by the instruction unit;
An image output unit for outputting an image transformed by the image transformation unit;
An image processing apparatus. An image input unit for inputting an image of the processing object;
A display unit for displaying an image input by the image input unit and displaying an instruction of an operator;
A condition recording unit for recording an established value of the processing object;
An instruction unit for instructing a symmetrical position, a center position, and a size index of the processing object in the image input by the image input unit;
The image input by the image input unit is transformed using the symmetrical position and the center position specified by the instruction unit, and the ready-made value recorded in the condition recording unit and the size specified by the instruction unit An image transformation unit that transforms an image using an index;
An image output unit for outputting an image transformed by the image transformation unit;
An image processing apparatus. An image input unit for inputting an image of the processing object;
A display unit for displaying an image input by the image input unit and displaying an instruction of an operator;
A condition recording unit for recording an established value of the processing object;
An instruction unit for instructing a symmetrical position and a central position, a size index, and an unused portion of the processing object in the image input by the image input unit;
The image input by the image input unit is subjected to image deformation using the symmetrical position and the center position specified by the instruction unit, and the pre-set value recorded in the condition recording unit is obtained for the image obtained thereby. An image deforming unit that deforms an image using the size index instructed by the instruction unit, and further supplements an unused portion instructed by the instruction unit with an image of a symmetric part;
An image output unit for outputting an image transformed by the image transformation unit;
An image processing apparatus.   The image processing apparatus according to claim 10, wherein the image output unit outputs the non-use part information instructed by the instruction unit together with the pre-replenishment image.   The image deforming unit deforms an image for each region divided by a line segment connecting the symmetrical position and the central position when there are a plurality of pairs of the symmetrical position and the central position specified by the instruction unit. Item 12. The image processing device according to any one of Items 11. An image input step for inputting an image of the processing object;
A display step of displaying an image input in the image input step and displaying an instruction of an operator;
An instruction step for instructing the symmetrical position and the center position of the processing object in the image input in the image input step according to an instruction of the operator;
An image deformation step of deforming the image input in the image input step using the symmetrical position and the center position instructed in the instruction step;
An image output step for outputting the image transformed in the image transformation step;
An image processing method comprising: An image input step for inputting an image of the processing object;
A display step of displaying an image input in the image input step and displaying an instruction of an operator;
A condition recording step for recording an established value of the processing object;
An instruction step for instructing a symmetric position, a center position and a size index of the processing object in the image input in the image input step;
The image input in the image input step is deformed using the symmetrical position and the center position specified in the instruction step, and the ready-made value recorded in the condition recording step and the size index specified in the instruction step An image transformation step for transforming the image using
An image output step for outputting the image transformed in the image transformation step;
An image processing method comprising: An image input step for inputting an image of the processing object;
A display step of displaying an image input in the image input step and displaying an instruction of an operator;
A condition recording step for recording an established value of the processing object;
An instruction step for instructing a symmetrical position and a center position of the processing object in the image input in the image input step, a size index, and an unused portion;
The image input in the image input step is subjected to image deformation using the symmetrical position and the center position specified in the instruction step, and for the image obtained thereby, the pre-formed value recorded in the condition recording step and the Image deformation using the size index indicated in the instruction step, and further, an image deformation step of replenishing the unused portion indicated in the instruction step with an image of a symmetric part;
An image output step for outputting the image transformed in the image transformation step;
An image processing method comprising:   The image processing method according to claim 15, wherein in the image output step, the non-use part information instructed in the instruction step is output together with the pre-replenishment image.   14. The image transformation step, wherein when there are a plurality of pairs of the symmetric position and the central position instructed in the instruction step, the image is transformed for each region divided by a line segment connecting the symmetric position and the central position. Item 17. The image processing method according to any one of Items 16.   An image processing program for causing a computer to function as the image processing apparatus according to any one of claims 8 to 12.   A recording medium storing the image processing program according to claim 18.

Images