JP2009003842A - Image processor and image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processor having a labeling function for surely extracting a specific object in a short time from an image, without causing an overflow, even when having a stack memory of small capacity. <P>SOLUTION: This image processor has a reduction processing part 20 for generating a reduction image by reducing an original image at a reduction rate determined according to an expected pixel size of the specific object included in the original image, a reduction image labeling processing part 24 for extracting a specific object area included in the reduction image by attaching the same label to a plurality of continuous pixels having the same attribute among the pixels constituting the reduction image, and an object identifying processing part 26 identifying an object as the specific object included in the original image by enlarging the specific object area extracted by the reduction image labeling processing part 24 into a size of the original image. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention provides an image processing apparatus or image having a labeling processing unit that extracts a region of a specific object included in an image by attaching the same label to a plurality of continuous pixels having the same attribute among pixels constituting the image. More particularly, the present invention relates to an image processing apparatus and an image processing method for extracting characteristic information of discolored parts of facial skin such as stains, pores, wrinkles, and the like based on human face image information captured by an imaging apparatus. .

  In cosmetic surgery and esthetic salons, an imaging device is used to objectively determine how much the skin condition has improved before and after cosmetic treatments that improve the skin condition of the face, such as stains, pore dirt, wrinkles, etc. Thus, it is necessary to take a human face image, perform predetermined image processing on the taken face image, and evaluate the state of stain, pore dirt, wrinkles, and the like.

  Therefore, in Patent Document 1, at least one processing region is set in the captured image data of the entire face captured by the digital camera, the processing region image data is cut out, and the cut out processing region image data is processed to process the skin. In a digital zoom skin diagnostic device that calculates parameters, a high-resolution image of the entire face with a large output data size is taken with a digital camera with a large number of pixels of the image sensor, and analysis lines that can be arbitrarily arranged on the image are superimposed Along with the enlarged image of the processing area, a graph of the concentration distribution along this analysis line is displayed on the monitor screen, and the skin parameters are determined by counting the unevenness of the concentration distribution based on the threshold value set on this graph. There has been proposed a digital zoom skin diagnostic apparatus characterized by the demand.

  In the above-described apparatus, an enlarged image cut out from the photographed image data is used as a processing region, an XY two-dimensional density distribution is created from the R image data in the processing region, and a threshold is changed to binarize and extract a stain and extract. The stain parameters were calculated from the size and number of stains.

When calculating the size of the stain, the same label is attached to a plurality of continuous pixels having the same attribute among the pixels constituting the image, thereby extracting the region of the object included in the image, here the stain region. A labeling process is generally used.
Japanese Patent No. 3530523

  However, if a labeling process accompanied by a recursive call process that calls the procedure itself again is executed in the procedure, the processing time becomes extremely long, and the number of recursive call processes increases as the pixel size of the object increases. For this reason, if a large-capacity stack memory is not prepared, an overflow occurs and processing is interrupted, resulting in a problem that the stability of the system is impaired.

  Therefore, if the number of recursive call processes is limited, the occurrence of overflow can be avoided even with a stack memory with a limited capacity, but the problem is that the size of the area of a specific object cannot be detected accurately. Arise.

  Further, when the search range of the object is limited, an object exceeding the range cannot be detected, and the same object is counted as a plurality of objects.

  Such a problem appears not only in the case of detecting a region size such as a stain from a face image, but also in various situations such as when a specific object is extracted from a photographed frame image by a labeling process.

  The present invention has been made in view of the above-described conventional problems. Even when a stack memory having a small capacity is provided, the occurrence of overflow is avoided, and a specific object is reliably extracted from an image in a short time. An object of the present invention is to provide an image processing apparatus and an image processing method capable of labeling.

  In order to achieve the above object, the characteristic configuration of the image processing apparatus according to the present invention is determined according to the assumed pixel size of the specific object included in the original image, as described in claim 1 of the document of the claims. A reduction processing unit that generates a reduced image obtained by reducing the original image at a reduced rate, and a plurality of consecutive pixels having the same attribute among the pixels constituting the reduced image, by attaching the same label. A reduced image labeling processing unit that extracts a region of a specific object included in an image, and a region of the specific object extracted by the reduced image labeling processing unit is enlarged to the size of the original image, and is included in the original image An object identification processing unit for identifying a specific object is provided.

  According to the above-described configuration, even if the specific object included in the original image is a large area, the reduced image is generated by the reduction processing unit with the reduction rate determined according to the assumed pixel size of the specific object. Since the labeling process is executed by the reduced image labeling processing unit, the number of target pixels for the labeling process decreases. As a result, the reduced image labeling processing unit can execute the labeling process at high speed without causing an overflow of the stack memory. The specific object extracted from the reduced image by the reduced image labeling processing unit is enlarged by the object identification processing unit and identified as a specific object included in the original image.

  In the second feature configuration, as described in claim 2, in addition to the first feature configuration described above, the reduced image labeling processing unit uses a reduced image generated by the reduction processing unit and having a large reduction rate. The labeling process is executed in the order of the reduced images with the smaller reduction ratios, and the labeling process is executed except for the specific object extracted from the reduced images with the higher reduction ratio and the surrounding area when the reduced images with the lower reduction ratios are labeled. It is in.

  In the reduction processing unit, the reduction ratio corresponding to the assumed pixel size of the specific object included in the original image is determined. However, the actual pixel size of the specific object varies, and if the reduction ratio is too large, the reduction image labeling process is performed. There is a possibility that the specific object cannot be extracted by the part.

  Therefore, it may be necessary to perform labeling processing on a plurality of reduced images having different reduction ratios by the reduced image labeling processing unit. In such a case, the labeling process is performed in the order of a reduced image with a large reduction ratio and a reduced image with a low reduction ratio, and the reduction ratio is reduced except for a specific object extracted in the reduced image with a large reduction ratio and its surrounding area. If labeling processing is performed on a small reduced image, duplicate labeling processing on the same specific object can be omitted, the specific object can be extracted reliably, and the labeling processing can be executed at high speed. .

  In the third feature configuration, in addition to the first or second feature configuration described above, the specific feature is defined with respect to the contour region of the specific object identified by the object identification processing unit. An outline region labeling processing unit that extracts the region of the specific object included in the original image by attaching the same label to a plurality of continuous pixels having the same attribute as the object is provided.

  When the specific object extracted by the labeling process for the reduced image is enlarged, the outline of the specific object existing in the original image may not be extracted accurately due to the influence of the missing pixel at the time of reduction. In such a case, if the contour region labeling processing unit performs labeling processing on the contour region of the specific object identified by the object identification processing unit, the contour of the specific object in the original image can be accurately extracted. In this case, since the labeling processing target pixels of the contour region labeling processing unit are limited to the contour region of the specific object identified by the object identification processing unit, the contour region labeling processing unit does not cause an overflow of the stack memory, Labeling processing can be executed at high speed.

  In the fourth feature configuration, in addition to any one of the first to third feature configurations described above, a plurality of pixels having the same attribute among the pixels constituting the original image may be provided. An original image labeling processing unit for extracting a region of a specific object included in the original image by attaching the same label to continuous pixels, and the reduction processing based on an assumed pixel size of the specific object included in the original image And a switching processing unit that switches between operating the reduced image labeling processing unit and the object identification processing unit or operating the original image labeling processing unit.

  The labeling process needs to be performed on the reduced image when the number of pixels of the specific object included in the original image is large and there is a possibility that inconvenience such as overflow of the stack memory may occur. Therefore, when there is no possibility that such an inconvenience will occur, a specific object can be accurately extracted from the original image by operating the original image labeling processing unit by the switching processing unit. When there is a possibility of occurrence, the switching processing unit operates the reduction processing unit, the reduced image labeling processing unit, and the object identification processing unit, so that a specific object can be extracted stably and at high speed.

  In the fifth feature configuration, in addition to any one of the first to third feature configurations described above, the specific object is a stain, pore, It is in a point that is a discoloration region of skin such as wrinkles.

  When a skin discoloration region such as a stain, pore, or wrinkle included in a human face image is used as a specific object, an image processing apparatus having any one of the first to third characteristic configurations functions suitably.

  In the sixth feature configuration, as described in claim 6, in addition to the fourth feature configuration described above, the specific object is a skin discoloration region such as a stain, pore, or wrinkle included in a human face image. The switching processing unit operates the original image labeling processing unit when the pore is a specific object, and the reduction processing unit, the reduced image labeling processing unit, and the object identification processing when the stain is a specific object. It is in the point which switches so that a part may be operated.

  According to the first characteristic configuration of the image processing method of the present invention, as described in claim 7, the original image is reduced at a reduction rate determined in accordance with an assumed pixel size of a specific object included in the original image. A reduction processing step for generating a reduced image, and extracting a region of a specific object included in the reduced image by attaching the same label to a plurality of continuous pixels having the same attribute among the pixels constituting the reduced image. A reduced image labeling processing step; and an object identification processing step for enlarging the region of the specific object extracted in the reduced image labeling processing step to the size of the original image to identify a specific object included in the original image. There is in point.

  In the second feature configuration, as described in claim 8, in addition to the first feature configuration described above, the reduced image labeling processing step includes a reduction image having a large reduction ratio generated in the reduction processing step. The labeling process is executed in the order of the reduced images with the smaller reduction ratios, and the labeling process is executed except for the specific object extracted from the reduced images with the higher reduction ratio and the surrounding area when the reduced images with the lower reduction ratios are labeled. It is in.

  According to the third feature configuration, in addition to the first or second feature configuration described above, the specific feature is defined with respect to the contour region of the specific object identified in the object identification processing step. An outline region labeling process step for extracting a region of a specific object included in the original image by attaching the same label to a plurality of continuous pixels having the same attribute as the object is provided.

  As described above, according to the present invention, even when a stack memory having a small capacity is provided, the occurrence of overflow can be avoided, and a labeling process that can reliably extract a specific object from an image in a short time is possible. An image processing apparatus and an image processing method can be provided.

  Embodiments of an image diagnosis system incorporating an image processing apparatus according to the present invention will be described below.

  As shown in FIG. 1, the diagnostic imaging system includes an imaging device 1 that captures a client's face image, and a hard disk that stores the client's facial image that has been captured by the imaging device 1 and converted into electronic data. And a display unit 3 having a liquid crystal screen, and an image processing apparatus 4 in which an application program embodying the present invention is installed in a personal computer.

  The imaging device 1 includes a face holding unit 1a having a chin guide and a head guide for keeping the client's face in a fixed posture, an illumination device 1b for illuminating the client's face held by the face holding unit 1a, An imaging unit 1c that captures the face of the client illuminated by the apparatus 1b is provided.

  The image processing device 4 is generated by an analysis target region generation unit 4a that defines an analysis target element, that is, an analysis target region, in the client's face image region, and a pixel group or analysis target region generation unit 4a that constitutes the entire face image. A skin characteristic extraction processing unit 4b that extracts characteristic information that is a discolored part of the skin such as a stain, pore dirt, wrinkles, and the like based on the pixel values of the pixel group that constitutes the analysis target element, and a skin characteristic extraction processing unit 4b A diagnostic processing unit 4c for diagnosing the skin condition based on the extracted skin characteristic information is provided.

  The client's face image captured by the imaging device 1 is stored as an original image in the memory 2, read from the memory 2 by the image processing device 4, and displayed on the display unit 3. The analysis target region generation unit 4a defines an analysis target element R, that is, an analysis target region as shown in FIG. 2 on the skin of the client's face image displayed on the display unit 3. Specifically, as the analysis target element R, a forehead region R1, a cheek region R2, a nose region R3, and a chin region R4 are demarcated. be able to.

  The skin characteristic extraction processing unit 4b specifies a predetermined range of skin color areas from the constituent pixel data of the analysis target element defined in the face image area, and the blue component values of a plurality of pixels constituting the skin color area are larger than a predetermined threshold value. Or small, using a labeling method, and extracting areas where facial stains d1, pore stains d2, wrinkles d3, etc. as shown in FIG. 3B are extracted.

  As will be described in detail later, as shown in FIGS. 7 (a) to 7 (c), the labeling process is an analysis target with thresholds appropriately set for extraction target objects such as stains, pore dirt, and wrinkles. After the pixels in the element R are binarized, the analysis target element R is divided into a plurality of regions Ra, Rb, Rc, Rd larger than the maximum size of the extraction target object, and within the divided regions Ra, Rb, Rc, Rd A specific label Tn (n is a positive integer and is the same in the same region) for any pixel having a pixel value below the threshold (here, set to 1 in the binarization process and shown in gray in the figure) Is repeated, and the process of attaching the same label Tn to the pixels adjacent to the pixel to which the label Tn is attached, which is similarly below the threshold, is repeated until there are no adjacent pixels below the threshold. Call processing , By performing labeling processing for all the pixels in the analyzed elements R, stains, pores dirt, single or a plurality of areas corresponding to the wrinkles are extracted.

  As a result, in the example shown in FIG. 7, the object indicated by the label T1 is extracted from the region Ra, the object indicated by the label T2 is extracted from the region Rb, the object indicated by the label T3 is extracted from the region Rc, and the region An object indicated by a label T4 is extracted from Rd. In the region between the regions (the regions indicated by labels T3 and T4), the continuity of the object existing at the boundary of the divided region is subsequently determined based on the binarized pixel values, and the continuous If they do, they are integrated as the same area.

  During the labeling process, an algorithm that repeatedly determines whether or not to attach a specific label to each target pixel is repeatedly executed. The processing is executed while a plurality of necessary parameters are stacked in a stack area defined in the memory area of the personal computer.

  Based on the area (number of pixels), shape, pixel value, etc. of the region extracted by the skin characteristic extraction processing unit 4b, the diagnosis processing unit 4c determines whether the stain is a stain, pore dirt, or wrinkle, and performs analysis Characteristic information including each position, area in the target element, an area ratio with respect to the analysis target element R, a pixel value, and the like is generated.

  At this time, by changing the display color of the discolored part extracted by the skin characteristic extraction processing unit 4b from the face image displayed on the display unit 3 to a color distinguishable from the surrounding skin region, the visual image is displayed. Can be easily recognized.

  For example, in a face image of 8 million pixels, an area of about 3 × 3 pixels to 7 × 7 pixels corresponds to a pore, so that an area with a larger number of pixels can be determined as a stain or a wrinkle, and the areas in the orthogonal direction are continuous. Of the pixel values, if the smaller continuous pixel value is several pixels to a dozen pixels and the smaller continuous pixel value and the larger continuous pixel value are equal to or greater than a predetermined value, it can be determined that it is a habit.

  In general, when extracting the above-described skin characteristics, it is preferable to employ the value of the blue component among the red, green, and blue color components constituting the pixel. This is because the condition of the deep part of the skin appears in the blue component. Therefore, it is preferable to use a blue transmission filter or an ultraviolet lamp as the light source of the illumination device 1b.

  The analysis target element R defined by the analysis target region generation unit 4a, the skin characteristic information extracted by the skin characteristic extraction processing unit 4b, and the diagnosis information diagnosed by the diagnosis processing unit 4c are IDs for identifying the client. It is stored in the memory 2 in association with data, diagnosis date, treatment contents, and the like.

  The diagnostic processing unit 4c generates diagnostic information as to whether or not the skin condition has been improved based on the client's past skin characteristic information and current characteristic information stored in the memory 2. A doctor, an esthetician, or a beauty counselor determines what treatment or prescription is preferable or what cosmetic is preferable based on the diagnosis information and appropriately responds.

  As illustrated in FIG. 1B, the analysis target region generation unit 4a includes the arrangement information of the standard face elements and the arrangement information of the standard analysis target elements constituting the face obtained by statistically processing a plurality of sample face images. And the analysis target face image B (see FIG. 3B) photographed by the imaging device and displayed on the display unit 3 in a superimposed manner (see FIG. 3). 3 (c)) Standard face element movement for moving the corresponding specific standard face element e to the display position of the face image display processing unit 10 and the specific face element E of the analysis target face image B displayed on the display unit 3 The analysis target stored in the storage unit 2 in association with the analysis target face image B as the analysis target element R for the analysis target face image B, the processing unit 12 and the standard analysis target element r moved by the standard face element movement processing unit 12 An element generation processing unit 18 is provided.

  Further, the standard face element movement processing unit 12 extracts another standard face element e ′ and standard analysis target element r located in a predetermined range S centered on the specific standard face element e, and a movement element extraction processing unit 14 that extracts the standard analysis target element r. The other standard face element e ′ and the standard analysis target element r extracted by the movement element extraction processing unit 14 are moved along the movement direction calculated based on a predetermined movement function F along with the movement of the specific standard face element e, and An element movement processing unit 16 that moves the movement distance is provided.

  The arrangement information of the standard face element e and the standard analysis target element r is defined by a plurality of reference points p and a connection line q connecting the reference points p, and the movement function F is a reference before the movement of the specific standard face element e. When ΔL ≦ L1 with respect to the inter-element distance L1 connecting the point p1 and another standard face element e ′ or the reference point p of the standard analysis target element r and the movement distance ΔL of the reference point of the specific standard face element e The movement distance L of the reference point p of the other standard face element e ′ is derived using ΔL / L1 as a variable, and the reference point p2 after movement of the specific standard face element e and another standard face element e ′ or standard analysis When the reference point p2 after the movement of the specific standard face element e moves away from the other standard face element e 'or the reference point p of the standard analysis target element r on a straight line connecting the reference point of the target element r , The reference point p2 after the movement of the specific standard face element e is another standard face element e 'or a standard solution It is defined as a function to be separated when approaching the reference point p of the analysis target element r.

  A standard face image A shown in FIG. 3A is obtained by comparing the size of face elements such as contours, hairline, eyebrows, eyes, ears, nose, and mouth constituting a face with respect to a large number of sample face images collected in advance. Relative distance is measured, and the result is statistically processed to calculate the size and position of standard face elements. Based on the calculated values, a standard face image with standard face elements arranged is generated. An image obtained by demarcating an analysis target element r, that is, an analysis target area, in which an operator manually operates a pointing device such as a mouse to search for a state of characteristics such as discoloration of the skin. This is an image defined by the face information such as the outline, the hairline, the eyebrows, the eyes, the ears, the nose and the mouth, that is, the placement information of the standard face element e and the placement information of the standard analysis target element r.

  Here, the arrangement information is composed of coordinates where the reference point p is located in a predetermined X, Y two-dimensional coordinate system and attribute information of the connection line q. The attribute information includes reference points p at both ends of the connection line q. Reference point specifying information to be specified, line type information to specify whether the connecting line is a straight line or a curve, and information such as curvature in the case of a curve are included.

  The standard face image A and the analysis target face image B stored in advance in the memory 2 are read out by the face image display processing unit 10 and displayed superimposed on the display unit 3 as shown in FIG. It is.

  The arrangement information of the standard face element e and the standard analysis target element r is defined by a plurality of reference points p and a connection line q connecting the reference points p as shown in FIG. The standard face element movement processing unit 12 moves the reference point p or the connection line q while being dragged with the mouse by the operator, and when the drop operation is performed, moves the reference point p or the connection line q to that position. .

  As shown in FIG. 4B, when the reference point p is moved, only the reference point p is moved, and the reference point p that is not moved is connected to the moved reference point p. When the line q is changed and the connection line q is moved as shown in FIG. 4C, the connection line q and all the reference points p connected by the connection line q are maintained in a positional relationship. Move with.

  At this time, as shown in FIG. 4D, the moving element extraction processing unit 14 extracts other standard face elements e ′ and standard analysis target elements r located in a predetermined range S centering on the specific standard face element e. As a specific standard face element e, it is extracted as an object to be moved at the same time. The predetermined range S is the movement distance of the specific standard face element e, that is, the movement distance ΔL of the reference point P when only one reference point is moved, and the movement distance ΔL of the connection line when the connection line q is moved. For example, it is set to a circular area whose radius is in the range of N × ΔL with the position of the reference point before movement or the center position of the connection line q as the center.

  Moreover, the length of one side centering on the position of the reference point P before the movement or the center position of the connection line q may be set to a rectangular area of N × ΔL. Here, N is a natural number, and typically 2 to 3 is selected, but may be appropriately changed depending on the number of pixels of the target image.

  That is, the predetermined range S described above is variably set based on the movement distance ΔL of the specific standard face element e. Therefore, if the movement distance ΔL of the specific standard face element e is long, the number of other standard face elements e ′ and standard analysis target elements r that are affected increases, and if the movement distance ΔL of the specific standard face element e is short, the influence is increased. Other standard face elements e ′ and standard analysis target elements r received are reduced. For simplicity, the predetermined range S described above can be set as a fixed range.

  The element movement processing unit 16 converts the other standard face element e ′ and the standard analysis target element r extracted by the movement element extraction processing unit 14 based on the movement function F described above with the movement of the specific standard face element e. The moving direction and the moving distance calculated as described above are moved.

  FIG. 5 is a diagram for explaining the characteristics of the transfer function F with respect to the eyebrows and eyes (closed state) of the standard face element and the nose region of the standard analysis target element. As shown in FIG. 5 (a), the movement function F is determined by the reference point p1 before the movement of the specific standard face element e and the reference point P (or the center position of the connecting line) of the other standard face element e ', or The inter-element distance L1 connecting the reference point p1 before the movement of the specific standard face element e and the reference point p of the standard analysis target element r (or the center position of the connecting line) and the movement distance ΔL of the reference point of the specific standard face element e On the other hand, when ΔL ≦ L1, the movement distance L of the reference point p of the other standard face element e ′ or the reference point p of the standard analysis target element r is expressed as L = a × (ΔL / L1) + b (a , B are constants) on the straight line connecting the reference point p2 after the movement of the specific standard face element e and the reference point p of the other standard face element e ′ or the standard analysis target element r, Is the reference point p2 after the movement of the specific standard face element e is the reference point p of another standard face element e 'or the standard analysis target element r? It is defined as a function that allows the reference point p2 after the movement of the specific standard face element e to move away from the reference point p2 and move away from the reference point p of the other standard face element e ′ or the standard analysis target element r. Yes. In the present embodiment, a = 1 and b = 0 are set, but may be set as appropriate based on the number of pixels of the analysis target image. The movement distance L of the reference point p is calculated for each X component and Y component in the X, Y two-dimensional coordinate system.

  Further, the movement function F sets the movement distance L of the reference point p of the other standard face element e ′ to ΔL when ΔL> L1, and the reference point p2 after the movement of the specific standard face element e and other reference face p2 On the straight line connecting the standard face element e ′ or the reference point of the standard analysis target element r, the reference point p2 after the movement of the specific standard face element e is the reference point of another standard face element e ′ or the standard analysis target element r. It is defined as a function that is approached when moving away from p, and moved away when the reference point p2 after the movement of the specific standard face element e approaches another reference face element e 'or the reference point p of the standard analysis target element r. ing. That is, as shown in FIG. 5B, when ΔL> L1, the element movement processing unit 16 sets the reference point p1 before the movement of the specific standard face element e and the reference point p of another standard face element e ′. Move while keeping the separation distance.

  Here, when ΔL ≦ L1, if the movement amount ΔL of the reference point p or the connection line q is a minute value, only the reference point p or the connection line q is moved, and the other standard face elements e ′ are moved. The movement function F may be defined so that the position of the reference point p or the reference point p of the standard analysis target element r is maintained as it is.

  For example, a positive integer K is set in advance with respect to the total number of pixels of the face image, and the reference point p of the other standard face element e ′ positioned closest to the moved reference point p or the connection line q or the standard analysis target When there is a relationship of L1 / ΔL ≧ K (K is a positive integer) between the inter-element distance L and the movement distance ΔL with respect to the reference point p of the element r, the reference point p of the other standard face element e ′ or the standard analysis target element The movement function is defined so as to maintain the original position as it is without moving the position of the reference point p of r.

  As shown in FIG. 6, the standard face element movement processing unit 12 moves each standard face element e of the standard face image A to the position of the corresponding face element E of the analysis target face image B. The standard analysis target element r that has been moved along with the movement process of the face element e is defined as the analysis target element R of the analysis target face image B, that is, the analysis target area.

  In addition to the analysis target area defining process for the analysis target face image B described above, a drawing processing unit is provided in the analysis target area generating unit 4a, and a plurality of points input by an operator using a mouse are connected by the drawing processing unit. The analysis target area may be defined by connecting with a line.

  In this way, an appropriate analysis target element R defined in the analysis target face image B is stored in the memory 2 in association with ID data, diagnosis date, treatment contents, and the like that specify the client.

  When the client's face is newly photographed thereafter, the analysis target element R stored in the memory 2 is read, and the analysis target element R is defined in the client's new face image. Is done. Needless to say, the number of pixels, magnification, and reference coordinates of the face image need to be matched with the magnification and reference coordinates of the analysis target element R.

  Hereinafter, the configuration and operation of the skin characteristic extraction processing unit 4b in which the image processing apparatus or the image processing method according to the present invention is employed will be described in detail.

  As shown in FIG. 1B, the skin characteristic extraction processing unit 4b generates a reduced image obtained by reducing the original image at a reduction rate determined in accordance with the assumed pixel size of the specific object included in the original image. A reduction processing unit 20, a reduction image labeling processing unit 22 that extracts a region of a specific object included in the reduction image by attaching the same label to a plurality of continuous pixels having the same attribute among the pixels constituting the reduction image; The object identification processing unit 24 that enlarges the area of the specific object extracted by the reduced image labeling processing unit 22 to the size of the original image and identifies it as a specific object included in the original image, and the object identification processing unit 24 identifies The same label is attached to a plurality of continuous pixels having the same attribute as the specific object with respect to the contour region of the specific object. , And a contour region labeling processing unit 26 for extracting a region of a specific object included in the original image.

  Further, the skin characteristic extraction processing unit 4b attaches the same label to a plurality of continuous pixels having the same attribute among the pixels constituting the original image, thereby extracting an original image labeling that extracts a specific object region included in the original image. A processing unit 28 is provided, and the reduction processing unit 20, the reduced image labeling processing unit 22, the object identification processing unit 24, and the contour region labeling processing unit 26 are sequentially operated based on the assumed pixel size of the specific object included in the original image. A switching processing unit 29 for switching whether to operate the original image labeling processing unit 28 is provided.

  Specifically, the switching processing unit 29 operates the original image labeling processing unit 28 when a pore or wrinkle having a small number of pixels is set as a specific object, and reduces when a stain or the like having a large number of pixels is set as a specific object. The processing unit 20, the reduced image labeling processing unit 22, the object identification processing unit 24, and the outline region labeling processing unit 26 are switched to operate in order.

  When the original image labeling processing unit 28 sets a pore as a specific object, the pixel having a pixel value lower than the threshold value of the blue component for extracting the pore from the skin region of the original image B1 as the analysis target face image B is set to 1. After performing binarization processing for setting pixels having pixel values equal to or greater than the threshold value to 0, for example, as shown in FIG. 8A, each pixel is divided into pixel sizes of about 10 × 10 pixels. A labeling process is performed on the pixel group. As a result, a pore region d2 of about 3 × 3 pixels to 7 × 7 pixels is extracted.

  However, as shown in FIG. 8B, when a specific area is a stain area d1 that may cover several hundred pixels × several hundred pixels, the original face image B as the analysis target face image B as shown in FIG. The entire image B1 or the skin area in each analysis target element R is subjected to a binarization process based on a blue component threshold value for extracting a stain, and then divided into a pixel size of several hundred pixels × several hundred pixels. When the labeling process is executed for each divided pixel group, the stack may overflow and the process may be interrupted.

  Therefore, the reduction processing unit 20 is activated by the switching processing unit 29. When the stain included in the original image B1 as the analysis target face image B is a specific object, the reduction processing unit 20 assumes, for example, a relatively large stain size of 10,000 pixels (100 × 100 pixels) as an assumed pixel size. As a result, a reduced image B2 in which the entire original image B1 or the analysis target element R is reduced to 100 pixels (10 × 10 pixels) that is 1/100 of that is generated. That is, the reduction ratio is set to 1/10. As the reduction processing algorithm at this time, it is sufficient to use the nearest neighbor method, and the reduction processing can be performed at high speed. Note that a reduction / enlargement algorithm other than the nearest neighbor method may be employed.

  For example, a reduced image B2 of about 230 × 350 pixels as shown in FIG. 9B is converted to an original image B1 of 8 million pixels made up of about 2300 × 3500 pixels as shown in FIG. 9A. Generate.

  The reduced image labeling processing unit 22 performs binarization processing on the pixels of the reduced image B2 based on the threshold value of the blue component for extracting the stain, and then, as shown in FIG. Included in the reduced image by dividing the pixel size of the pixel and assigning the same label T2 to a plurality of continuous pixels having the same attribute for each divided pixel group, that is, a binarized pixel having a value of 1. The region U2 of the specific object to be extracted is extracted.

  In other words, by reducing the number of pixels in the division area to be labeled to 1/100, if the original image B1, it is necessary to execute the labeling process in units of 100 × 100 pixels in the reduced image B2. If so, the labeling process can be executed in units of 10 × 10 pixels, and the occurrence of stack overflow can be avoided without increasing the memory capacity. Further, since the number of target pixels for the labeling process is reduced, the labeling process can be executed at high speed.

  As shown in FIG. 10B, the object identification processing unit 24 enlarges the stain area extracted by the reduced image labeling processing unit 22 by a factor of 100 based on the reduction ratio (1/100). By associating with the coordinates of the stain area in the original image B1, the specific object U1 included in the original image B1, that is, the stain area is identified.

  When the assumed pixel size of the specific object in the original image is set to 10,000 pixels, if the reduction ratio is set to 1/100, the size of the specific object is 1 pixel in the reduced image, and the labeling process can be performed at a higher speed, but there is an error as much. Increase. If the reduction ratio is set to ½, the size of the specific object in the reduced image is 2500 pixels, and the labeling process can be performed without any significant decrease in accuracy, but the stack capacity becomes a problem. Therefore, the reduction ratio needs to be determined in terms of the accuracy of the labeling process and the capacity of the stack.

  The specific object U1 of the original image B1 thus identified may not be able to accurately extract the contour of the specific object existing in the original image due to the effect of missing pixels at the time of reduction. There is a risk of inconvenience.

  Therefore, the contour region labeling processing unit 26 attaches the same label to a plurality of continuous pixels having the same attribute as the specific object U1 with respect to the contour region V of the specific object U1 identified by the object identification processing unit 24. The area of the specific object included in the original image is extracted.

  As shown in FIG. 10C, for example, a range of 10 pixels is set as a contour region V around the edge pixel of the identified specific object U <b> 1 on the left and right and up and down. The number of pixels for setting the contour region V is a value determined in accordance with the reduction ratio. In this embodiment, the reduction ratio is 1/100. Therefore, it is considered that rounding errors of 10 pixels occur vertically and horizontally. is there.

  The contour region labeling processing unit 26 divides such a contour region V into a plurality of regions of 20 × 20 pixels so that an arbitrary edge pixel of the specific object U1 identified by the object identification processing unit 24 is located at the center. Then, the labeling process is executed for each divided region. In this way, an accurate specific object U as shown in FIG. 10 (d) is detected. At this time, by performing the labeling process from the pixel farthest from the edge pixel toward the specific object U1, only the pixels that are continuous with the specific object U1 can be labeled accurately and at high speed.

  Note that the contour region labeling process may be configured to be executed as necessary based on a selection operation by an operator as a selective process.

  An example of an image processing method for extracting a stain by the skin characteristic extraction processing unit 4b described above will be outlined based on the flowchart shown in FIG.

  As shown in FIG. 12, the display area 40 of the monitor image B3 obtained by thinning the entire original image on which the analysis target element R is superimposed on the screen of the display unit 3, the display area 41 of the reduced image B2, and the analysis target A display area 42 for displaying an image obtained by enlarging a part of the element R is set, and a diagnostic report display area 39 is arranged on the right side thereof.

  Below the display area 40 of the monitor image B3, keys 30, 31, and 32 for selecting one of stain, wrinkle, and pores are displayed as specific objects, and keys 33, 34, 35 is arranged. A start key 36 for starting the specific object extraction process, a cancel key 37 for canceling the operation input immediately before, a print key 38 for printing the display content of the screen, and the like are arranged at the lower center of the screen.

  When the operator operates the key 30 corresponding to the stain, the switching processing unit 29 activates the reduction processing unit 20, and a message that the operator should operate the reduction ratio setting key in the message display field 43 at the top of the screen. Is displayed, and the automatic reduction key 33 or manual setting key 34 for the reduction ratio is activated. When the automatic setting key 33 is selected, the reduction ratio is set to 1/10. When the manual setting key 34 is selected, a numerical value input from the keyboard is set as the reduction ratio and displayed on the reduction ratio display unit 35. .

  When the operator selects the analysis target element R (R1 to R4) displayed in the display area 40 of the monitor image B3 and operates the start key 36, the reduction processing unit 20 sets the original image B1. A reduction processing step is executed to display the reduced image B2 in the reduced image display area 41 by performing reduction processing at the reduction ratio (S1).

  The reduced image labeling processing unit 22 performs the above-described labeling processing on the reduced image region corresponding to the selected analysis target element R, and performs a reduced image labeling processing step of extracting the stain region U2 included in the reduced image B2. Execute (S2). The result is reflected in the reduced image B2 displayed in the reduced image display area 41, and the corresponding area is displayed in a different color from the others.

  The object identification processing unit 24 enlarges the stain area extracted in the reduced image labeling processing step to the size of the original image, and executes an object identification processing step that identifies the stain as included in the original image. The identified stain area is displayed so as to be identifiable with a color different from the others, and the extracted stain and its surrounding enlarged image are displayed in the enlarged image display area 42 (S3).

  The outline region labeling processing unit 26 applies the same label to a plurality of continuous pixels having the same attribute as the stain for the outline region of the stain identified in the object identification processing step (see reference numeral V in FIG. 10C). The contour region labeling processing step for extracting the stain area included in the original image is executed, and the stain area included in the original image is displayed in a different color from the others (S4). At this time, the contour region V may be appropriately divided into a plurality of regions, the contour region labeling process may be performed on each divided region, and then the same region may be integrated.

  As a result, the state of the extracted stain area is diagnosed by the diagnosis processing unit 4c, and the diagnosis result is displayed in the display area 39 as a diagnosis report.

  Another embodiment will be described below. In the above-described embodiment, the case where the size of the stain that is the specific object is assumed has been described. However, when the size of the stain is large or small, the reduction processing unit 20 generates a plurality of reduced images with different reduction rates, The reduced image labeling processing unit 22 may perform a labeling process for each reduced image.

  At this time, the reduced image labeling processing unit 22 executes the labeling process in the order of the reduced image having the larger reduction rate generated from the reduction processing unit 20 and the reduced image having the lower reduction rate, and performs the labeling process on the reduced image having the lower reduction rate. If the configuration is such that the labeling process is executed by excluding the specific object extracted from the reduced image having a large reduction ratio and its peripheral area, the duplicate labeling process for the same specific object can be omitted. The specific object can be extracted and the labeling process can be executed at high speed.

  In the above-described embodiment, the example in which the reduction processing unit 20 determines the reduction rate to be 1/10 assuming that the assumed pixel size of the stain that is the specific object included in the original image is 10000 pixels is described. The reduction ratio corresponding to the size is not limited to such a value, and may be set as appropriate according to the size of the stain to be detected. The same applies to the case where a specific object having a large size other than a stain is extracted.

  In the above-described embodiment, the reduction ratio is determined on the assumption that the stain that is the specific object has a substantially circular shape. However, when the stain or another specific object has another shape such as an elliptical shape. Therefore, the reduction rate may be determined in consideration of this. For example, in the case of an elliptical shape, the reduction ratio may be determined based on the minor axis. Also, the vertical and horizontal reduction ratios may be set to different values.

  In the case described above, the number of pixels in the vertical and horizontal directions of the divided area where the labeling process is performed can be set to a different value to efficiently process.

  An input processing unit that can variably set a reduction ratio that is automatically selected when an automatic setting key for reduction magnification is selected by an operator's setting operation via an input unit of a personal computer constituting the image processing apparatus. It is also possible to provide.

  The binarization threshold at the time of labeling processing is a value that is appropriately set depending on the target specific object, and is an input that can be variably set by an operator's setting operation via an input unit of a personal computer constituting the image processing apparatus. It is also possible to provide a processing unit. Further, an upper limit value and a lower limit value are set as the binarization threshold, and a pixel having a value defined by the upper limit value and the lower limit value is set to 1, and a pixel having a deviating value is set to 0. Also good.

  In the above-described embodiment, among the red, green, and blue color components constituting the pixel, the binarization based on the blue component has been described. However, depending on the characteristics of the specific object, other color components may be used. The image may be binarized based on the luminance, or may be binarized based on the luminance value. Further, it may be determined which of specific color components or luminance should be binarized in relation to the wavelength of the light source used when photographing the original image.

  In the above-described embodiment, the moving element extraction processing unit has been configured to extract other standard face elements and standard analysis target elements located in a predetermined range centered on a specific standard face element. The extraction processing unit may be configured to extract only the standard analysis target element associated with the specific standard face element. In this case, the analysis target element can be arranged more naturally and appropriately by configuring so as to extract the standard analysis target element located in a predetermined range centered on the specific standard face element.

  In this case, each standard face element and the standard analysis target element need to be associated in advance. For example, the eyebrow of the standard face element is associated with the standard analysis target element in the forehead region R1, and the eyes of the standard face element are nose The standard face element is associated with the standard analysis target element of the region R3 and the cheek region R2, the mouth of the standard face element is associated with the standard analysis target element of the jaw region R4, and each standard face element and the standard analysis target element are stored in the memory 2 together with the association information. You just have to.

  In addition, what is necessary is just to comprise the predetermined area | region which extracts a transfer function and a standard analysis object element similarly to the above-mentioned.

  In the above-described embodiment, an example in which the specific object is a skin discoloration region such as a stain, a pore, or a wrinkle included in a human face image is described. However, the specific object targeted by the image processing apparatus according to the present invention is as follows. The present invention is not limited to such a case, and the present invention can also be applied when extracting an arbitrary object included in an arbitrary image. For example, the present invention can be applied to a case where a human face existing in a photographic image is extracted.

  The embodiment described above is merely an example of the present invention, and it goes without saying that the specific configuration and the like of each functional block can be changed and designed as appropriate within the scope of the effects of the present invention.

(A) is a functional block diagram of an image diagnostic system, (b) is a block diagram of an image processing apparatus. Explanatory drawing of the analysis target element defined in the skin of the face image (A) is an explanatory diagram of a standard face image displayed on the display unit, (b) is an explanatory diagram of an analysis target face image displayed on the display unit, and (c) is a standard face displayed superimposed on the display unit. Illustration of the image and the face image to be analyzed (A) is an explanatory view of the arrangement information of the standard face element and the standard analysis target element, (b) is an explanatory view of the movement mode of the standard face element and the standard analysis target element when the reference point is moved by the operator. c) is an explanatory diagram of a movement mode of the standard face element and the standard analysis target element when the connection line is moved by the operator, and (d) is another standard face element and standard analysis target element that moves together with the specific standard face element. Explanatory drawing of the predetermined range where is extracted (A) is explanatory drawing which shows the movement distance and direction of the reference point of another standard face element when (DELTA) L <= L1, (b) is a reference | standard of another standard face element and a standard analysis object element when (DELTA) L> L1. Explanatory drawing showing the moving distance and direction of a point Explanatory drawing of the analysis target element defined in the analysis target face image Illustration of labeling process (A) is an explanatory diagram of a target pixel block for labeling processing when the specific object is a pore (black circle is a pixel for pores), and (b) is a target pixel block for labeling processing when the specific object is a stain (black circle is Illustration of pore pixel) Illustration of original image and reduced image (A) is an explanatory diagram of a stain area obtained by the labeling process on the reduced image, (b) is an explanatory diagram of a main part of the stain area identified by the object identification process, and (c) is set by the outline area labeling process. Explanatory drawing of the principal part of the outline area | region which shows, (d) is explanatory drawing of the principal part which shows the result of a contour area labeling process Flowchart explaining operation of skin characteristic extraction processing unit Illustration of the screen displayed on the display

Explanation of symbols

2: Storage unit 3: Display unit 4b: Skin characteristic extraction processing unit 20: Reduction processing unit 22: Reduced image labeling processing unit 24: Object identification processing unit 26: Contour region labeling processing unit 28: Original image labeling processing unit 29: Switching Processing part

Claims (9)

A reduction processing unit that generates a reduced image obtained by reducing the original image at a reduction rate determined in accordance with an assumed pixel size of a specific object included in the original image;
A reduced image labeling processing unit for extracting a region of a specific object included in the reduced image by attaching the same label to a plurality of continuous pixels having the same attribute among the pixels constituting the reduced image;
An image processing apparatus comprising: an object identification processing unit that enlarges a region of a specific object extracted by the reduced image labeling processing unit to a size of the original image and identifies the specific object as a specific object included in the original image.   The reduced image labeling processing unit executes a labeling process in the order of a reduced image with a large reduction rate and a reduced image with a low reduction rate generated by the reduction processing unit, and performs a reduction rate during a labeling process for a reduced image with a low reduction rate. The image processing apparatus according to claim 1, wherein the labeling process is executed except for a specific object extracted from a large reduced image and a peripheral area thereof.   By extracting the same label from a plurality of continuous pixels having the same attribute as the specific object, the specific object area included in the original image is extracted from the contour area of the specific object identified by the object identification processing unit. The image processing apparatus according to claim 1, further comprising a contour region labeling processing unit. An original image labeling processing unit for extracting a region of a specific object included in the original image by attaching the same label to a plurality of continuous pixels having the same attribute among pixels constituting the original image;
Switching between operating the reduction processing unit, the reduced image labeling processing unit, and the object identification processing unit or operating the original image labeling processing unit based on an assumed pixel size of a specific object included in the original image. The image processing apparatus according to claim 1, further comprising a switching processing unit.   The image processing apparatus according to claim 1, wherein the specific object is a skin discoloration region such as a stain, a pore, or a wrinkle included in a human face image.   The specific object is a skin discoloration region such as a stain, a pore or a wrinkle included in a human face image, and the switching processing unit activates an original image labeling processing unit when the pore is a specific object, and stains The image processing apparatus according to claim 4, wherein the reduction processing unit, the reduced image labeling processing unit, and the object identification processing unit are switched to operate when the object is a specific object. A reduction processing step for generating a reduced image obtained by reducing the original image at a reduction ratio determined in accordance with an assumed pixel size of a specific object included in the original image;
A reduced image labeling processing step of extracting a region of a specific object included in the reduced image by attaching the same label to a plurality of continuous pixels having the same attribute among the pixels constituting the reduced image;
An image processing method comprising: an object identification processing step for enlarging a region of a specific object extracted in the reduced image labeling processing step to a size of the original image and identifying the specific object as a specific object included in the original image.   The reduced image labeling processing step executes a labeling process in the order of a reduced image with a large reduction ratio and a reduced image with a small reduction ratio generated in the reduction processing step, and performs a reduction ratio during a labeling process for a reduced image with a small reduction ratio. The image processing method according to claim 7, wherein the labeling process is executed except for a specific object extracted from a large reduced image and a peripheral area thereof.   For the contour region of the specific object identified in the object identification processing step, the region of the specific object included in the original image is extracted by attaching the same label to a plurality of continuous pixels having the same attribute as the specific object. The image processing method according to claim 7, further comprising a contour region labeling processing step.

Images