JP5286215B2 - Outline extracting apparatus, outline extracting method, and outline extracting program - Google Patents

Description

  The present invention relates to a contour extraction device, a contour extraction method, and a contour extraction program.

  As a method for extracting only an object from an image and a video, the snake method, which is one method of the dynamic contour method, has attracted attention as an algorithm that can be used by improving the processing capability of a computer. The snake method is one of the methods for processing a matching of a two-dimensional image and a line figure shape as a kind of energy minimization problem, and targets an outline of an object existing in the image as an initial line figure pattern ( The shape is collated while dynamically deforming the initial contour (see, for example, Non-Patent Document 1). Since the snake method is repeatedly performed until it is determined that the collation has been converged, the contour can be clearly extracted. FIG. 12 is a diagram for explaining the outline of the contour extraction procedure using the snake method. FIG. 12A shows an input image 501 that includes an extraction object 511. With respect to this input image 501, an initial contour 521 is set so as to surround the entire object 511 as shown in FIG. 12B, and it is determined that the initial contour has converged while dynamically deforming using the snake method. Do computations until Thereby, the object 531 can be extracted as shown in FIG.

  As an initial contour setting method using the snake method, a method of manually giving an initial contour (for example, refer to Patent Document 1), a method in which a user of the contour extraction device manually sets an initial contour (for example, Patent Document 2).

JP-T-2004-516589 Japanese Patent Laid-Open No. 10-187936

Hideyuki Tamura, “Computer Image Processing”, Ohmsha, December 20, 2002, P261-262

  Thus, in the contour extraction method using the snake method, it is necessary to set an initial contour as a range for specifying an object to be extracted from an acquired image. If the initial contour is not appropriately set in accordance with the shape of the object to be extracted, there has been a problem that the number of processes until the calculation of the extraction increases and the time required for extraction increases. An example of increase / decrease in the number of times of convergence due to the difference in initial contour setting will be described with reference to FIG. FIG. 13A shows an example in which an initial contour 621 that surrounds the entire object 611 is set for the object 611 of the input image 601, and the contour 641 is formed as in the state of FIG. The number of operations until convergence was 1000 times. On the other hand, FIG. 13B is an example in which an initial contour 631 having a shape approximate to the object 621 is set, and the number of calculations until the contour 641 converges as in the state of FIG. there were. Thus, in the snake method, it is important to set the initial contour appropriately in order to extract at high speed.

  However, as a method for setting the initial contour, the methods of Patent Document 1 and Patent Document 2 have a problem that it is necessary to manually provide the initial contour. In addition, as a method of automatically generating the initial contour, there is a method of generating and using a general shape (square or circle) as shown in FIGS. 12B and 13A, for example. In this case, as already described, there has been a problem that the number of convergence calculations increases.

  The present invention has been made in view of the above-described problems. In contour extraction using the dynamic contour method, an appropriate initial contour is generated from an input image even if the background and the object are different, and a convergence calculation is performed. It is an object of the present invention to provide a contour extraction device, a contour extraction method, and a program that can reduce the number of times and increase the extraction time.

  In order to achieve the above object, according to the present invention, in a contour extraction apparatus that performs contour extraction using a dynamic contour method, a color space separation unit that separates an input image into a plurality of color space components, and the color space separation unit includes: A color space component binarization unit that binarizes each separated color space component, and an initial stage that generates initial contour image data based on the image data of each color space component binarized by the color space component binarization unit A contour generation unit, and a contour extraction unit that extracts a contour of a graphic included in the input image using the initial contour image data generated by the initial contour generation unit by a dynamic contour method.

  Further, the present invention provides a contour extraction apparatus that performs contour extraction using a dynamic contour method, wherein the initial contour generation unit includes an image center portion of each color space component binarized by the color space component binarization unit. If the pixel is determined to be a white pixel or a black pixel, and determined to be a black pixel, the initial contour image data is generated based on image data obtained by reversing the image data of each color space component after binarization into black and white, When it is determined that the pixel is a white pixel, the initial contour image data is generated based on the image data of each color space component after binarization.

  Further, the present invention provides a contour extraction apparatus that performs contour extraction using a dynamic contour method, wherein the initial contour generation unit includes image data of each color space component binarized by the color space component binarization unit. The initial contour image is generated using at least one.

  Further, the present invention provides a contour extraction apparatus that performs contour extraction using a dynamic contour method, wherein the initial contour generation unit synthesizes image data of each color space component binarized by the color space component binarization unit. Then, the initial contour image data is generated.

  Further, the present invention provides a contour extraction apparatus that performs contour extraction using a dynamic contour method, wherein the initial contour generation unit applies image data of each color space component binarized by the color space component binarization unit. Then, after weighting each color space component, initial contour image data is generated by combining the weighted color space components.

  The present invention also provides a contour extraction apparatus that performs contour extraction using a dynamic contour method, and performs a smoothing process on the input image before the color space separation unit separates into color space components. And further comprising.

Further, the present invention provides a contour extraction apparatus that performs contour extraction using a dynamic contour method, wherein the input image is grayscaled, and each pixel of the grayscaled image data includes a pixel or an object constituting a background. It is determined based on the initial contour image data generated by the initial contour generation unit whether the pixel is a constituent pixel, and based on the determination result, the background pixel value is multiplied by a constant of 1 or more to make the pixel thinner, or the object A contrast enhancement unit that performs density enhancement by multiplying the pixel value by a constant value of 1 or less to darken the pixel, and generates density enhancement image data, and the contour extraction unit includes: It is characterized by extracting the contour of the input image from the generated the grayscale enhanced image data by using the initial contour image data, wherein the initial contour generation unit has generated.

  Further, the present invention provides the contour extraction apparatus that performs contour extraction using the dynamic contour method, wherein the shading emphasis unit multiplies the background pixel value by a constant of 1 or more and adds the background pixel value to the object pixel value. It is characterized by multiplying the inverse of the value multiplied by.

  Further, the present invention relates to a contour extraction device that performs contour extraction using a dynamic contour method, wherein the shading emphasis unit is based on a histogram generated from the gray-scaled image data, and is a pixel of an object according to contrast. It is characterized in that a value to be multiplied by the value and the background pixel value is set.

Further, the present invention provides a contour extraction apparatus that performs contour extraction using a dynamic contour method, wherein the input image is grayscaled, and each pixel of the grayscaled image data includes a pixel or an object constituting a background. It is determined based on the initial contour image data generated by the initial contour generation unit whether the pixel is a component, and based on the determination result, the average value of all the pixel values of the background and the average of all the pixel values of the object Calculate the value, compare the calculated average value, calculate the multiplier of each of the background and the object based on the comparison result, use the calculated multipliers to emphasize the density of the background and the object, anda shading enhancement unit that generates a grayscale enhanced image data, wherein the contour extraction unit, the initial contour generator initial contour image de generated from the gray weighted image data to which the shading enhancement unit has generated It is characterized by extracting the contour of the input image using the data.

  The present invention also relates to a contour extraction method of a contour extraction device that performs contour extraction using a dynamic contour method, wherein a color space separation unit separates an input image into a plurality of color space components, A space component binarization unit binarizes each color space component separated by the color space separation step, and an initial contour generation unit binarizes the color space component binarization step. An initial contour generation step for generating initial contour image data based on the converted image data of each color space component, and a contour extraction unit in the input image using the initial contour image data generated by the initial contour generation step And a contour extracting step of extracting a contour of the contained figure by a dynamic contour method.

  According to another aspect of the present invention, there is provided a contour extraction program of a contour extraction apparatus that performs contour extraction using a dynamic contour method, wherein a computer performs a color space separation step of separating an input image into a plurality of color space components, and the color space separation. A color space component binarization step for binarizing each color space component separated in the process, and initial contour image data is generated based on the image data of each color space component binarized by the color space component binarization step. Performing an initial contour generation step, and a contour extraction step of extracting a contour of a figure included in the input image by using a dynamic contour method using the initial contour image data generated by the initial contour generation step. It is said.

  According to the present invention, an appropriate initial contour is generated from an input image even if the background and the object are different, and the number of convergence operations for performing contour extraction by the dynamic contour method using the generated initial contour is reduced, and the extraction time It becomes possible to speed up.

It is a block diagram which shows an example of a structure of the outline extraction apparatus which concerns on embodiment of this invention. It is a figure which shows an example of a structure of each pixel value of the RGB component which the image elementary value memory | storage part 107 concerning the embodiment has memorize | stored. It is a figure explaining an example of the outline of the outline extraction procedure concerning the embodiment. It is a figure explaining an example of the predominance of smoothing concerning the embodiment. It is a figure explaining an example of the predominance of smoothing concerning the embodiment. It is a figure which shows the outline of the production | generation procedure of the initial outline image data which concerns on the embodiment. Is a flow chart for explaining the outline extraction procedure according to the embodiment. It is a flowchart of the production | generation procedure of the initial outline image data which concerns on the same embodiment . It is a flowchart of the production | generation procedure of the initial outline image data which concerns on the embodiment. It is a flowchart of the mask data generation procedure of each color space component according to the embodiment. It is a flowchart of the density emphasis procedure which concerns on the same embodiment. It is a figure explaining the outline of the procedure of the contour extraction using the snake method which concerns on the conventional embodiment. It is a figure explaining an example of increase / decrease in the frequency | count of convergence by the difference in the setting of the initial outline which concerns on the conventional embodiment.

  Hereinafter, an embodiment of the present invention will be described with reference to FIGS. In addition, this invention is not limited to the structure of the embodiment which concerns, A various change is possible within the range of the technical idea of this invention.

  The configuration of the contour extraction apparatus according to this embodiment will be described with reference to a block diagram illustrating an example of the configuration of the contour extraction apparatus in FIG. The contour extracting apparatus 10 includes an external setting value capturing unit 101, an image data capturing unit 102, a smoothing unit 103, a color space separation unit 104, a color space component binarization unit 105, and an initial contour generation unit 106. A pixel value storage unit 107, a gray scale conversion unit 108, a density enhancement unit 109, a contour extraction unit 110, and a contour extraction result output unit 111. Further, a camera 11 and an image display device 13 are connected to the contour extraction device 10.

  The camera 11 is composed of, for example, a light receiving lens and a CCD camera, captures an input image including an object, and transmits the captured input image to the contour extraction device 10. The image display device 13 receives image data for display of the contour extraction result from the contour extraction device 10 and displays the received image data.

  The external setting value capturing unit 101 captures external setting values that are various settings used for the contour extraction processing from the outside. The external setting value includes various parameters of smoothing performed by the smoothing unit 103 (processing method, number of processings, range to be processed, shape to be processed, weighting coefficient, etc.), color space separation method performed by the color space separation unit 104 Types (RGB, CMY, etc.), parameters of density emphasis processing performed by the density emphasis unit 109 (setting of whether to perform processing, multiplier at the time of processing, etc.), various parameters of the dynamic contour method performed by the contour extraction unit 110 ( Processing method, number of calculations, etc.). In addition, the external setting value capturing unit 101 outputs the captured external setting value to the smoothing unit 103, the color space separation unit 104, the density enhancement unit 109, and the contour extraction unit 110.

  The image data capturing unit 102 captures an input image including an object photographed by the camera 11, and the image data (hereinafter referred to as input image data) of the captured input image is smoothed by a smoothing unit 103, a grayscale conversion unit 108, and an outline. The result is output to the extraction result output unit 111.

  The smoothing unit 103 captures the input image data captured by the image data capturing unit 102 and the external setting value output by the external setting value capturing unit 101. Also, the smoothing unit 102 reads out the type of smoothing method and various parameters for smoothing from the acquired external setting values. Further, the smoothing unit 102 smoothes the acquired input image data by the number of acquired external setting values, for example, using a Gaussian filter, based on the type of the read smoothing method. Also, the smoothing unit 103 outputs the smoothed image data to the color space separation unit 104.

  The color space separation unit 104 captures the smoothed image data smoothed by the smoothing unit 103 (hereinafter referred to as smoothed image data) and the external setting value output by the external setting value capturing unit 101. Further, the color space separation unit 104 separates the captured smoothed image data into image data of each channel of the color space component based on an instruction for a separation method of external setting values. Color space component separation methods are RGB (red green blue), CMY (cyan, magenta, yellow), HSV (hue, saturation, brightness), YCbCr (luminance, blue signal color difference, red signal color difference), YPbPr (luminance, blue signal). Color difference, red signal color difference), and the like. In addition, the color space separation unit 104 outputs the image data of each channel separated into color space components to the color space component binarization unit 105.

  The color space component binarization unit 105 takes in the image data of each channel separated by the color space separation unit 104. In addition, the color space component binarization unit 105 binarizes the captured image data of each channel using a threshold setting method (hereinafter referred to as Otsu's threshold setting method) proposed by, for example, Noriyuki Otsu in 1979. Turn into. Further, the color space component binarization unit 105 outputs the binarized image data of each channel to the initial contour generation unit 106.

  The initial contour generation unit 106 captures image data of each channel binarized by the color space component binarization unit 105 (hereinafter referred to as binarized image data). Further, the initial contour generation unit 106 writes and stores each pixel value of the captured binary image data of each channel in the pixel value storage unit 106. In addition, the initial contour generation unit 106 generates initial contour image data according to a procedure described later, using the captured binary image data of each channel and the pixel values stored in the binarization storage unit 106. Further, the initial contour generation unit 106 outputs the generated initial contour image data to the density enhancement unit 108 and the contour extraction unit 110.

The pixel value storage unit 107 stores each pixel value of the binarized image data of each channel. FIG. 2 is a diagram illustrating an example of the configuration of each pixel value of the RGB components stored in the pixel value storage unit 107. When the color space separation unit 103 separates the smoothed image data into RGB, as shown in FIG. 2, it is realized as tabular data, and the pixel value P _R (1 or 0) of the R channel of the pixel 1 Each data includes a G channel pixel value P _G (1 or 0) and a B channel pixel value P _B (1 or 0).

  The gray scale conversion unit 108 captures the input image captured by the image data capturing unit 102. Further, the gray scale conversion unit 108 determines whether the captured input image is color image data or a gray scale image. Further, when the grayscale conversion unit 108 determines that the input image is a color image, the grayscale conversion unit 108 converts the input image into grayscale image data by a general method. Further, the gray scale conversion unit 108 outputs the converted gray scale image data to the density enhancement unit 109 and the contour extraction unit 110.

  The density enhancement unit 109 captures the external setting values captured by the external setting value capturing unit 101, the initial contour image data generated by the initial contour generation unit 106, and the grayscale image data converted by the grayscale conversion unit 108. Further, the light / dark emphasis unit 109 reads out an instruction value as to whether or not to perform light / dark emphasis processing and various parameters of the light / dark emphasis processing from the acquired external setting values. In addition, the gray level enhancement unit 109 uses the instruction value as to whether or not to perform the gray level enhancement process of the read external setting value, the multiplier at the time of the gray level enhancement process, and the initial contour image data, and the grayscale image data by a method described later. Is emphasized, and density-enhanced image data is generated. In addition, the density enhancement unit 109 outputs the generated density enhancement image data to the contour extraction unit 110.

  The contour extraction unit 110 outputs external setting values output from the external setting value capturing unit 101, initial contour image data output from the initial contour generation unit 106, grayscale image data output from the grayscale conversion unit 108, and density enhancement. The grayscale emphasized image data generated by the unit 109 is taken in. The contour extraction unit 110 reads information indicating the type of the dynamic contour method and various parameters of the dynamic contour method from the acquired external setting values. In addition, the contour extraction unit 110 uses the initial contour image data and various parameters (processing method, number of calculations, etc.) of the dynamic contour method of the external setting value to input the grayscale emphasized image data by the dynamic contour method. Using the data, the contour data of the object is extracted from the density-enhanced image data. In addition, the contour extraction unit 110 outputs the extracted contour data to the contour extraction result output unit 111.

  The contour extraction result output unit 111 captures the input image output from the image data capturing unit 102 and the contour data of the object extracted by the contour extracting unit 110. In addition, the contour extraction result output unit 111 extracts image data of the target object from which the background region has been removed using the contour data from the input image that has been captured. The contour extraction result output unit 110 outputs the extracted image data of the object to the image display device 13.

  Next, an example of the outline of the contour extraction procedure in the present embodiment will be described with reference to FIG. An input image 201 as shown in FIG. 3A is acquired, and the acquired input image 201 is subjected to blurring processing (smoothing) to generate smoothed image data 202 as shown in FIG. Next, the smoothed post-smoothing image data 202 is separated into RGB color space components, and the separated image data of each channel is binarized. After binarization, initial contour image data 203 as shown in FIG. 3C is generated using the binarized image data. On the other hand, after the acquired input image 201 is converted to gray scale, density enhancement is performed to generate density enhanced image data 204 as shown in FIG. Next, the contour data of the object 205 is extracted by the dynamic contour method using the initial contour image data 203 and the grayscale emphasized image data 204. Next, using the contour data, image data of the object 205 from which the background has been removed from the input image as shown in FIG.

Next, an example of superiority in performing smoothing (blurring processing) will be described with reference to FIGS. FIG. 4A is an example of an input image including an object with a complicated pattern. FIG. 4B is an example of initial contour image data generated by binarization processing of FIG. When the pattern of the object is complex as shown in FIG. 4A, if the initial contour image data is generated as it is by general binarization processing as it is, this image data becomes as shown in FIG. Contains a lot of data. When the image data of FIG. 4B is used as the initial contour and the contour is extracted by the snake method, the total length of the line segments included in the initial contour image data is long, so that the number of calculations increases, resulting in the contour extraction time. Will increase.
On the other hand, if smoothing processing using a Gaussian filter is performed three times on the input image of FIG. 4A, the input image is blurred as shown in FIG. When the initial contour image data is generated from the image data of FIG. 5A, initial contour image data in which the pattern data is reduced as compared with FIG. 4B can be generated as shown in FIG. 5B. Even in the state where the input image is blurred in this way, the initial contour image data as shown in FIG. 5B is a component of the contour of the target object that can extract the contour of the target object by the dynamic contour method. Contains information. If the contour is extracted using the initial contour image data shown in FIG. 5B, the processing can be performed at a higher speed than the case where the contour is extracted using the initial contour image data shown in FIG.
That is, by performing an appropriate number of smoothings on an input image by an appropriate method, the amount of contour extraction can be reduced, and as a result, contour extraction can be performed at high speed.

  Next, the contour extraction method will be described with reference to the procedure for generating the initial contour image data in FIG. 6 and the flowcharts in FIGS. FIG. 6 is a diagram showing an outline of a procedure for generating initial contour image data. FIG. 7 is a flowchart for explaining a contour extraction procedure, FIGS. 8 and 9 are flowcharts for a procedure for generating initial contour image data, and FIG. 10 is a flowchart for a mask data generation procedure for each color space component. 11 is a flowchart of the contrast enhancement procedure. In the present embodiment, when the input image is a color image, RGB is designated as the color space separation method, Gaussian processing is instructed as the smoothing method, and tone enhancement processing is instructed by the external setting value Will be described.

  First, the external setting value reading unit 101 takes in external setting values that are various settings used in the contour extraction process (step S1). The external setting value reading unit 101 outputs the acquired external setting value to the smoothing unit 103, the color space separation unit 104, the density enhancement unit 109, and the contour extraction unit 110.

  Next, the contour extracting apparatus 10 performs initial contour image data generation processing (step S2). The procedure for generating the initial contour image data in step S2 will be described with reference to the flowcharts of FIGS. 8 and 9 and FIG. First, the image data capturing unit 102 captures the input image 301 shown in FIG. 6 captured by the camera 11 (step S101). Further, the image data capturing unit 102 outputs the captured input image 301 to the smoothing unit 103, the grayscale conversion unit 108, and the contour extraction result output unit 111.

  The smoothing unit 103 captures the input image 301 captured by the image data capturing unit 102 and the external setting value captured by the external setting value capturing unit 101. In addition, the binarization unit 103 reads the smoothing method and various parameters for smoothing from the acquired external setting values. Based on the read external setting value, the smoothing unit 103 performs Gaussian processing on the input image 301 the number of times set by the external setting value, for example, three times (steps S102 to S104).

  After finishing the Gaussian processing for the number of times of external set value instruction, the smoothing unit 103 writes and stores the smoothed image data 311 shown in FIG. 6 subjected to the Gaussian processing in the smoothing unit 103 (step S105). Also, the smoothing unit 103 reads the smoothed image data 311 and outputs it to the color space separation unit 104 (step S105).

  Next, the color space separation unit 104 captures the smoothed image data 311 smoothed by the smoothing unit 103 and the external setting value captured by the external setting value capturing unit 101. Further, the color space separation unit 104 reads the type of the color space separation method from the acquired external setting value. In addition, the color space separation unit 104 separates the image data 311 into RGB channel image data using a general method based on the external setting value (step S106). Further, the color space separation unit 104 outputs the image data (321, 322, 323) of the separated RGB channels shown in FIG. 6 to the color space component binarization unit 105.

  Next, the color space component binarization unit 105 takes in the image data (321, 322, 323) of each of the RGB channels separated by the color space separation unit 104. Further, the color space component binarization unit 105 performs contour mask processing for each channel on the captured RGB image data (321, 322, 323) (steps S107 to S108).

The contour mask process for each channel in steps S107 to S108 will be described with reference to the flowchart of FIG. Note that the contour mask processing for each of the RGB channels performs the same steps S301 to S306, and thus the description of the overlapping portions is omitted.
First, R channel contour mask processing will be described. The color space component binarization unit 105 generates a histogram of the captured R channel image data 321 (step S301).

  Next, the color space component binarization unit 105 uses the generated histogram to calculate a binarization threshold using, for example, the Otsu threshold determination method (step S302).

  Next, the color space component binarization unit 105 binarizes the R channel image data 321 using the calculated threshold value (step S303). Further, the color space component binarization unit 105 outputs the binarized image data to the initial contour generation unit 106.

Next, the initial contour generation unit 106 takes in the image data (hereinafter referred to as binarized image data) 321 binarized by the color space component binarization unit 105. Further, the initial contour generation unit 106 determines whether the pixel 331 at the center of the captured binarized image data is a black pixel, that is, whether the pixel value P _R = 0 or the white pixel P _R = 1 (step S1). S304).
When it is determined that the pixel 331 at the center of the binarized image data is a black pixel (pixel value P _R = 0) (step S304; 0), the initial contour generation unit 106 selects all the pixels of the binarized pixel data 321. The values 0 and 1 are exchanged to generate image data with black and white inverted (step S305). Further, the initial contour generation unit 106 writes and stores the inverted binarized image data in the pixel value storage unit 107 as contour mask image data.
Also, when the pixel 321 at the center of the binarized image data is determined to a white pixel (pixel value P _{R =} 1) (Step S304; 1), the initial contour generation unit 106 inverts the binarized pixel data Instead, the binarized image data is written and stored in the initial contour generation unit 106 as contour mask image data as it is.
In the case of the R channel image data 321, the initial contour generation unit 106 determines that the pixel 331 at the center of the binarized image data is a white pixel, does not invert the binarized pixel data, and does not invert the binarized image. The data is written and stored in the initial contour generation unit 106 as contour mask image data.

  In the case of the image data 322 of the G channel, the same processing is performed from step 301 to step 306, and the initial contour generation unit 106 determines that the pixel 332 at the center of the binarized image data is a white pixel and binarizes it. The pixel data is not inverted, and the binarized image data is written and stored in the initial contour generation unit 106 as contour mask image data as it is.

  In the case of the B channel image data 323, the same processing is performed from step 301 to step 306, and the initial contour generation unit 106 determines that the pixel 333 at the center of the binarized image data is a black pixel and inverts it. The valued image data is written and stored in the initial contour generation unit 106 as contour mask image data.

The reason for determining whether the pixel at the center of the binarized image data is a black pixel or a white pixel and performing the black and white inversion of the binarized image data will be described with reference to FIG. In FIG. 6, the image data binarized after being separated into RGB channels is R channel image data 321, G channel image data 322, and B channel image data 323. In FIG. 6, since the color of the object is blue, when binarized after separation into RGB channels, the B channel image data 323 is not only the background, but also the pixel value P _B of the object is 0, that is, black. It will be represented. When these image data 321 to 323 are superimposed, an image like the image data 343 is obtained. That is, the object is represented by a pixel value = 0 (black pixel), and it is difficult to generate initial contour image data for extracting the object. The object to be extracted is assumed to be included in the center of the input image, and if the pixel value at the center of the image data is 0, it is determined that the object is generated as a black pixel by binarization. The pixel value of the object is reset to 1 by reversing the pixels of the entire image data in black and white. With these processes, the pixel value of the object = 1 can be maintained as in the image data 341 even when the image data of each of the RGB channels is superimposed.

Returning to the flowchart of FIG. 8, the initial contour generation unit 106 reads out the stored contour mask image data of each RGB channel, superimposes the read out RGB contour mask image data, and generates superimposed image data 341. (Step S110). Further, the initial contour generation unit 106 writes and stores the superimposed image data and the pixel values (P _R , P _G , P _B ) of the RGB channels in the pixel value storage unit 108 (step S111).
The reason for superimposing the decomposed RGB contour mask image data is as follows. For example, when only the center of the object is blue (B), the entire B channel image data 323 is reversed in black and white according to the determination in step S304. In this case, for example, if the initial contour image data is generated using only the decomposed B channel, the initial contour image data cannot be generated properly. For this reason, by superimposing the RGB contour mask image data, it is possible to prevent erroneous detection of generating the initial contour image data with only one channel, so that the initial contour image data can be appropriately generated.

Next, moving to the flowchart of FIG. 9, the initial contour generation unit 106 determines whether the total value of the pixel values (P _R , P _G , P _B ) of the RGB channels for all the pixels of the superimposed image data = 3. The determination of whether or not is repeated (steps S201 to S204).
When the total value of the pixel values of each RGB channel is other than 3 (step S202; No), 0 is substituted for each pixel value of each RGB channel of the pixel, that is, a black pixel is obtained (step S203). That is, P _R = 0, P _G = 0, and P _B = 0 are substituted, and the pixel value stored in the pixel value storage unit 108 is rewritten.
Sum = 3 of the pixel values of each RGB channel, i.e., if all the pixels are white pixels (Step S202; Yes), the pixel value of each RGB channel _{P R,} P G, the value of _{P B} is not changed ( Step S203).

After the pixel value processing for all image data is completed, the initial contour generation unit 106 reads out the pixel values (P _R , P _G , P _B ) of the RGB channels stored in the pixel value storage unit 107 and performs initial processing. Contour image data is generated. Further, the initial contour generation unit 106 writes and stores the generated initial contour image data in the initial contour generation unit 106 (step S205).
When the initial contour image data is generated, when at least one of the pixel values (P _R , P _G , P _B ) of each RGB channel is 0, that is, a black pixel, the initial contour generation unit 106 sets the pixel value of the pixel. Set to 0 to generate initial contour image data. Further, when the pixel values (P _R , P _G , P _B ) of each of the RGB channels are all 1, that is, white pixels, the initial contour generation unit 106 sets the pixel value of the pixel to 1, and the initial contour image Data 351 is generated.

In addition, the initial contour generation unit 106 outputs the generated initial contour image data 351 to the density enhancement unit 109 and the contour extraction unit 110 (step S206).
This completes the initial contour image data generation processing in step S2.

  Returning to the flowchart of FIG. 7, the shading enhancement unit 109 performs shading enhancement processing (step S <b> 3). The density enhancement processing procedure will be described with reference to the flowchart of FIG.

The density enhancement unit 109 captures the external setting value captured by the external setting value capturing unit 101. Also, the density enhancement unit 109 reads an instruction as to whether or not to perform density enhancement processing and various parameters of density enhancement processing from the acquired external setting value. In the case of an instruction to perform density emphasis processing, the processing in steps S401 to S410 is performed based on the read density setting parameter. In the case of an instruction not to perform density emphasis processing, the processing proceeds to step S4 without performing density emphasis processing.
In the case of an instruction to perform density enhancement processing, the grayscale conversion unit 108 captures the input image captured by the image data capturing unit 102 and determines whether the captured input image is color image data or grayscale image data. When the grayscale conversion unit 108 determines that the input image is color image data, the grayscale conversion unit 108 converts the input image into 256-level grayscale image data using a general method, and converts the converted grayscale image data to grayscale conversion. The data is written and stored in the unit 108 (step S401).

  Next, the gray scale conversion unit 108 outputs the stored gray scale image data to the density enhancement unit 109 (step S402). Further, the initial contour image data generated by the initial contour generation unit 106 is captured (step S402).

  Next, the density enhancement unit 109 uses the grayscale image data output from the grayscale conversion unit 108 and the captured initial contour image data to perform the processing of steps S404 to S410 for all the pixels of the grayscale image data. The density emphasis is repeatedly performed, and the density emphasized image data in which the density is emphasized is generated. In the processing from step S404 to step S410, the various parameters of the shading enhancement processing are instructions for doubling the pixel value when the pixel is darkened (emphasized) and doubling the pixel value when thinning the pixel. An example will be described.

Next, the shading enhancement unit 109 reads one pixel that has not been read out from the read grayscale image data. Also, the density enhancement unit 109 compares the read one pixel with the captured initial contour image data to determine whether it is a background or an object (step S405).
A specific example will be described with reference to FIG. As shown in FIG. 3, in the initial contour image data 203, the pixel of the background 206 is a black pixel (pixel value = 0), and the pixel of the object 207 has a white pixel (pixel value = 1). The density enhancement unit 109 compares each pixel of the grayscale image data with each pixel at the same position in the initial contour image data. If the pixel value of the initial contour image data is 0 as a result of the comparison, the density enhancement unit 109 determines that the background is present, and if the pixel value of the initial contour image data is 1, it determines that the object is an object.

  As a result of the determination in step 405, when it is determined that the read pixel is an object (step S405; 1 (object)), the shading enhancement unit 109 sets the pixel value of the pixel to 1 / based on the external setting value. The pixel value is reduced by doubling the pixel value (step S406).

  As a result of the determination in step 405, when it is determined that the read pixel is the background (step S405; 0 (background)), the shading enhancement unit 109 doubles the pixel value of the pixel based on the external setting value. Thus, the pixel value is reduced (step S407).

In addition, after the step S407 ends, the density emphasis unit 109 determines whether the pixel value is greater than 255 (S408).
As a result of the determination in step S408, when it is determined that the value is 255 or less (step S407; false), the density enhancement unit 109 ends the density enhancement process for this pixel, and if there are pixels that have not been processed yet, Steps S404 to S410 are repeated until the tone emphasis processing for the pixel is completed (step S410).
As a result of the determination in step S408, if it is determined that it is greater than 255 (step S407; true), the density enhancement unit 109 substitutes 255 for this pixel value, and ends the density enhancement process for this pixel. In addition, if there is a pixel that has not been processed yet, the gray level enhancement unit 109 repeats Steps S404 to S410 until the gray level enhancement process for all pixels is completed (Step S410).
In this embodiment, since the input image is converted into grayscale image data of 256 gradations, 255 is substituted for all values greater than 255 in the processing from step S408 to step S409, and all pixel values are 256th floor. To make it fit.

After the density enhancement processing is completed for all the pixels, the density enhancement unit 109 generates the density enhanced image data using the pixel values for which the density enhancement processing has been completed, and outputs the generated density enhanced image data to the contour extraction unit 110. To do.
Thus, the shading enhancement process in step S3 is completed, and the process returns to FIG.

  Next, the gray scale conversion unit 108 outputs the stored gray scale image data to the contour extraction unit 110. The contour extraction unit 110 also includes an external setting value captured by the external setting value capturing unit 101, initial contour image data output by the initial contour generation unit 106, grayscale image data output by the grayscale conversion unit 108, The grayscale emphasized image data generated by the grayscale enhancement unit 109 is captured. In addition, the contour extraction unit 110 reads the type of the dynamic contour method and various parameters of the snack process from the acquired external setting values. In addition, the contour extraction unit 110 extracts contour data by the known snake method from the grayscale emphasized image data using the various parameters of the read snake process and the initial contour image data (step S4). In addition, the contour extraction unit 110 outputs the extracted contour data to the contour extraction result output unit 111.

  Next, the contour extraction result output unit 111 captures the input image captured by the image data capturing unit 102 and the contour data extracted by the contour extracting unit 110. In addition, the contour extraction result output unit 111 extracts image data of the target object from which the background region has been removed using the contour data from the input image that has been captured. The contour extraction result output unit 110 outputs the extracted image data of the object to the image display device 13 (step S4).

The image display device 13 displays the contour extraction result data fetched from the contour extraction result output unit 111.
This completes the contour extraction process for the object from the input image.

As described above, according to the present embodiment, the input image is smoothed, binarized after being separated into color space components after smoothing, and the initial color image components are synthesized by combining the binarized color space components. Generate data. Further, gray scale conversion is performed on the input image, and after the gray scale conversion, the gray level enhancement processing is performed on the background and the object using the initial contour image data, and the gray level enhancement image data is generated by the gray level enhancement processing. Then, using the generated initial contour image data, the contour of the object is extracted from the input image by the snake method, which is one of the dynamic contour methods, from the grayscale emphasized image data. However, an appropriate initial contour can be generated from the input image. For this reason, it is possible to reduce the number of convergence calculations for performing contour extraction by the dynamic contour method, and to realize a faster extraction time.
Even if the acquired image is unclear, the input image is converted to grayscale, and after performing grayscale conversion, after performing grayscale enhancement processing, contour data is extracted from the image data after grayscale enhancement processing using the initial contour. Thus, it is possible to reduce the number of convergence calculations for performing contour extraction by the dynamic contour method and to increase the extraction time.

 Furthermore, according to the present embodiment, it is possible to obtain a background image in which the object is not captured in advance, compare the background image with the input image, generate an initial contour, and extract the contour of the target. An appropriate initial outline can be automatically generated from the captured image. Furthermore, according to the present embodiment, a standard image (a plain background or no background) obtained by imaging an object acquired in advance from the front, an initial outline is not generated from the standard image, and the outline of the object is not extracted. However, it is possible to automatically generate an appropriate initial contour from the captured image.

  In the present embodiment, the example in which the smoothing unit 103 performs smoothing using a Gaussian filter has been described, but smoothing is performed using other general median filters, bilateral filters, blur filters, and the like. May be performed.

  Further, in the present embodiment, the example in which the smoothing number of the smoothing unit 103 is designated by the external setting value captured by the external setting value capturing unit 101 has been described. The number of times may be set according to the object. In this case, for example, the complexity of the input image may be determined based on the magnitude of the high-frequency component, the distribution degree of the component in the histogram, and the like, and the number of times of smoothing may be increased for complex image data. Further, as a result of the calculation as described above, when the number of times of averaging is 1, for example, color space separation may be performed without performing the averaging process.

  In this embodiment, an example in which the smoothed image data is stored in the smoothing unit 103 has been described. However, a ROM (Read Only Memory) connected to a CPU (Central Processing Unit) (not shown) of a contour extraction device (not shown), You may write in and memorize | store in memory | storage devices, such as a USB memory connected via HDD (Hard Disk Drive) or USB (Universal Serial Bus) I / F.

  Further, in the present embodiment, an example in which the color space component binarization unit 105 and the initial contour generation unit 106 are each performed is described. However, according to the method of the color space to be separated, for example, when separating into RGB, The RGB color space component binarization unit 105 or the RGB initial contour generation unit 106 may be included.

  In the present embodiment, the color space component binarization unit 105 has been described as an example of setting with the Otsu threshold selection method. However, another general threshold selection method may be used.

  In this embodiment, the initial contour generation unit 106 binarizes the separated image data after separation into RGB color space components, and combines the binarized image data of each channel to generate an initial contour. Although an example of generating image data has been described, weighting or the like may be used as a synthesis method for each channel. In the case of FIG. 6, for example, a histogram of each channel is generated, and based on the generated histogram, the R channel image data 321 that is considered to have a high degree of separation of each component is synthesized with the highest weight. . Alternatively, the initial contour image data may be generated using any one of the image data of each channel.

  Further, in this embodiment, the example in which the initial contour image data is extracted as it is by the snake method has been described. However, when the initial contour image data of the target object has a hole-like black region, for example, a small pixel by a black pixel in FIG. The area 352 is extracted by the initial contour generation unit 106 using a general labeling technique or the like. Further, the initial contour generation unit 106 converts the extracted hole-shaped black pixel region into a white pixel, thereby forming a hole-shaped black pixel that is a small region of the black pixel surrounded by the white pixel around the object. Can be filled with white pixels, and an initial contour capable of extracting a contour at high speed can be generated.

  In addition, a small area of white pixels surrounded by black pixels is noise when viewed from the initial contour image data, and in the case of such noise in the background, the initial contour generation unit 106 similarly performs general labeling. Extracted by processing method. Further, the initial contour generation unit 106 converts the extracted white pixel area of the noise into black pixels, thereby removing the white pixels that are the background noise by converting them into black pixels. An initial contour that can be extracted can be generated.

  Further, in the present embodiment, the example of performing the grayscale enhancement by thinning the background pixels and darkening the target object for the grayscale enhancement process has been described, but this method is effective for an input image with a bright background and a dark target object. It is. For this reason, in consideration of an input image with a dark background and a bright object, the shading enhancement unit 109 uses the grayscale image data and the initial contour image data to determine the pixel value of the region determined as the background by the grayscale image data. The average value and the average value of the pixel values of the area determined as the object by the grayscale image data are calculated. Next, the shading enhancement unit 109 determines which pixel area of the background or the object is brighter based on the calculated average value of each pixel value, and based on this determination result, the brighter one becomes brighter By calculating a multiplier for making (darker) and darker (darker) darker ones and performing density enhancement processing, it is possible to extract contours under various conditions.

  In the present embodiment, the binarized image data of each channel separated into color space components is determined whether the pixel value at the center of the input image is a black pixel or a white pixel, and the binarized image of each channel is determined. Although the example of inverting the black and white of the data has been described, the central pixel used for the determination may be determined using a plurality of pixels near the center instead of one pixel. The central portion of the binarized image data may be the centroid or the center of gravity of the binarized image data. Alternatively, labeling processing is performed on the binarized image data, the barycentric position of each label is obtained, and the pixel value of the area where the barycentric position is closest to the image center of the binarized image data is extracted. In other words, it may be determined to perform black and white reversal so as to be white.

  In the present embodiment, the contrast enhancement process is performed on the input image output from the image data capturing unit 103 using the initial contour image data to generate the density enhancement image data, and the initial contour image data is used from the density enhancement image data. Although the example of extracting the contour data has been described, contour extraction may be performed using the initial contour image data from the input image according to the input image or the object.

  Further, in this embodiment, the example in which the contour is extracted by using the generated initial contour image data by the snake method which is one of the dynamic contour methods has been described, but the contour extraction is performed by using another dynamic contour method. Alternatively, contour extraction may be performed using another contour extraction method. Furthermore, the generated initial contour image data may be used, for example, for limiting the range of an object to be extracted, and this initial contour image data may be used for specifying a region when performing feature point extraction in graphic recognition. .

It should be noted that a program for realizing the function of each unit in FIG. 1 of the embodiment is recorded on a computer-readable recording medium, and the program recorded on this recording medium is read into a computer system and executed, thereby executing each unit. Processing may be performed. Here, the “computer system” includes an OS and hardware such as peripheral devices.
Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.
The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, it also includes those that hold a program for a certain period of time, such as a volatile memory inside a computer system serving as a server or client in that case. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

DESCRIPTION OF SYMBOLS 10 ... Contour extraction apparatus 11 ... Camera 13 ... Image display apparatus 101 ... External setting value taking-in part 102 ... Image data taking-in part 103 ... Smoothing part 104 ... Color space separation Unit 105 ... color space component binarization unit 106 ... initial contour generation unit 107 ... pixel value storage unit 108 ... gray scale conversion unit 109 ... shade enhancement unit 110 ... contour extraction unit 111 ... contour extraction result output unit

Claims (8)

In a contour extraction apparatus that performs contour extraction using the dynamic contour method,
A color space separation unit that separates an input image into a plurality of color space components;
A color space component binarization unit that binarizes each color space component separated by the color space separation unit;
An initial contour generation unit that generates initial contour image data based on the image data of each color space component binarized by the color space component binarization unit;
A contour extracting unit that extracts a contour of a figure included in the input image by using an initial contour image data generated by the initial contour generating unit;
The input image is grayscaled, and whether each pixel of the grayscaled image data constitutes a background or a pixel constituting an object is determined based on the initial contour image data generated by the initial contour generation unit, Based on the result of the determination, the background pixel value is multiplied by a constant of 1 or more to make the pixel thinner, or the object pixel value is multiplied by a constant of 1 or less to make the pixel darker, thereby enhancing the shading. And a tone emphasis unit for generating tone emphasis image data,
With
The contour extraction unit
A contour extracting apparatus, wherein the contour of the input image is extracted from the grayscale emphasized image data generated by the grayscale emphasizing unit using the initial contour image data generated by the initial contour generating unit. The shading emphasis part is
Multiplied by a constant pixel value of the background with one or more values, a contour extraction apparatus according to claim 1, characterized in that multiplying the reciprocal of the value obtained by multiplying the background pixel value of the object. The shading emphasis part is
Based on said histogram generated from the gray scale of image data, the contour of claim 1 or claim 2, characterized in that setting the value to be multiplied by the pixel value of the pixel value and the background object in accordance with the contrast Extraction device. In a contour extraction apparatus that performs contour extraction using the dynamic contour method,
A color space separation unit that separates an input image into a plurality of color space components;
A color space component binarization unit that binarizes each color space component separated by the color space separation unit;
An initial contour generation unit that generates initial contour image data based on the image data of each color space component binarized by the color space component binarization unit;
A contour extracting unit that extracts a contour of a figure included in the input image by using an initial contour image data generated by the initial contour generating unit;

The input image is grayscaled, and it is determined based on the initial contour image data generated by the initial contour generation unit whether each pixel of the grayscaled image data is a pixel constituting a background or a pixel constituting an object. The average value of all the pixel values of the background and the average value of all the pixel values of the object are calculated based on the determined result, the calculated average values are compared, and the background and the target are calculated based on the comparison result. Calculating the respective multipliers of the objects, and using the calculated multipliers to perform density emphasis on the background and the object, and generating a density emphasized image data,
With
The contour extraction unit
The gray highlighted portion contour extractor you and extracts the contour of the input image using the initial contour image data, wherein the initial contour generation unit is generated from the gray weighted image data generated. In the contour extraction method of the contour extraction apparatus that performs contour extraction using the dynamic contour method,
A color space separation unit that separates the input image into a plurality of color space components;
A color space component binarization unit that binarizes each color space component separated by the color space separation step; and
An initial contour generation step for generating initial contour image data based on the image data of each color space component binarized by the color space component binarization step;
A contour extracting step in which a contour extracting unit extracts a contour of a figure included in the input image using the initial contour image data generated by the initial contour generating step by a dynamic contour method;
The initial contour image data generated by the initial contour generation unit, in which the gray level enhancement unit grayscales the input image and whether each pixel of the grayscaled image data is a pixel constituting a background or a pixel constituting an object Based on the determination result, the background pixel value is multiplied by a constant of 1 or more to make the pixel thinner, or the object pixel value is multiplied by a value of 1 or less to make the pixel darker. And a light / dark emphasis process for generating light / dark emphasis image data .
In the contour extraction process, the contour extraction unit extracts the contour of the input image using the initial contour image data generated by the initial contour generation unit from the grayscale emphasized image data generated by the grayscale enhancement unit. An outline extraction method characterized by the above. In the contour extraction method of the contour extraction apparatus that performs contour extraction using the dynamic contour method,
A color space separation unit that separates the input image into a plurality of color space components;
A color space component binarization unit that binarizes each color space component separated by the color space separation step; and
An initial contour generation step for generating initial contour image data based on the image data of each color space component binarized by the color space component binarization step;
A contour extracting step in which a contour extracting unit extracts a contour of a figure included in the input image using the initial contour image data generated by the initial contour generating step by a dynamic contour method;
The initial contour image data generated by the initial contour generation unit, in which the gray level enhancement unit grayscales the input image and whether each pixel of the grayscaled image data is a pixel constituting a background or a pixel constituting an object Based on the determination result, the average value of all the pixel values of the background and the average value of all the pixel values of the object are calculated, and the calculated average values are compared and compared. A gray scale emphasizing process in which a multiplier for each of the background and the object is calculated based on the image, and a gray scale emphasis of the background and the object is performed using each of the calculated multipliers, and a gray scale emphasized image data is generated.
With
In the contour extraction process, the contour extraction unit extracts the contour of the input image using the initial contour image data generated by the initial contour generation unit from the grayscale emphasized image data generated by the grayscale enhancement unit. Contour extraction method characterized by On the computer,
A color space separation step of separating the input image into a plurality of color space components;
A color space component binarization step for binarizing each color space component separated by the color space separation step;
An initial contour generation step of generating initial contour image data based on the image data of each color space component binarized by the color space component binarization step;
A contour extraction step of extracting a contour of a figure included in the input image by an active contour method using the initial contour image data generated by the initial contour generation step;
The initial contour image data generated by the initial contour generation unit, in which the gray level enhancement unit grayscales the input image and whether each pixel of the grayscaled image data is a pixel constituting a background or a pixel constituting an object Based on the determination result, the background pixel value is multiplied by a constant of 1 or more to make the pixel thinner, or the object pixel value is multiplied by a value of 1 or less to make the pixel darker. A contour extraction program for performing a grayscale enhancement process and a grayscale enhancement process for generating grayscale-weighted image data ,
In the contour extraction process, the contour extraction unit extracts the contour of the input image using the initial contour image data generated by the initial contour generation unit from the grayscale emphasized image data generated by the grayscale enhancement unit.
Contour extraction program characterized by On the computer,
A color space separation step of separating the input image into a plurality of color space components;
A color space component binarization step for binarizing each color space component separated by the color space separation step;
An initial contour generation step of generating initial contour image data based on the image data of each color space component binarized by the color space component binarization step;
A contour extraction step of extracting a contour of a figure included in the input image by an active contour method using the initial contour image data generated by the initial contour generation step;
The initial contour image data generated by the initial contour generation unit, in which the gray level enhancement unit grayscales the input image and whether each pixel of the grayscaled image data is a pixel constituting a background or a pixel constituting an object Based on the determination result, the average value of all the pixel values of the background and the average value of all the pixel values of the object are calculated, and the calculated average values are compared and compared. The contours for calculating the respective multipliers of the background and the object based on the image, and performing the gradation enhancement process for performing the gradation enhancement of the background and the object using the calculated multipliers and generating the gradation-enhanced image data An extraction program ,
In the contour extraction process, the contour extraction unit extracts the contour of the input image using the initial contour image data generated by the initial contour generation unit from the grayscale emphasized image data generated by the grayscale enhancement unit.
Contour extraction program characterized by