JP5653308B2 - Region detection apparatus, region detection method, and program - Google Patents

Description

  Embodiments described herein relate generally to an area detection apparatus, an area detection method, and a program for detecting an area surrounded by a frame from an image.

  As a technique for detecting a region surrounded by a frame from an image, a technique (first technique) for detecting a red frame from a road sign and identifying a label therein is conventionally known.

  A technique (second technique) for extracting and recognizing characters by detecting a connected region of white pixels surrounded by black ruled lines on a form has also been known.

Japanese Patent Publication No. 6-10838 JP-A-6-119491 JP-A-6-60220

  As described above, the first technique for detecting a frame of a specific color from a road sign requires a color image. For this reason, when the means for acquiring an image is, for example, a monochrome camera, it is difficult to apply the first technique to frame detection (region detection) because color information cannot be used.

  In the first and second technologies, it is necessary to distinguish the outside and the inside of the frame surrounding the area to be detected. However, when the frame and the characters in the frame are in contact with each other, the distinction criteria may be unclear.

  The problem to be solved by the present invention is that the area detection can improve the detection performance of the area surrounded by the frame by making the distinction between the inside and the outside of the frame when the frame and the characters in the frame are in contact with each other. An apparatus, a region detection method, and a program are provided.

  According to the embodiment, the region detection device includes a convex hull extraction unit and a region acquisition unit. The contour extracting means extracts a first contour pixel series from the input image for each contour. The convex hull extracting means extracts, for each series of the extracted first contour pixels, a second contour pixel series constituting a convex hull of the first contour pixel series. The area acquisition unit acquires the extracted second contour pixel series as information of an area detected from the input image.

The block diagram which shows the hardware constitutions of the area | region detection apparatus which concerns on 1st Embodiment. The block diagram which shows the function structure of the area | region detection apparatus which concerns on 1st Embodiment. The figure which shows the example of the input image applied in 1st Embodiment. The figure which shows a mode that the pixel in which the edge was not detected between the said 2 pixels was connected by the line segment among the groups of the pair of 2 pixels adjacent in the input image shown in FIG. The figure which shows a part of flowchart which shows the procedure of the outline extraction process performed by the outline tracking means in 1st Embodiment. The figure which shows the remainder of the flowchart which shows the procedure of the outline extraction process performed by the outline tracking means in 1st Embodiment. The figure which shows the example of the data structure of the information regarding the group of an outline applied with 1st Embodiment. The figure which shows all the series of the outline pixel extracted by the outline tracking means from the input image shown in FIG. The figure which shows the series of a part of outline pixel of the group of the series of outline pixels shown in FIG. 8 separately. The figure which shows the series of the remaining outline pixels of the group of outline pixel series shown in FIG. 8 separately. FIG. 11 is a diagram illustrating an example of a set of contour pixel series extracted by a selection unit from a group of contour pixel series shown in FIGS. 9 and 10 under a predetermined condition. The block diagram which shows the function structure of the character recognition mechanism added to the area | region detection apparatus which concerns on 1st Embodiment. The figure which shows the example of the binary image produced | generated by binarizing the input image shown in FIG. When the contour pixel series finally extracted from the input image shown in FIG. 3 is a set of the contour pixel series shown in FIG. 11, examples of the mask image and the extracted image generated by the image extraction unit FIG. The block diagram which shows the function structure of the area | region detection apparatus which concerns on 2nd Embodiment. The figure which shows the example of the input image applied in 2nd Embodiment. The figure which shows all the series of the outline pixel extracted by the outline tracking means from the input image shown in FIG. The figure which shows the series of a part of outline pixel of the group of the series of outline pixels shown in FIG. 17 separately. The figure which shows separately the series of the remaining outline pixels of the group of outline pixel series shown in FIG. The figure which shows the example of the series of the outline pixel which comprises the convex hull extracted by the convex hull extraction means from one series in the group of the series of outline pixels shown in FIG. The figure which shows the example of the mask image and extraction image which are produced | generated by the image extraction means in the case of the input image shown in FIG. FIG. 17 is a diagram showing an example of a mask image and an extracted image generated by the image extraction unit in the first embodiment for comparison with the second embodiment in the case of the input image shown in FIG. 16. The block diagram which shows the function structure of the area | region detection apparatus which concerns on 3rd Embodiment. The figure which shows the example of the input image applied in 3rd Embodiment. The figure which shows a mode that the pixel by which the edge was not detected between the said 2 pixels among the group of two pairs of adjacent pixels in the input image shown in FIG. 24 was connected with the line segment. The figure which shows collectively all the series of the outline pixel extracted by the outline tracking means from the input image shown in FIG. The figure which shows the series of a part of outline pixel of the group of the series of outline pixels shown in FIG. 24 separately. The figure which shows the series of the remaining outline pixels of the group of outline pixel series shown in FIG. 24 separately. The figure which shows the example of the positional relationship of the pixel used as the object of the light / dark determination by the light / dark determination means, and the nearby pixel. The flowchart which shows the procedure of the light / dark determination process performed by the light / dark determination means in 3rd Embodiment. The present invention relates to a group of contour lines stored in a storage device, including an array indicating a determination result of whether each pixel of a sequence of extracted contour pixels is brighter or darker than neighboring pixels, which is obtained by a light / dark determination means The figure which shows the example of the data structure of information. The flowchart which shows the procedure of the change point detection process performed by the change point detection means in 3rd Embodiment. A data structure of information relating to a group of contour lines stored in the storage device, including an array representing the number of change points on each contour line and an array representing the number of contour pixels of the change points, obtained by the change point detection means FIG. The figure which shows the example of the data structure of the information regarding the group of the outline stored in the memory | storage device 12, including the series of outline pixels which comprise the fragment of an outline. The figure which shows a part of flowchart of the procedure of the contour cutting process performed by a contour cutting means. The figure which shows another part of the flowchart of the procedure of the outline cutting process performed by the outline cutting means. The figure which shows the remainder of the flowchart of the procedure of the contour cutting process performed by a contour cutting means. The figure which shows collectively the series of all the outline pixels including the series of outline pixels of the fragment | piece in which the outline shown in FIG. 26 was cut | disconnected by the outline cutting means. The figure which shows the series of a part of outline pixel of the group of the series of outline pixels shown in FIG. 38 separately. The figure which shows separately the series of the remaining outline pixels of the group of outline pixel series shown in FIG. The figure which shows a part of flowchart which shows the procedure of the connection fragment | piece determination process performed by an outline reconnection means. The figure which shows the remainder of the flowchart which shows the procedure of the connection fragment | piece determination process performed by an outline reconnection means. FIG. 39 is a diagram collectively showing all the contour pixel series after reconnection of the outline pixel series of cut fragments included in the group of outline pixel series shown in FIG. 38. The figure which shows the series of a part of outline pixel of the group of the series of outline pixels shown in FIG. 43 separately. The figure which shows the series of the remaining outline pixels of the group of outline pixel series shown in FIG. 43 separately. The block diagram which shows the function structure of the area | region detection apparatus which concerns on 4th Embodiment.

Hereinafter, embodiments will be described with reference to the drawings.
[First Embodiment]
FIG. 1 is a block diagram showing a hardware configuration of the area detection apparatus according to the first embodiment. The area detection apparatus shown in FIG. 1 is realized using a computer (hereinafter referred to as a PC) 10. The PC 10 includes a CPU 11, a storage device 12, and an input / output control unit 13. The CPU 11, the storage device 12, and the input / output control unit 13 are interconnected by a bus 14.

  The storage device 12 is used to store software groups such as the area detection software 121 and the character recognition software 122, input images, work data, and the like. The CPU 11 reads the area detection software 121 from the storage device 12 and executes it to perform area detection processing for detecting an area surrounded by a frame from the input image. The area detection process includes a process of outputting the area detection result to the storage device 12 or an external device. The CPU 11 also performs character recognition processing for recognizing characters in the region detected by the region detection processing by reading and executing the character recognition software 122 from the storage device 12. The character recognition process includes a process of outputting a character recognition result to the storage device 12 or an external device.

  The input / output control unit 13 is connected to a camera 21 as an image acquisition unit that captures an image. The input / output control unit 13 is also connected to the serial interface 22. The input / output control unit 13 controls input of an image from the camera 21 and serial output to an external device (external device other than the camera 21) via the serial interface 22 (that is, input to an external device). Control the output). The input / output control unit 13 also transfers the image input from the camera 21 to the storage device 12 by direct memory access (DMA), for example. The input / output control unit 13 further serially outputs the execution result of the area detection software 121 or the character recognition software 122 (that is, the area detection result or the character recognition result) by the CPU 11 to the external device via the serial interface 22. Alternatively, the data is output to the storage device 12 via the bus 14.

  FIG. 2 is a block diagram showing a functional configuration of the area detection apparatus according to the first embodiment. The functional configuration of the area detection apparatus shown in FIG. 2 is realized by the CPU 11 reading and executing the area detection software 121.

The region detection apparatus shown in FIG. 2 includes a contour line extraction unit 210, a selection unit 220, and a region output unit 230.
The contour line extraction unit 210 detects a series of contour pixels for each contour line from an image input from the camera 21 (that is, an input image), thereby extracting a set S1 of the contour pixel series.

The selection means 220 extracts one or more contour pixel sequences from the contour pixel sequence set S1, thereby extracting the selected contour pixel sequence set S4 (that is, a subset S4 of the set S1). To do.
The area output unit 230 outputs the selected (extracted) pixel series set S4 to the external device or the storage device 12 as a result of the area detection.

  next. The operation of the region detection apparatus shown in FIG. 2 will be described. In the following description, the i-th element of the one-dimensional array α is expressed as α (i), the j-th element in the horizontal direction of the two-dimensional array β, and the i-th element in the vertical direction as β (i, j). Here, the symbol α represents an arbitrary one-dimensional array variable, and β represents an arbitrary two-dimensional array variable. In the following description, it is assumed that the input image is in principle a monochrome grayscale image (so-called monochrome image).

  FIG. 3 is a diagram illustrating an example of an input image. The input image shown in FIG. 3 includes a character string “10” in a round frame. The input image is an array of pixel values having a horizontal width of w pixels and a height of h pixels, and is stored in the storage device 12. In the following description, the pixel at the coordinates (x, y) of the input image is denoted as pixel (x, y). It is assumed that the pixel value of the pixel (x, y) can be read and written as a two-dimensional array f, that is, f (x, y). However, x is positive in the direction of arrow 31 (right direction) in FIG. 3, and y is positive in the direction of arrow 32 (downward direction) in FIG.

  The contour line extraction unit 210 detects a series of contour pixels for each contour line from the input image shown in FIG. 3, thereby extracting a set S1 of the contour pixel series. In order to extract the set S1, the contour line extraction unit 210 includes an edge detection unit 211 and a contour line tracking unit 212.

  The edge detection unit 211 detects an edge between adjacent pixels in the input image. More specifically, the edge detection unit 211 detects an edge based on a difference in pixel values between adjacent pixels. Here, the edge detection unit 211 detects an edge when the absolute value of the difference between pixel values between adjacent pixels is equal to or greater than the threshold value Te, and does not detect an edge when the absolute value is less than Te.

  The contour line tracking unit 212 tracks the contour line of the pixel block for each pixel block divided by the detected edge. The contour tracking unit 212 extracts a series of contour pixels constituting the contour detected by the contour tracking for each contour corresponding to the pixel block.

  Details of edge detection by the edge detection unit 211 will be described below. Here, a variable representing the presence / absence of an edge in the horizontal direction (x direction) is denoted as h (x, y), and a variable representing the presence / absence of an edge in the vertical direction (y direction) is denoted as v (x, y). .

  The edge detection unit 211 sets a variable h (x, y) and a variable v (x, y) indicating the presence / absence of an edge in the vertical direction from an input image by a method described later, −1 ≦ y ≦ h−1 and 1 ≦. Calculation is performed for all coordinates (x, y) in a range defined by x ≦ w−1.

Whether the variable h (x, y) is detected as a horizontal edge crossing a line segment connecting the pixel (x, y) at the coordinates (x, y) of the input image and the pixel (x, y + 1) adjacent in the vertical direction Indicates whether or not.
For h (x, y), 0 is substituted when a horizontal edge is detected, and 1 is substituted when the horizontal edge is not detected. Therefore h (x, y) is
| F (x, y) −f (x, y + 1) | ≧ Te
Then 0,
| F (x, y) -f (x, y + 1) | <Te
Then it becomes 1.

Whether the variable v (x, y) is detected as a vertical edge crossing a line segment connecting the pixel (x, y) at the coordinate (x, y) of the input image and the pixel (x + 1, y) adjacent in the horizontal direction Indicates whether or not.
For v (x, y), 0 is substituted when a vertical edge is detected, and 1 is substituted when the vertical edge is not detected. Therefore h (x, y) is
| F (x, y) −f (x + 1, y) | ≧ Te
Then 0,
| F (x, y) -f (x + 1, y) | <Te
Then it becomes 1.

  However, the edge detection unit 211 determines that the pixel value f (x, y) is a predetermined value C for a pixel outside the input image, that is, a pixel where x <0 or x ≧ w or y <0 or y ≧ h. H (x, y) and v (x, y) are calculated.

  FIG. 4 illustrates a state in which pixels in the pair of adjacent two pixels in the input image illustrated in FIG. 3 are connected to each other by a line segment in which no edge is detected between the two pixels by the edge detection unit 211. FIG.

  The contour line tracking unit 212 traces the contour line of a block of pixels divided by the detected edge (lateral edge or vertical edge), thereby forming the contour pixel constituting the contour line for each contour line. Detect a series of The contour line tracking unit 212 detects a series of contour pixels for each contour line, thereby extracting a set S1 of contour pixel series representing a set of detected contour lines.

Hereinafter, details of contour detection by the contour tracking means 212 will be described.
The contour line tracking unit 212 uses the following variable group for contour detection: (a1) An integer array g of horizontal width w pixels and height h pixels, similar to the input image
(A2) Integer variables x0 and y0 representing the coordinates of the tracking start pixel (that is, the tracking start point) of the contour line
(A3) Integer variables x and y representing the coordinates of the currently tracked pixel (contour point)
(A4) Integer variables d0, d1, and d2 representing the up, down, left, and right directions as 0 for the right direction, 1 for the upward direction, 2 for the left direction, and 3 for the downward direction
(A5) Integer variable N1 representing the number of the contour line being tracked (and the number of contour lines that have been tracked)
(A5) Integer variable M1 representing the number of contour pixels extracted (enumerated) (the number of contour pixels)
(A6) Array variable (array) p1 for storing the two-dimensional coordinates of the extracted (enumerated) contour pixels
(A7) An integer array variable (array) r1 in which the element r1 (n) represents the position in the array p1 of the first pixel (that is, the start point) of the series of pixels constituting the nth tracked contour line
(A8) The element m1 (n) is an integer array variable (array) indicating the number of contour pixels (number of contour pixels) included in a series of pixels (that is, contour pixels) constituting the nth tracked contour line ) M1
Is used. These variable groups are stored in the storage device 12.

  Next, the procedure of the contour line extraction process executed by the contour line tracking unit 212 will be described with reference to the flowcharts of FIGS. In this contour line extraction process, the contour pixels constituting the contour line are extracted (enumerated) as described below.

  First, the outline tracking unit 212 initializes N1 to 0 and M1 to 0 (step 501). The contour tracking unit 212 initializes all elements of the array g to 0 (step 502). The contour tracking unit 212 initializes y0 to 0 (step 503).

Next, the contour line tracking unit 212 repeats the following process of loop A (step 504) until y0 reaches h.
First, the contour tracking unit 212 initializes x0 to 0 (step 505). Next, the contour tracking unit 212 repeats the following process of loop B (step 506) until x0 reaches w.

  First, the contour tracking unit 212 determines whether g (x0, y0) = 0 and h (x0, y0) = 0 (step 507). If the determination in step 507 is Yes, the contour line tracking means 212 substitutes 0 for m1 (N1) and M1 for r (N1) (step 508). The contour tracking unit 212 substitutes x0 for x and y0 for y (step 509). The contour tracking means 212 substitutes 2 for d0 (step 510). Next, the contour line tracking unit 212 substitutes d0 for d1 (step 511).

Then, the contour line tracking unit 212 repeats the following process of loop C (step 601) until x = x0, y = y0, and d1 = d0.
First, the contour tracking means 212 substitutes coordinates (x, y) for p1 (M1) (step 602). By repeating this step 602, pixels (x, y) at coordinates (x, y) are enumerated (extracted) as contour pixels.

  Next, the contour tracking unit 212 substitutes m1 (N1) +1 for m1 (N1) (step 603) and substitutes M1 + 1 for M1 (step 604).

Next, the contour line tracking unit 212 has an edge between the pixel (x, y) at the coordinate (x, y) and the pixel (adjacent pixel) adjacent to the pixel (x, y) in the direction indicated by d2. It is determined whether it is detected (step 605). In step 605, the contour line tracking unit 212 sets d2 to the following value based on the determination result until it indicates a direction in which no edge is detected between adjacent pixels: (d1 + 3) MOD 4
・ D1
・ (D1 + 1) MOD 4
・ (D1 + 2) MOD 4
Change in the order. However, if the pixel in the direction of d2 is outside the range of the input image, the contour line tracking unit 212 regards the value of the adjacent pixel as 0. Here, “a MOD b” represents a remainder obtained by dividing a by b.

The contour tracking unit 212 determines whether an edge is detected between pixels adjacent to the direction indicated by d2 based on the value of d2.
If d2 = 0 and h (x, y) = 0, an edge has been detected.
If d2 = 1 and v (x, y-1) = 0, an edge is detected.
If d2 = 2 and h (x-1, y) = 0, an edge has been detected.
If d2 = 3 and v (x, y) = 0, an edge has been detected.
In other cases, no edge is detected.

After completing the execution of step 605, the contour tracking unit 212 substitutes d2 for d1 (step 606) and substitutes 1 for g (x, y) (step 607).
Next, the contour line tracking unit 212 performs one of the following processes according to the value of d1 (step 608).
If d1 = 0, the contour tracking means 212 substitutes x + 1 for x.
If d1 = 1, the contour tracking means 212 substitutes y−1 for y.
If d1 = 2, the contour tracking means 212 substitutes x−1 for x.
If d1 = 3, the contour tracking means 212 substitutes y + 1 for x.

After step 608, if x = x0, y = y0, and d1 = d0 are not satisfied, the contour line tracking unit 212 executes loop C (step 601) again.
On the other hand, if x = x0, y = y0, and d1 = d0 are established, the contour line tracking unit 212 ends the loop C (step 601).

Then, the contour line tracking unit 212 substitutes N1 + 1 for N1 (step 609).
Next, the contour line tracking unit 212 increases x0 by 1 (step 610). Even when the determination in step 507 is No, the contour line tracking unit 212 increments x0 by 1 (step 610).

After step 610, if x0 = w does not hold, the contour tracking means 212 executes loop B (step 506) again.
On the other hand, if x0 = w is established, the contour tracking unit 212 ends the loop B (step 506).
Then, the contour line tracking unit 212 increases y0 by 1 (step 611).

After step 611, if y0 = h is not established, the contour line tracking means 212 executes loop A (step 504) again.
On the other hand, if y0 = h is established, the contour tracking unit 212 ends the loop A (step 504).

  As a result of the processing of the contour tracking unit 212 described above, the storage device 12 stores information related to a group of extracted contour lines (a series of contour pixels constituting the contour line). FIG. 7 is a diagram illustrating an example of a data structure of information regarding the group of contour lines. This data structure includes an integer variable N1, an integer variable M1, an array variable (array) r1, an array variable (array) p1, and an array variable (array) m1.

  FIG. 8 is a diagram showing all the series of contour pixels extracted by the contour line tracking unit 212 from the input image shown in FIG. FIG. 8 shows a series of eleven contour pixels 801 to 811 (more specifically, a series of contour pixels 801 to 811 constituting eleven contour lines). The series 801 to 811 constitute a set S1 of outline pixel series.

  FIG. 9 is a diagram individually showing the series 801 to 806 of the outline pixel series 801 to 811 shown in FIG. 8, and FIG. 10 is the series 807 to 811 of the outline pixel series 801 to 811. FIG.

  Referring to FIG. 2 again, the selection unit 220 selects one or a plurality of contour pixel series from the set S1 extracted by the contour line extraction unit 210. Then, the selection unit 220 extracts the set S4 including the selected series of contour pixels as a subset S4 of the set S1. Details of selection (extraction) of the contour pixel series (that is, the subset S4) by the selection means 220 will be described below.

The selection means 220 determines, for each of n = 0,..., N1-1, whether or not to select a series of contour pixels that constitute the nth tracked contour line by the following method. .
(1-1) First, the selection unit 220 calculates the area surrounded by the nth tracked contour line. When the calculated area is negative, that is, when the contour tracked in the nth form an outer contour of a block of pixels divided by edges, the selection unit 220 selects the pixels constituting the contour line. It is determined that no series is selected. In the case of the set of contour pixel series (contour line) shown in FIG. Sequence 808 of sequences 808 and 809 and sequence 810 of sequences 810 and 811 are not selected to form the outer contour.

The area surrounded by the nth tracked contour is the following value for each of i = 0,..., M1 (n) −1: (p1 (r1 (n) + i) .x
+ P1 (r1 (n) + i + 1 MOD m1 (n)). x)
X (p1 (r1 (n) + i) .y-p1 (r1 (n) + i + 1) .y) / 2
It is calculated by adding together. Here, p1 (r1 (n) + i). x indicates the x component of the two-dimensional coordinates stored in p1 (r1 (n) + i), and p1 (r1 (n) + i). y represents the y component of the two-dimensional coordinate stored in p1 (r1 (n) + i). p1 (r1 (n) + i + 1). The same applies to.

  (1-2) Next, the selection unit 220 determines that p = 0 (r1 (n) + i) .i regarding i = 0,..., M1 (n) -1. A minimum value xmin (n) and a maximum value xmax (n) of x are calculated. Then, the selection unit 220 calculates the width xmax (n) −xmin (n) +1 of the nth tracked contour line.

  If the calculated width xmax (n) −xmin (n) +1 is less than the predetermined threshold value wmin or exceeds the predetermined threshold value wmax, the selection unit 220 determines that the nth Do not select the series of contour pixels that make up the second tracked contour. When wmin is 16, as will be described later, the above-described series 807, 809, and 811 that do not form the outer contour are not selected.

  (1-3) Similarly, the selection unit 220 determines that p = 0 (r1 (n) + i). A minimum value ymin (n) and a maximum value ymax (n) of y are calculated. Then, the selection unit 220 calculates the height ymax (n) −ymin (n) +1 of the nth tracked contour line.

  If the calculated height xmax (n) −xmin (n) +1 is less than the predetermined threshold value hmin or exceeds the predetermined threshold value hmax, the selection means 220 will A series of contour pixels constituting the nth tracked contour line is not selected.

  The wmax and hmax are set in order to prevent the selection of the contour pixel series of the contour line irrelevant to the frame, such as the contour pixel series 801 of the contour line corresponding to the periphery of the input image. On the other hand, wmin and hmin are set in order to prevent selection of a contour pixel series of a contour line of a small region that is not used as a frame.

(1-4) If none of the above (1-2) and (1-3) applies, the selection means 220 selects a series of contour pixels that constitutes the nth tracked contour line.
In this way, the selection unit 220 selects a series of contour pixels that meet the selection condition from the set S1, and stores the selected series of contour pixels in the set (subset) S4.

  The data structure in the storage device 12 of the set S4 is the same as that of the set S1. That is, the data structure of the set S4 is the same as that of S1 except that the variables M4, N4, r4, p4, and m4 are used instead of the variables M1, N1, r1, p1, and m1. The same applies to the set S2, which will be described later, and variables M2, N2, r2, p2, m2 are used instead of the variables M1, N1, r1, p1, m1. The same applies to the set S3 described later, and variables M3, N3, r3, p3, and m3 are used instead of the variables M1, N1, r1, p1, and m1.

  FIG. 11 is a diagram illustrating an example of a set S4 of contour pixel series extracted by the selection unit 220 from the contour pixel series 801 to 811 shown in FIGS. 9 and 10 under the above-described selection conditions. The selection conditions include, for example, wmin = 16, hmin = 16, wmax = 34, and hmin = 34. In FIG. 11, a series of contour pixels (that is, a series of selected contour pixels) constituting the set S4 is represented by a solid line, and a series of contour pixels not selected is represented by a broken line. In the example of FIG. 11, the set S4 includes two contour pixel series 803 and 805, and represents a region detected by the region detection device.

  Note that the selection unit 220 may extract a set including only the series having the largest area among the sets extracted by the above-described method as the set S4. Further, the selection unit 220 may include a series of all the contour pixels constituting the set S1 in S4. In other words, by setting S4 = S1, not only a series of contour pixels in a region surrounded by a frame, but also a sequence of contour pixels in another region may be extracted. As described above, the selection unit 220 may select any method of selecting a contour pixel series as long as conditions such as the load of subsequent processing and available computer resources permit.

  When the selection unit 220 extracts the set S4 of contour pixels, the region output unit 230 functions as a region acquisition unit, and the set of contour pixels S4 is obtained from the information on the region detected from the input image ( That is, it is acquired as a result of area detection. The area output unit 230 outputs the acquired area detection result (that is, the set S4) to an external device (external device) or the storage device 12 of the area detection apparatus. Here, the output of the set S4 by the area output means 230 is the variable M4, the variable M4, This is done by making a copy of N4, r4, p4, m4. Note that the output of the set S4 by the area output unit 230 is performed by transmitting the contents of the variables M4, N4, r4, p4, and m4 to the external device via the input / output control unit 13 and the serial interface 22, for example. It doesn't matter. That is, the output of the set S4 by the region output means 230 may be performed by another method as long as the set S4 is performed by a method used for information processing in some form.

  FIG. 12 is a block diagram showing a functional configuration of a character recognition mechanism added to the area detection apparatus according to the first embodiment. The functional configuration of the character recognition mechanism shown in FIG. 12 is realized by the CPU 11 reading and executing the character recognition software 122. Note that a character recognition mechanism may be provided in an external device. In this case, the output of the set S4 by the area output unit 230 may be performed on the external device.

The character recognition mechanism shown in FIG. 12 includes binarization means 240, image extraction means 250, recognition means 260, and recognition result output means 270.
The binarization unit 240 binarizes the input image and generates a binary image.

  The image extraction unit 250 extracts a corresponding image from the binary image for each of the outline pixel series included in the set S4 output by the region output unit 230. That is, the image extraction means 250 extracts an image of the area detected by the area detection device shown in FIG.

  The recognizing unit 260 recognizes characters on the image extracted by the image extracting unit 250 (that is, characters in the region detected by the region detecting device). The recognition result output unit 270 outputs the result recognized by the recognition unit 260 (recognition result) to the storage device 12 or an external device.

Details of the operation of the character recognition mechanism shown in FIG. 12 will be described below.
The binarization unit 240 binarizes the input image and generates a binary image. For this binarization, for example, a well-known Niblack algorithm is used. FIG. 13 is a diagram illustrating an example of a binary image generated by binarizing the input image illustrated in FIG. 3.

  The image extraction means 250 extracts pixels inside the region indicated by the series of contour pixels included in the set S4 (that is, the region detected by the region detection device) from the binary image. The image extraction unit 250 then fills the pixels outside the detected area with white pixels, thereby generating an extracted image in which the pixels outside the area are filled with white pixels. The generation of the extracted image by the image extracting unit 250 will be described more specifically.

  First, the image extraction means 250 uses a well-known solid scan conversion algorithm and the pixel value in the region (detected region) surrounded by the series of contour pixels included in the set S4 is 1, and the region A mask image having an outer pixel value of 0 is generated. Then, the image extracting unit 250 generates an extracted image in which the pixel value at the coordinate whose pixel value is 1 on the mask image matches that of the input image, and the pixel values of the remaining pixels indicate white.

  FIG. 14 shows a mask image and an extracted image generated by the image extraction means 250 when the series of contour pixels finally extracted from the input image shown in FIG. 3 is the series 803 and 805 shown in FIG. It is a figure which shows the example of. Here, the mask image is shown in FIG. 14A, in which a pixel having a pixel value of 1 is represented in white and a pixel having a pixel value of 0 is represented in gray. An area of a pixel having a pixel value of 1 (that is, a pixel represented by white) is an area inside the outline indicated by the series 805 shown in FIG. 11, and is included in the input image shown in FIG. Corresponds to the area within the circle. On the other hand, the extracted image is shown in FIG. The extracted image illustrated in FIG. 14B includes the character string “10” within the round frame included in the input image illustrated in FIG.

  The recognizing unit 260 recognizes a character (character string) on the character string image by a known character recognition process as follows. First, the recognition unit 260 extracts a connected component of black pixels from the extracted image by labeling, and then extracts a partial image from the circumscribed rectangle of the connected component. Next, the recognition unit 260 recognizes the extracted partial image by the partial space method, and acquires a character code corresponding to the recognition result.

  The recognition result output unit 270 outputs the character code acquired by the recognition unit 260 to the storage device 12. If a plurality of character codes are acquired by the recognition unit 260, the recognition result output unit 270 sorts the plurality of character codes in the order of the center coordinates of the circumscribed rectangle. Then, the recognition result output unit 270 outputs the sorted character code string to the storage device 12 as character string data representing the character recognition result. The recognition result output means 270 may output the character string data and the line feed code (0x0a for ASCII code) to the external device via the input / output control unit 13 and the serial interface 22 as the character recognition result. That is, the output of the character recognition result by the recognition result output unit 270 may be performed by another method as long as the character recognition result is provided by a method used for information processing in some form.

  According to the first embodiment, while an input image is a monochrome gray scale, an edge is detected between adjacent pixels in the input image, and the pixel is detected for each block of pixels divided by the detected edge. By tracing the contour line of the lump of the object, it is possible to detect the contour line forming the region, that is, to detect the region.

  In the first embodiment, it is assumed that the input image is a monochrome gray scale. However, the input image may be in color. When the input image is in color, a two-dimensional or higher vector (hereinafter referred to as a pixel value vector) expressing the color pixel value may be applied to each element of the pixel value array f.

  When applying the pixel value vector, in the first embodiment, instead of the absolute value of the difference between the pixel values calculated by the edge detection unit 211 for comparison with the threshold value Te, The norm of the difference between pixel value vectors between adjacent pixels may be used. The pixel value vector components may be RGB values representing red, green, and blue color components, or binary values representing red and green color components. In short, any combination of values representing the color of the image captured by the camera 21 may be used as the component of the pixel value vector.

  In addition, preprocessing may be performed in which a combination of color components of pixel values input from the camera 21 is converted into another color component more suitable for edge detection. As this preprocessing, for example, conversion from an RGB color system to an HSV color system can be cited.

[Second Embodiment]
Next, a second embodiment will be described.
The hardware configuration of the region detection apparatus according to the second embodiment is the same as that of the first embodiment. However, the configuration of the area detection software 121 is different from that of the first embodiment. On the other hand, the configuration of the character recognition software 122, that is, the configuration of the character recognition mechanism is the same as that of the first embodiment. Considering these points, FIG. 1 and FIG. 12 are used for convenience in the following description.

  FIG. 15 is a block diagram showing a functional configuration of an area detection apparatus according to the second embodiment. 15, components equivalent to those in FIG. 2 are denoted by the same reference numerals, and detailed description thereof is omitted.

  The area detection device shown in FIG. 15 includes a contour line extraction unit 210 and a region output unit 230, similarly to the region detection apparatus (first embodiment) shown in FIG. The region detection apparatus further includes a convex hull extraction unit 280 and a selection unit 221.

  The convex hull extracting means 280 is a series of contour pixels (hereinafter referred to as a first contour pixel series) extracted by the contour extracting means 210 (more specifically, the contour tracking means 212 of the contour extracting means 210). ) To extract each convex hull of the first contour pixel series. Specifically, the convex hull extracting means 280 applies the GrahamScan algorithm to each of the first contour pixel series included in the set S1 to obtain the vertex of the convex hull of the first contour pixel series. A series in which the contour pixels located are arranged in order on the outer ring of the convex hull (hereinafter referred to as a second contour pixel series) is extracted. The convex hull extracting means 280 acquires a set S3 including all the extracted second contour pixel series. That is, the convex hull extracting means 280 acquires a set S3 including all the second contour pixel series extracted corresponding to the first contour pixel series included in the set S1. Information on the series of second contour pixels constituting the set S3 (variables M3, N3, r3, p3, m3) is stored in the storage device 12.

  The selecting unit 221 selects one or a plurality of contour pixel series from the second set of contour pixel series S3 acquired by the convex hull extracting unit 280, thereby setting the selected set of contour pixel series S4. (In other words, a subset S4 of the set S3) is extracted.

Next, features of the second embodiment will be described.
First, it is assumed that the contour line extraction unit 210 extracts a set S1 of contour pixel series from the input image by detecting a series of contour pixels for each contour line as in the first embodiment. FIG. 16 is a diagram showing an example of this input image. In the input image shown in FIG. 16, unlike the input image shown in FIG. 3, a part of the character “0” in the character string “10” in the circle frame is in contact with the inside of the circle frame.

  FIG. 17 is a diagram collectively showing all the contour pixel series extracted by the contour tracking unit 212 from the input image shown in FIG. FIG. 17 shows a series of ten contour pixels 1701 to 1710 (more specifically, a series of contour pixels 1701 to 1710 constituting ten contour lines). The series 1701 to 1710 constitute a set S1 of outline pixel series.

  Here, series 1701 to 1704 correspond to series 801 to 804 in FIG. 8, respectively, and series 1706, 1707, 1708, 1709, and 1710 correspond to series 806, 807, 809, 810, and 811 in FIG. 8, respectively. To do. On the other hand, the series 1705 corresponds to the two series 805 and 808 in FIG. This is because, as described above, the inner contour line of the round frame corresponding to the series 805 in FIG. 8 is in contact with the outer contour line of the character “0”.

  18 is a diagram individually showing the series 1701 to 1706 of the series of contour pixels 1701 to 1710 shown in FIG. 17, and FIG. 19 is a series of 1707 to 1710 of the series of contour pixels 1701 to 1710. FIG.

  FIG. 20 is a diagram illustrating an example of a series of contour pixels constituting a convex hull extracted by the convex hull extracting unit 280 from the series 1705 among the series of contour pixels 1701 to 1710 shown in FIG. Here, as can be inferred from the first embodiment, the set S4 of contour pixel series is extracted from the series of contour pixels (see FIG. 20) constituting the convex hull extracted from the series 1705 and from the series 1703. And a series of contour pixels constituting the convex hull.

  FIG. 21 is a diagram showing an example of a mask image and an extracted image generated by the image extracting unit 250 in the case of the input image shown in FIG. Here, the mask image is shown in FIG. 21A, in which a pixel having a pixel value of 1 is represented in white and a pixel having a pixel value of 0 is represented in gray. This mask image is generated by the image extracting means 250 based on the input image shown in FIG. 16 and the series of contour pixels shown in FIG. 20 (that is, the series of contour pixels constituting the convex hull extracted from the series 1705). It has been done.

  On the other hand, the extracted image is shown in FIG. FIG. 21B shows an example of an extracted image generated based on the input image shown in FIG. 16 and the mask image shown in FIG.

  FIG. 22 is a diagram showing an example of a mask image and an extracted image generated by the image extraction unit 250 in the first embodiment in the case of the input image shown in FIG. 16 for comparison with the second embodiment. It is. Here, the mask image is shown in FIG. 22 (a), and the extracted image is shown in FIG. 22 (b). The mask image shown in FIG. 22 (a) is generated based on the input image shown in FIG. 16 and the contour pixel series shown in FIG. 18 (e), and the pixel whose pixel value is 1 is white. Pixels with a pixel value of 0 are represented in gray.

  In the mask image shown in FIG. 22A, the outline surrounding the character “0” and the outline inside the frame are connected, and the character “0” is not enclosed. Therefore, when an extracted image is generated based on the mask image shown in FIG. 22A and the input image shown in FIG. 16, the character “0” disappears and the character string “0” is lost as shown in FIG. Only the character “1” of “10” remains. In the second embodiment, the series of contour pixels constituting the convex hull is applied because of such disappearance of the character (more specifically, the contour line of the region to be detected is in contact with the frame. This is for the purpose of preventing the loss of characters caused by the above.

  According to the second embodiment, by extracting the convex hull of each of the series of contour pixels constituting the extracted contour line, even if the frame and the characters in the frame are in contact with each other, The distinction between the inside and the outside can be made more reliable, and the detection performance of the area surrounded by the frame can be improved.

[Third Embodiment]
Next, a third embodiment will be described.
The hardware configuration of the area detection apparatus according to the third embodiment is the same as that of the first embodiment. However, the configuration of the area detection software 121 is different from that of the first embodiment. On the other hand, the configuration of the character recognition software 122, that is, the configuration of the character recognition mechanism is the same as that of the first embodiment. Considering these points, FIG. 1 and FIG. 12 are used for convenience in the following description.

  FIG. 23 is a block diagram illustrating a functional configuration of an area detection device according to the third embodiment. In FIG. 23, components equivalent to those in FIG. 2 are denoted by the same reference numerals, and detailed description thereof is omitted.

  The area detection device shown in FIG. 23 includes a contour line extraction unit 210 and a region output unit 230, similarly to the region detection apparatus (first embodiment) shown in FIG. The area detection apparatus shown in FIG. 23 further includes light / dark determination means 290, change point detection means 300, contour correction means 310, and selection means 222.

  Based on the set S1 of the contour pixel series extracted by the contour line extraction unit 210 (more specifically, the contour line tracking unit 212 of the contour line extraction unit 210), the brightness / darkness determination unit 290 determines that the series of the contour pixels is a neighborhood. It is judged whether it is brighter or darker than the other pixels.

  The change point detection unit 300 detects a set of points (pixels) on the contour pixel series where the brightness determined by the brightness determination unit 290 changes on the contour line indicated by the corresponding contour pixel sequence.

  The contour correcting means 310 reconnects the contour lines at the points detected by the change point detecting means 300, and acquires a set S2 of series of pixels (contour pixels) constituting the reconnected contour lines.

  The selecting unit 222 selects one or a plurality of contour pixel series from the set S2 acquired by the contour correcting unit 310, thereby selecting the set S4 of the selected contour pixel series (that is, a subset of the set S3). S4) is extracted.

next. The operation of the region detection apparatus shown in FIG. 23 will be described.
Here, it is assumed that a region surrounded by a frame is detected from the input image shown in FIG. 24 by the region detection apparatus shown in FIG. In the input image shown in FIG. 24, unlike the input image shown in FIG. 3, so-called whiteout occurs in a part of the frame (round frame). That is, the series of pixels that should be detected as black pixels gradually changes to white pixels.

The edge detection unit 211 of the contour line extraction unit 210 detects an edge between adjacent pixels in the input image shown in FIG.
FIG. 25 shows a state in which pixels in the pair of two adjacent pixels in the input image shown in FIG. 24 are connected by a line segment where no edge is detected between the two pixels by the edge detection unit 211. FIG.

  The outline tracking unit 212 of the outline extraction unit 210 tracks the outline of the pixel block for each pixel block divided by the edge detected by the edge detection unit 211. The contour tracking unit 212 extracts a series of contour pixels constituting the contour detected by the contour tracking for each contour corresponding to the pixel block.

  FIG. 26 is a diagram collectively showing all contour pixel series extracted by the contour line tracking unit 212 from the input image shown in FIG. FIG. 26 shows a series of eight contour pixels 2601 to 2608 (more specifically, a series of contour pixels 2601 to 2608 constituting eight contour lines). The series 2601 to 2608 constitute a set S1 of outline pixel series.

  Here, series 2603 to 2608 correspond to series 806 to 811 in FIG. 8, respectively. On the other hand, the series 2601 corresponds to the two series 802 and 803 in FIG. 8, and the series 2602 corresponds to the two series 804 and 805 in FIG. The reason is that, as described above, whiteout occurs in the round frame portions corresponding to the series 802 and 803 and 804 and 805 in FIG. In FIG. 26, a series of contour pixels corresponding to the series 801 in FIG. 8 is omitted.

  FIG. 27 is a diagram individually showing the series 2601 to 2606 of the outline pixel series 2601 to 2608 shown in FIG. 26, and FIG. 28 is a series 2607 and 2608 of the outline pixel series 2601 to 2608. FIG.

  The light / dark determination means 290 determines whether each of the contour pixels in the contour pixel series included in the set S1 is brighter or darker than the neighboring pixels. Specifically, the brightness determination unit 290 determines whether the contour pixel is brighter or darker than neighboring pixels (pixels within a predetermined range), the pixel value of the contour pixel, and the upper and lower sides of the contour pixel. The determination is made by comparing the average value of the pixel values inside a square including D pixels on the left and right (that is, a region of 2D + 1 in the vertical and horizontal directions centering on the contour pixel). FIG. 29 shows an example of the positional relationship between a pixel that is subject to light / dark determination by the light / dark determination means 290 and a neighboring pixel. In FIG. 29, a symbol x is attached to a pixel to be determined.

  The above-described light / dark determination by the light / dark determination means 290 is performed for each of the M1 contour pixels. The brightness / darkness determination results of each of the M1 contour pixels are stored in the integer array s having the number of elements M1. This integer array s is stored in the storage device 12.

Hereinafter, the procedure of the brightness determination process executed by the brightness determination means 290 will be described with reference to the flowchart of FIG.
First, the brightness determination means 290 initializes an integer variable i representing the number of the contour pixel to 0 (step 3001).

  Next, the brightness determination means 290 determines whether i is M1 or more (step 3002). If i ≧ M1 is not satisfied (No in step 3002), the light / dark determination means 290 determines the pixel value of the pixel at the coordinates (p1 (i) .x, p1 (i) .y) and the top, bottom, left and right of the pixel. The difference Δ (i) from the average value of the pixel values inside the square including the D pixel is calculated (step 3003).

  Next, the brightness determination means 290 determines whether Δ (i) is positive (step 3004). If Δ (i) is positive (Yes in step 3004), the brightness / darkness determining means 290 substitutes 1 indicating “bright” for s (i) (step 3005), and otherwise (in step 3004). No), 0 indicating “dark” is substituted for s (i) (step 3006). s (i) is stored in the storage device 12 in association with the coordinates (p1 (i) .x, p1 (i) .y) of the array (array variable) p1 shown in FIG.

  Next, the brightness determination means 290 substitutes i + 1 for i (step 3007) and returns to step 3002. When the above process is repeated M1 times and i ≧ M1 is satisfied (Yes in step 3002), the light / dark determination unit 290 ends the light / dark determination process.

  FIG. 31 is an array showing a determination result of whether each pixel in the array p1 (that is, each pixel in the series of extracted contour pixels) obtained by the light / dark determination means 290 is brighter or darker than neighboring pixels. It is a figure which shows the example of the data structure of the information regarding the group of the contour line stored in the memory | storage device 12 containing (array variable) s1.

  The change point detection unit 300 detects a point (hereinafter referred to as a change point) where the light and dark determined by the light and dark determination unit 290 changes on the contour line indicated by the corresponding contour pixel series on the contour pixel series. To do. The change point detection means 300 stores the detected change point in the storage device 12 as the order number of the change points on the contour pixel series of the corresponding contour line.

Hereinafter, the procedure of the change point detection process executed by the change point detection means 300 will be described with reference to the flowchart of FIG.
First, the change point detection means 300 initializes an integer variable j representing the contour number to 0 (step 3201). In the following description, the contour line indicated by j may be referred to as a contour line j.

  Next, the change point detection means 300 determines whether j is N1 or more (step 3202). If j ≧ N1 is not satisfied (No in step 3202), 0 is substituted into an integer array variable e (j) indicating the number of change points on the contour line j (step 3203). Next, the change point detection means 300 determines whether s (m1 (j) -1) and s (0) are different (step 3204). That is, the change point detection unit 300 determines whether the end point (last pixel) and the start point (first pixel) of the contour line j are different from each other.

  If s (m1 (j) -1) ≠ s (0) (Yes in step 3204), the change point detection unit 300 determines that the start point of the contour line j is a change point with respect to the end point. To do. In this case, the change point detection means 300 substitutes 0 for an integer array variable c (e (j) + r1 (j)) representing the number of the contour pixel on the contour line j of the change point (step 3205), e (J) is increased by 1 (step 3206). Then, the change point detecting means 300 initializes an integer variable i representing the number of the contour pixel on the contour line j to 1 (step 3207). In the following description, the contour pixel indicated by i may be referred to as a contour pixel i. On the other hand, if s (m1 (j) −1) ≠ s (0) is not satisfied (No in step 3204), the change point detection means 300 jumps to step 3207 and initializes i to 1.

  Next, the change point detection means 300 determines whether i is equal to or greater than M1 as in step 3002 (step 3208). If i ≧ M1 is not satisfied (No in Step 3208), the change point detection unit 300 determines whether s (i−1) and s (i) are different (Step 3209). That is, the change point detection unit 300 determines whether the brightness of the (i−1) -th pixel and the i-th pixel on the contour line j are different.

  If s (i-1) .noteq.s (i) (Yes in step 3209), the change point detection means 300 determines that the i-th pixel is a change point, and c (e (j) I is substituted into + r1 (j)) (step 3210), and e (j) is incremented by 1 (step 3211). Next, the change point detection means 300 substitutes i + 1 for i (step 3212) and returns to step 3208. On the other hand, if s (i−1) ≠ s (i) is not satisfied (No in step 3209), that is, if s (i−1) = s (i), the change point detecting means 300 is the i th This pixel is determined not to be a change point. In this case, the change point detection unit 300 jumps to step 3212, substitutes i + 1 for i, and returns to step 3208.

  In step 3208, the change point detection means 300 determines whether i ≧ m1 (j). If i ≧ m1 (j) (Yes in Step 3208), the change point detection means 300 substitutes j + 1 for j (Step 3213) and returns to Step 3202. In step 3202, the change point detection means 300 determines whether j ≧ N1. If j ≧ N1 (Yes in step 3202), the change point detection unit 300 ends the change point detection process.

  FIG. 33 shows the contour stored in the storage device 12, which includes the array e representing the number of change points on each contour line and the array c representing the number of the contour pixels of the change points, acquired by the change point detection means 300. It is a figure which shows the example of the data structure of the information regarding the group of a line.

  The contour correcting means 310 reconnects the contour lines at the change points detected by the change point detecting means 300. Then, the contour correcting means 310 acquires a set S2 of a series of pixels (contour pixels) constituting the reconnected contour. The contour line correcting unit 310 includes a contour line cutting unit 311 and a contour line reconnecting unit 312 for reconnecting (that is, correcting) the above-described contour line.

  The contour cutting means 311 cuts the corresponding contour pixel series (that is, the contour line) at the change point detected by the change point detection means 300. The contour cutting means 311 obtains a series of contour pixels that constitute a fragment of the cut contour line by cutting the contour line. The acquired series of contour pixels constituting the fragment of the contour line is stored in the storage device 12.

  The contour reconnecting means 312 is an end point in a series (one or more series) of contour pixels constituting one or a plurality of fragments (that is, one or more fragments) of the contour line acquired by the contour cutting means 311. Connect together. By this connection (that is, fragment connection), the contour reconnecting unit 312 acquires a series of pixels constituting the connected (corrected) contour.

  In the third embodiment, the coordinates of pixels (contour pixels) included in a series of contour pixels that form a fragment of the contour line are stored in the array p8 having the number of elements M1. In addition, in the array m8 in which the number of pixels (contour pixel number) included in the series of pixels constituting the fragment of the contour line is an integer, the position (start position) in p8 of the start point of the series (that is, the fragment) is Stored in an integer array r8. FIG. 34 is a diagram illustrating an example of a data structure of information relating to a group of contour lines stored in the storage device 12, including a series of contour pixels constituting a contour line fragment.

Next, the procedure of the contour cutting process executed by the contour cutting means 311 will be described with reference to the flowcharts of FIGS.
First, the outline cutting means 311 initializes a variable j representing the number of the outline being cut to 0 (step 3501). In step 3501, the contour cutting means 311 initializes k to 0 and N8 to 0, respectively. k is an integer variable representing the position (number) of the start point of the fragment from which the contour line is cut in the contour pixel series of the contour line, and N8 is an integer variable representing the number of elements of the array p8.

  Next, the outline cutting means 311 determines whether j is N1 or more (step 3502). If j ≧ N1 is not satisfied (No in step 3502), the contour cutting means 311 determines whether e (j) is 0 (step 3503). That is, the contour cutting means 311 determines whether the number of change points on the contour j is zero.

  If e (j) = 0 (Yes in step 3503), the contour cutting means 311 sets m1 (j) to m8 (N8), k to r8 (N8), 0 to i, respectively. Substitution is performed (steps 3504, 3505, 3506).

  Next, the outline cutting means 311 determines whether i is greater than or equal to m1 (j) (step 3507). If i ≧ m1 (j) is not satisfied, the contour cutting means 311 determines that p8 (k). x is p1 (k). x, p8 (k). p1 (k). y is substituted respectively (steps 3508 and 3509). That is, the contour cutting means 311 copies the coordinates of the kth pixel on the contour j to the array p8 from the array p1. Next, the contour cutting means 311 increments i and k by 1 (step 3510).

  Then, the outline cutting means 311 returns to step 3507 and determines whether i is m1 (j) or more. If i ≧ m1 (j) (Yes in step 3507), the contour cutting means 311 increases N8 by 1 (step 3511), and then increases j by 1 (step 3512). Then, the outline cutting means 311 returns to step 3502.

  On the other hand, if it is determined in step 3503 that e (j) = 0 is not satisfied, the contour cutting unit 311 proceeds to step 3601. In step 3601, the contour cutting unit 311 determines whether c (r1 (j)) is zero. If c (r1 (j)) = 0 (Yes in step 3601), the contour cutting means 311 represents an integer representing the number on the contour of the change point immediately before the last segment cut out on the contour. E (j) -1 is substituted into the variable ea (step 3602). Then, the outline cutting means 311 proceeds to step 3706 described later.

  On the other hand, if c (r1 (j)) = 0 is not satisfied (No in step 3601), the contour cutting means 311 substitutes c (e (j) -1) for si2 (step 3603). The contour cutting means 311 substitutes c (0) -1 for ei3 and e (j) -2 for ea (steps 3604 and 3605).

  When executing step 3605, the contour cutting means 311 substitutes 0 for m8 (N8) and si2 for i (steps 3606 and 3607). Next, the outline cutting means 311 determines whether i is greater than or equal to m1 (j) as in step 3507 (step 3608).

  If i ≧ m1 (j) is not satisfied (No in step 3608), the contour cutting means 311 determines that p8 (k). x is p1 (r1 (j) + i). x, p8 (k). y is p1 (r1 (j) + i). y is substituted respectively (steps 3609 and 3610). The contour cutting means 311 increments i, k and m8 (N8) by 1 (step 3611).

  Then, the contour cutting means 311 returns to step 3608 and determines whether i is greater than or equal to m1 (j). If i ≧ m1 (j) (Yes in step 3608), the contour cutting means 311 proceeds to step 3701. In step 3701, the contour cutting means 311 substitutes 0 for i.

  Next, the outline cutting means 311 determines whether i is larger than ei3 (step 3702). If i> ei3 is not satisfied (No in step 3702), the contour cutting means 311 determines that p8 (k). x is p1 (r1 (j) + i). x, p8 (k). y is p1 (r1 (j) + i). y is substituted respectively (steps 3703 and 3704). Next, the contour cutting means 311 increments i, k, and m8 (N8) by 1 (step 3705), and returns to step 2702. On the other hand, if i> ei3 (Yes in step 3702), the contour cutting means 311 proceeds to step 3706.

  In step 3706, the contour cutting means 311 assigns 0 to an integer variable a representing the number on the contour of the change point immediately before the newly cut contour segment. Next, the contour cutting means 311 determines whether a is larger than ea (step 3707). If a> ea (Yes in step 3707), the contour cutting means 311 jumps to step 3512. The contour cutting means 311 then increases j by 1 (step 3512) and returns to step 3502.

  On the other hand, if a> ea is not satisfied (No in step 3707), the contour cutting unit 311 substitutes c (a) for i and c (a) + e (a) -1 for ei4 (step 3708). ). Next, the outline cutting means 311 determines whether i is larger than ei4 (step 3709). If i> ei4 (Yes in step 3709), the contour cutting means 311 increments a and N8 by 1 (step 3710) and returns to step 3707.

  On the other hand, if i> ei4 is not satisfied (No in step 3709), the contour cutting means 311 determines that p8 (k). x is p1 (r1 (j) + i). x, p8 (k). y is p1 (r1 (j) + i). y is substituted respectively (steps 3711 and 3712). The contour cutting means 311 increments i, k and m8 (N8) by 1 (step 3713). Then, the contour cutting means 311 returns to Step 3709.

  As described above, the contour cutting means 311 returns to step 3502 when j is increased by 1 in step 3512. If j ≧ 1 as a result of increasing j by 1 (Yes in step 3502), the contour cutting means 311 ends the contour cutting processing.

By the processing of the contour cutting means 311 described above, a fragment is generated by cutting each of the contour lines extracted by the contour line extracting means 210 at each change point detected by the change point detecting means 300. , Stored in array variables p8, m8 and r8.
FIG. 38 shows all the contour pixels including the contour pixel series of the fragments in which the contour line shown in FIG. 26 (that is, the contour line extracted from the input image shown in FIG. 24) is cut by the contour cutting means 311. FIG. In FIG. 38, the same reference numbers are assigned to the same contour pixel series as in FIG. In FIG. 38, there are two contour pixel series 2601a and 2601b obtained by cutting the contour line composed of the contour pixel series 2601 shown in FIG. 26, and a contour line composed of the contour pixel series 2602. Shown is a series 2602a and 2602b of contour pixels of two cut pieces. FIG. 38 also shows the contour pixel series 2603 to 2608 shown in FIG.

  FIG. 39 is a diagram individually showing sequences 2601a, 2601b, 2602a, 2602b, 2603, and 2604 of the contour pixel series shown in FIG. 38, and FIG. 40 is a diagram of the contour pixel series shown in FIG. It is a figure which shows the series 2605 thru | or 2608 separately.

  As described above, the contour reconnecting unit 312 connects (that is, connects) the end points of a series of contour pixels that form one or more fragments of the contour acquired by the contour cutting unit 311. A series of pixels constituting the corrected contour line is acquired. A set S <b> 2 of a series of pixels constituting the acquired outline is stored in the storage device 12.

  In order for the contour reconnecting means 312 to connect the fragments of the contour line, it is necessary to determine the fragments to be connected after each fragment in the order stored in the storage device 12. Hereinafter, the procedure of connection fragment determination processing executed by the contour reconnecting means 312 will be described with reference to the flowcharts of FIGS. 41 and 42.

  First, the contour reconnecting means 312 substitutes 0 for F (j) for each of j = 0,..., N8-1 (step 4101). Here, F is an integer array with the number of elements N8, and F (j) indicates the jth element of the array F. Next, the contour reconnecting means 312 substitutes 0 for each of the integer variables k and j (steps 4102 and 4103).

  Next, the outline reconnecting means 312 determines whether j is N8 or more (step 4104). If j ≧ N8 is not satisfied as in this example (No in step 4104), the contour reconnecting unit 312 determines whether F (j) is 0 (step 4105). If F (j) = 0 as in this example (Yes in step 4105), the contour reconnecting means 312 is a series of adjacent pixels of the jth fragment starting from the pixel indicated by r8 (j). Is obtained from p8, and the series of adjacent pixels of the j-th fragment is stored in the set S2 as the k-th series of the set S2 (step 4106). Here, the number of adjacent pixels constituting the series of adjacent pixels of the jth fragment is indicated by m8 (j), and the series of adjacent pixels stored in the set S2 is the position of p8 (r8 (j)). Are composed of m8 (j) pixels starting from.

  Next, the contour reconnecting means 312 substitutes 1 for F (j) to indicate that the series of adjacent pixels of the jth fragment has already been stored in the set S2 (step 4107). Next, the contour reconnecting means 312 substitutes 0 for the integer variable flag (step 4108). On the other hand, if F (j) = 0 is not satisfied (No in step 4105), that is, if F (j) = 1, the contour reconnecting unit 312 sets the adjacent pixel series of the jth fragment to the set S2. Judge that it has already been stored. In this case, the contour reconnecting means 312 substitutes 0 for the integer variable flag (step 4108).

  The contour reconnecting means 312 substitutes 0 for the integer variable flag (step 4108), and substitutes 0 for score (j2) for each of j2 = 0,..., N8-1 (step 4109). score is an array of real numbers with N8 elements, and score_ (j2) indicates the j2th element of the array score.

  Next, the outline reconnecting means 312 substitutes −1 for the real variable score_max and 0 for the integer variable j2_opt (step 4110). The contour reconnecting means 312 substitutes 0 for the integer variable j2 (step 4111).

  Next, the contour reconnecting means 312 determines whether j2 is N8 or more (step 4201). If j2 ≧ N8 is not satisfied as in this example (No in step 4201), the contour reconnecting means 312 is F (j2) = 0 and s1 (r8 (j2)) = s1 (r8 (j)) And whether the Euclidean distance ud (j2, j) between the first contour pixel on the j2th fragment and the last contour pixel of the jth fragment is smaller than a predetermined positive constant ε (step) 4202). Here, the fact that s1 (r8 (j2)) = s1 (r8 (j)), that is, the lightness and darkness of the start point of the j2th fragment matches the lightness and darkness of the start point of the jth fragment is It shows that the contrast of the upper contour pixel with the neighboring contour pixels matches that on the jth fragment. Further, ud (j2, j) <ε is defined as a range in which the positions of the end point (for example, the start point) on the j2th fragment and the end point (for example, the end point) of the jth fragment are predetermined (that is, the vicinity). Range).

  If the determination in step 4202 is Yes, the contour reconnecting means 312 calculates a vector VA from the last contour pixel (that is, the end point) of the jth fragment to the last δth contour pixel ( Step 4203). δ is a predetermined natural number. The contour reconnecting unit 312 calculates a vector VB from the first contour pixel (that is, the start point) of the j2th fragment to the δth contour pixel from the beginning (step 4204).

  Next, the contour reconnecting means 312 calculates the angle θ formed by the vector VA and the vector VB by the cosine theorem (step 4205). Here, it is assumed that the unit of the angle θ is “degree”. Next, the contour reconnecting means 312 determines whether the angle θ is larger than score_max (step 4206).

  In step 4209, the angle θ formed by the vector VA and the vector VB is the largest for the j-th fragment of j2 being 0,..., N8-1, respectively (that is, closest to 180 degrees). In order to detect the fragments, the processing is repeated for the fragments satisfying the condition shown in step 4202. When the fragment having the largest angle θ between the vector VA and the vector VB with respect to the jth fragment is the j2_optth fragment, the end points (starting points in this case) of the j2_optth fragment are 0,..., N8-1. It can be said that the end point of the jth fragment is the most facing to the end point (here, the end point) of the jth fragment.

  If θ> score_max (Yes in step 4206), the contour reconnecting unit 312 substitutes θ for score_max and j2 for the integer variable j2_opt (step 4207). Next, the contour reconnecting means 312 substitutes 1 for flag (step 4208), and then increases j2 by 1 (step 4209). flag = 1 indicates that there is a fragment to be connected to the jth fragment. The integer variable j2_opt = j2 indicates that the fragment j2 of the fragments 0,..., J2 is optimal as the fragment to be connected to the jth fragment.

  On the other hand, if θ> score_max is not satisfied (No in step 4206), the contour reconnecting means 312 jumps to step 4209 to increase j2 by one. Even when the determination at step 4202 is No, the contour reconnecting means 312 jumps to step 4209 to increase j2 by one.

  The contour line reconnecting means 312 increases j2 by 1 (step 4209), returns to step 4201, and determines whether j2 ≧ N8. If j2 ≧ N8, the contour reconnecting means 312 determines that the fragments to be connected to the jth fragment have been checked for j2 = 0,..., N8-1.

  Therefore, the contour line reconnecting unit 312 determines whether flag is 1 (step 4210). If flag = 1 (Yes in step 4210), the contour reconnecting unit 312 sets the j2_opt-th fragment at the end of the k-th contour pixel series (the fragment of the contour line indicated) of the set S2. A series of contour pixels is added (step 4211).

  In this way, the addition of the j2_opt-th fragment to the k-th contour line fragment is performed by adding the contour pixel series included in the j2_opt-th fragment to the contour pixel series of the contour line. As a result, the j2_opt-th fragment is joined to the k-th contour fragment. Here, the sequence of contour pixels of the j2_opt-th fragment is a sequence of contour pixels of a fragment starting from the pixel indicated by the contour pixel r8 (j2_opt) among the sequence of contour pixels of the fragment stored in p8. , M8 (j), the number of contour pixels.

  The contour reconnecting means 312 adds the j2_opt-th fragment to the k-th contour fragment (step 4211), and substitutes 1 for F (j2_opt) (step 4212). Next, the contour reconnecting means 312 increases k by 1 (step 4112) and increases j by 1 (step 4113). On the other hand, if flag = 1 is not satisfied (No in step 4210), the contour reconnecting means 312 jumps to step 4112 to increase k by 1 and increase j by 1 (step 4113). The contour reconnecting means 312 increases j by 1 (step 4113), returns to step 4104, and determines whether j is N8 or more. If j ≧ N8 (Yes in step 4104), the contour line reconnecting unit 312 ends the contour line reconnection process.

  FIG. 43 is a diagram collectively showing all contour pixel sequences after reconnection of the contour pixel sequence of the cut fragment included in the group of contour pixel sequences shown in FIG. In FIG. 43, the same reference numbers are assigned to the same contour pixel series as in FIG. FIG. 43 shows a contour line pixel sequence 4301 to which the start point and end point of the fragment outline pixel series 2601a shown in FIG. 38 are connected, and a start point and an end point of the fragment outline pixel series 2601b shown in FIG. A series 4302 of contour pixels of the contour line to which is connected is shown.

  The series 4301 and 4302 are generated by storing the series 2601a and 2601b of the outline pixels of the fragment in the set S2 in the above step 4106, respectively. If a series of contour pixels of another fragment is not added to the end of the series 2601a and 2601b, that is, if step 4211 is not executed, the series 2601a and 2601b are only stored in the set S2, and both ends of the series 2601a and the series 2601b Both ends of are connected.

  43 also includes a contour line pixel sequence 4303 to which the start and end points of the fragment outline pixel series 2602a shown in FIG. 38 are connected and a start point and a start point of the fragment outline pixel series 2602b shown in FIG. A series 4304 of contour pixels of the contour line to which the end points are connected is shown. The series 4303 and 4304 are generated by storing the series 2602a and 2602b of the contour pixels of the fragment in the set S2 in the above step 4106, respectively. FIG. 43 also shows the series of contour pixels 2603 to 2608 shown in FIG.

  44 is a diagram individually showing sequences 4301 to 4304, 2603, and 2604 of the contour pixel series shown in FIG. 43, and FIG. 45 is a sequence 2605 of the contour pixel series shown in FIG. 1 to 2608 are diagrams individually showing.

  Note that the method of connecting one or more end points in a series of contour pixels that constitutes a fragment of the contour line matches the fragments whose end points are most directly facing each other, with the lightness and darkness compared with the neighboring pixels matching. As long as it is a technique to connect, it is not limited to the method applied in the third embodiment, and any method may be used.

  The contour reconnecting means 312 substitutes M1 for M2, N8 for N2, and for j = 1,..., N8, r8 (j) for r2 (j) and r8 for m1 (j). (J) may be substituted respectively. That is, the outline reconnecting means 312 may simply connect both ends of the fragment obtained by the outline cutting means 311 as a new outline.

  When the contour line reconnection process by the contour line reconnection unit 312 ends, the selection unit 222 selects one from the set S2 acquired by the contour line correction unit 310 in the same manner as the selection unit 220 in the first embodiment. Alternatively, by selecting a series of a plurality of contour pixels, a set S4 of the selected series of contour pixels is extracted.

  According to the third embodiment, a point (change point) that changes on a contour line indicated by a series of contour pixels is detected, and the contour line is reconnected at the change point, resulting in light reflection or the like. It is possible to realize area detection that is robust against whiteout.

[Fourth Embodiment]
Next, a fourth embodiment will be described.
The hardware configuration of the region detection apparatus according to the fourth embodiment is the same as that of the first embodiment. However, the configuration of the area detection software 121 is different from that of the first embodiment.

  FIG. 46 is a block diagram showing a functional configuration of an area detection apparatus according to the fourth embodiment. 46, components equivalent to those in FIG. 2, FIG. 15, or FIG. 23 are given the same reference numerals, and detailed descriptions thereof are omitted.

  The area detection device shown in FIG. 46 includes a contour line extraction unit 210 and a region output unit 230 in the same manner as the region detection devices shown in FIGS. 2, 15, and 23. The area detection device shown in FIG. 46 also includes light / dark determination means 290, change point detection means 300, and change point detection means 300, similarly to the area detection device shown in FIG. 46 further includes a convex hull extracting means 281 corresponding to the convex hull extracting means 280 in the area detecting apparatus shown in FIG. 15, and selecting means similar to the area detecting apparatus shown in FIG. 221 is provided.

  The convex hull extracting unit 281 extracts each convex hull of the series of contour pixels included in the set S2 acquired by the contour line correcting unit 310 (contour line reconnecting unit 312), and extracts the pixels constituting the convex hull. A set S3 of series is acquired. The convex hull extracting means 281 is the same as the convex hull extracting means 280 applied in the second embodiment except that the set of contour pixels to be convex hull extracted is not the set S1 but the set S2. It is.

  The feature of the region detection device shown in FIG. 46 is that, in the configuration of the region detection device shown in FIG. 23 applied in the third embodiment, between the contour line correction means 310 and the region output means 230, the second embodiment. The convex hull extracting unit 281 and the selecting unit 221 corresponding to the convex hull extracting unit 280 and the selecting unit 221 applied in the above are added.

  As is apparent from the second and third embodiments, according to the fourth embodiment, a region that is robust against whiteout is realized, and the frame and the characters in the frame are in contact with each other. The detection performance of the region surrounded by the frame in the case can be improved.

  According to at least one embodiment described above, it is possible to more reliably distinguish between the inside and outside of the frame when the frame and the characters in the frame are in contact, and to improve the detection performance of the region surrounded by the frame. An area detection apparatus, an area detection method, and a program can be provided.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

  DESCRIPTION OF SYMBOLS 10 ... Computer (PC), 11 ... CPU, 12 ... Storage device, 21 ... Camera, 210 ... Contour extraction means, 211 ... Edge detection means, 212 ... Contour tracking means, 220, 221 ... Selection means, 230 ... Area Output means, 240 ... binarization means, 250 ... image extraction means, 260 ... recognition means, 270 ... recognition result output means, 280,281 ... convex hull extraction means, 290 ... light / darkness determination means, 300 ... change point detection means, 310 ... contour line correcting means, 311 ... contour line cutting means, 312 ... contour line reconnecting means.

Claims (7)

Contour line extraction means for extracting a series of first contour pixels from the input image for each contour line related to the frame detected from the input image ;
A convex hull extracting means for extracting a second contour pixel series constituting a convex hull of the first contour pixel series for each of the extracted first contour pixel series;
An area detection apparatus comprising: an area acquisition unit configured to acquire the extracted second contour pixel series as information of an area surrounded by a frame detected from the input image. A mask image generating unit and an extracted image generating unit;
The convex hull extracting means is configured such that, for each series of the first contour pixels extracted by the contour line extracting means, a contour pixel positioned at a vertex of the convex hull of the first contour pixel series is the convex hull. Extracting a sequence arranged in order on the outer ring as a sequence of the second contour pixels;
The mask image generating means generates a mask image having a pixel value of 1 in an area surrounded by the frame and a pixel value of 0 outside the area;
The extracted image generating means generates an extracted image in which a pixel value of a pixel at a coordinate having a pixel value of 1 on the mask image matches a pixel value of the input image, and a pixel value of the remaining pixels indicates white. Item 1. The area detection apparatus according to Item 1. The contour line extracting means includes
Edge detection means for detecting an edge between adjacent pixels in the input image;
The contour line tracking means for extracting the series of the first contour pixels constituting the contour line for each contour line by tracking the contour line of the block of pixels divided by the edge. Area detection device. In an area detection device comprising an outline extraction means, a convex hull extraction means, and an area acquisition means, an area detection method for detecting an area surrounded by a frame from an input image,
The contour line extracting means extracting a series of first contour pixels from the input image for each contour line related to a frame detected from the input image ;
A convex hull extracting step in which the convex hull extracting means extracts, for each of the extracted first contour pixel series, a second contour pixel series constituting a convex hull of the first contour pixel series; ,
A step of acquiring the region of the extracted second contour pixel as information of a region surrounded by a frame detected from the input image; In the convex hull extracting step, the convex hull extracting means is positioned at the vertex of the convex hull of the first contour pixel series for each series of the first contour pixels extracted by the contour line extracting means. Extracting a series of contour pixels arranged in order on the outer ring of the convex hull as the second contour pixel series;
The region detection apparatus further includes a mask image generation unit and an extraction image generation unit,
The mask image generating means generating a mask image in which the pixel value in the region surrounded by the frame is 1 and the pixel value outside the region is 0;
The extracted image generating means generates an extracted image in which the pixel value of a pixel at a coordinate having a pixel value of 1 on the mask image matches the pixel value of the input image, and the pixel values of the remaining pixels are white. When,
The region detection method according to claim 4, further comprising: Computer
Contour line extraction means for extracting a series of first contour pixels from the input image for each contour line related to the frame detected from the input image ;
A convex hull extracting means for extracting a second contour pixel series constituting a convex hull of the first contour pixel series for each of the extracted first contour pixel series;
A program for causing the extracted second contour pixel series to function as area acquisition means for acquiring information of an area surrounded by a frame detected from the input image. Further causing the computer to function as a mask image generation unit and an extraction image generation unit,
The convex hull extracting means is configured such that, for each series of the first contour pixels extracted by the contour line extracting means, a contour pixel positioned at a vertex of the convex hull of the first contour pixel series is the convex hull. Means for extracting a series arranged in order on the outer ring as a series of the second contour pixels;
The mask image generation means is a means for generating a mask image in which the pixel value in the region surrounded by the frame is 1 and the pixel value outside the region is 0,
The extracted image generating means generates an extracted image in which a pixel value of a pixel at a coordinate having a pixel value of 1 on the mask image matches a pixel value of the input image, and a pixel value of the remaining pixels indicates white. The program according to claim 6.

Images