KR100238319B1 - Image processing method for distinguishing object - Google Patents

Abstract

The image processing method for identifying an object according to the present invention includes (a) receiving light reflected from an image region including an object. (b) The received light is processed and converted into image data representing luminance of each point of the image area. (c) Calculate the mode luminance in the bright region more than the predetermined boundary luminance and the mode luminance in the dark region less than the boundary luminance. (d) New boundary luminances are determined by reflecting the most frequent luminance in the bright region and the dark region in a predetermined division ratio. And (e) converts the image data so that the object is identified from the image area by applying a new boundary luminance.

Description

Image processing method for object identification

The present invention relates to an image processing method for object identification.

A system for imaging an object and processing image data, for example, a machine vision system includes a camera, for example, a charge coupled device (CCD) camera, an image processing board, a monitor, and the like. The camera performs the following functions. First, the light reflected from the image area including the object is received and converted into an analog signal. The image processing board converts an analog signal from a camera into image data, which is a digital signal, and then processes the image data to perform various functions. The image data represents luminance for each point of the image area.

The main function of such an image processing system is to identify the background and the object of the image, or the object and other objects. Accordingly, functions related to the background image data, the first object image data, the second object image data, and the like may be classified.

FIG. 1 shows an image in which an object is picked up on one screen 10. Such an image region includes a background region 101 and an object region 102. P ₁ , P ₂ , and P ₃ represent boundary pixels of the background area 101 and the object area 102. In order to identify the pixels of the coordinates corresponding to the background area 101 and the object area 102 in the image data, it is very important to appropriately set the threshold brightness, that is, the boundary brightness. In order to set such boundary luminance, conventionally, the average luminance in the bright region and the dark region is reflected in a predetermined division ratio to determine a new boundary luminance. Such conventional image processing methods are well described in US Pat. No. 5,138,671.

2 shows a distribution state of luminance for explaining a conventional image processing method. As shown in FIG. 1, when only a background and a single object are captured, when looking at a histogram of luminance, two distribution regions appear around the first boundary luminance M. FIG. An area of luminance lower than the first boundary luminance M represents the background region (101 of FIG. 1), and an region of luminance higher than the first boundary luminance M represents the object region (102 of FIG. 1). Distinguishing between the background image data and the object image data simply by using the first boundary luminance M is not accurate. Accordingly, conventionally, the new boundary luminance is determined by reflecting the average luminances md and mb in the dark region and the bright region in a predetermined division ratio. For example, thn = r ⋅ md + (1-r) ⋅ The new boundary luminance thn is determined according to the equation of mb. Here, r represents the division ratio. This split ratio r is set by experimental learning.

However, there is a tendency that the average luminances md and mb change relatively largely depending on whether the background or the surface of the object is glossy. Accordingly, according to the conventional image processing method, there is a problem that the accuracy of object identification is unstable depending on whether the background or the surface of the object is glossy.

It is an object of the present invention to provide an image processing method capable of object identification by maintaining uniform boundary luminance regardless of whether the background or the surface of an object is glossy.

1 is a schematic diagram illustrating an image in which an object is picked up on one screen.

2 is a distribution diagram of luminance for explaining a conventional image processing method.

3 is a distribution diagram of luminance for explaining an image processing method according to an exemplary embodiment.

4 is a distribution diagram of luminance for describing an image processing method according to another exemplary embodiment of the present invention.

5 is a distribution diagram of luminance obtained by photographing an object having no gloss on its surface.

Fig. 6 is a distribution diagram of brightnesses obtained by photographing a glossy object on its surface.

<Explanation of symbols for main parts of the drawings>

10 ... camera screen, 101 ... background area,

102 object region, P ₁ , P ₂ , P ₃ ... boundary pixels,

M ... first boundary luminance, md, mb ... average luminance,

fd, fb, f ₀ , f ₁ , f ₂ , f ₃ ... mode luminance.

S1 ... Area not included in the actual area.

The image processing method of the present invention for achieving the above object comprises the steps of (a) receiving the light reflected from the image area including the object. (b) The received light is processed and converted into image data representing luminance of each point of the image area. (c) Calculate the mode luminance in the bright region that is more than the predetermined boundary luminance and the mode luminance in the dark region that is less than the boundary luminance. (d) The new boundary luminance is determined by reflecting the most frequent luminance in the bright region and the dark region in a predetermined division ratio. And (e) converting the image data so that the object is identified from the image area by applying the new boundary luminance.

Since the image processing method of the present invention determines the new boundary luminance by reflecting the most frequent luminance at a predetermined division ratio, the uniform boundary luminance is maintained regardless of whether the background or the surface of the object is glossy, thereby identifying the object. can do.

Preferably, step (d) is characterized in that thn = r ⋅ fd + (1-r) ⋅ It is performed according to the formula of fb. Where thn represents the new boundary luminance, r represents the division ratio, fd represents the mode luminance in the dark region, and fb represents the mode luminance in the bright region.

Hereinafter, preferred embodiments of the present invention will be described in detail.

3 illustrates a distribution state of luminance for describing an image processing method according to an exemplary embodiment of the present invention. As described above, when only a background and a single object are captured, when looking at a histogram of luminance, two distribution regions appear around the first boundary luminance M. FIG. An area of luminance lower than the first boundary luminance M represents the background region (101 of FIG. 1), and an region of luminance higher than the first boundary luminance M represents the object region (102 of FIG. 1). It is not accurate to distinguish the background image data and the object image data simply by the first boundary luminance M. FIG. Accordingly, the newest luminance is determined by reflecting the most frequent luminances fd and fb in the dark region and the bright region in a predetermined division ratio. md and mb represent the average brightness in a bright area and a dark area. Here, the most frequent luminance fd and fb mean the luminance shared by the most pixels in the corresponding image area. Equation 1 below is a formula for determining a new boundary luminance thn by reflecting the most frequent luminances fd and fb in a predetermined division ratio.

thn = r⋅fd + (1-r) ⋅fb

Where r is the division ratio (0) ≤ r ≤ 1), fd represents the mode of brightness in the dark region, and fb represents the mode of brightness in the bright region. The division ratio r is appropriately set by experimental learning.

4 illustrates a distribution of luminance for describing an image processing method according to another exemplary embodiment of the present invention. Referring to the drawings, each divided area consisting of four peaks means a background area, a first object area, a second object area, and a third object area from the left. The most frequent luminance of each region is f ₀ , f ₁ , f ₂ and f ₃ . Using these, the equation for obtaining the new boundary luminance of each region is as shown in Equation 2 below.

thn _j = r _j ⋅f _j + (1-r _j ) ⋅f _{j + 1}

here, thn _j Is the boundary luminance between the j th region and the (j + 1) th region, r _j Is the split ratio between the j th region and the (j + 1) th region, f _j Is the mode of luminance in the j th region, and f _{j + 1} Denotes the mode of luminance in the (j + 1) th region. Split ratios between each region r _j Is appropriately set by experimental learning.

For example, the boundary luminance between the background region and the first object region thn ₁ silver thn ₁ = r ₁ ⋅f ₁ + (1-r ₁ ) ⋅f ₂ It is determined by the formula. here, r ₁ Is the split ratio between the background area and the first object area, f ₁ Is the mode of luminance in the background region, and f ₂ Represents the mode of luminance in the first object region.

Also, the boundary luminance between the first object region and the first object region thn ₂ silver thn ₂ = r ₂ ⋅f ₂ + (1-r ₂ ) ⋅f ₃ It is determined by the formula. here, r ₂ Is the split ratio between the first object zone and the second object zone, f ₂ Is the mode of luminance in the first object region, and f ₃ Represents the mode of luminance in the second object region.

In this way, by setting the boundary luminance between the divided regions using the most frequent luminance in each divided region, it is possible to identify the object by maintaining uniform boundary luminance regardless of whether the background or the surface of the object is glossy. Can be. The experimental content to demonstrate this is described below.

FIG. 5 shows a state of distribution of luminance caused by picking up an object having no gloss on its surface. FIG. 6 shows a state of distribution of luminance represented by imaging a glossy object on the surface thereof. Referring to the drawings, it can be seen that the concentration of the luminance distribution is relatively high in the case of the histogram of the object having no gloss on the surface. The images in the distributed state of FIGS. 5 and 6 were obtained by a camera of 256 gray scales. Table 1 below is a table showing the number of pixels for each luminance obtained from the image data of the distribution state of FIG.

Luminance The number of pixels 0, ..., 9 0, 0, 0, 0, 0, 0, 0, 3, 2, 10 10, ..., 19 7, 9, 11, 19, 23, 57, 43, 44, 46, 40 20, ..., 29 23, 32, 49, 85, 88, 44, 112, 145, 98, 90 30, ..., 39 87, 155, 163, 222, 153, 267, 300, 335, 228, 302 40, ..., 49 392, 543, 723, 476, 1363, 1702, 1744, 1228, 1704, 1597 50, ..., 59 1399, 1051, 1048, 1518, 1822, 2272, 2240, 3764, 5319, 7578 60, ..., 69 4708, 10873, 11823, 11914, 11192, 7546, 8610, 6902, 5111, 2055 70, ..., 79 3617, 2939, 2347, 1423, 1507, 1184, 1002, 846, 402, 733 80, ..., 89 726, 658, 478, 664, 625, 554, 344, 315, 354, 293 90, ..., 99 232, 160, 180, 157, 136, 64, 117, 106, 88, 80 100, ..., 109 65, 87, 68, 69, 28, 59, 58, 56, 50, 43 110, ..., 119 49, 58, 64, 20, 55, 37, 46, 35, 40, 22 120, ..., 129 38, 20, 26, 40, 41, 41, 23, 43, 27, 39 130, ..., 139 11, 43, 50, 48, 48, 32, 32, 43, 37, 21 140, ..., 149 34, 28, 27, 25, 36, 36, 35, 34, 14, 28 150, ..., 159 37, 37, 20, 35, 34, 47, 26, 21, 22, 38 160, ..., 169 25, 30, 30, 38, 34, 24, 35, 31, 35, 24 170, ..., 179 25, 22, 36, 34, 13, 21, 19, 28, 52, 119 180, ..., 189 64, 50, 48, 21, 36, 62, 79, 59, 81, 82 190, ..., 199 87, 47, 64, 51, 64, 48, 49, 65, 56, 36 200, ..., 209 19, 32, 16, 8, 9, 5, 0, 3, 5, 0 210, ..., 219 9, 2, 2, 0, 0, 0, 0, 0, 0, 0 220, ..., 229 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 230, ..., 239 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 240, ..., 249 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 250, ..., 255 0, 0, 0, 0, 0, 0

Referring to Table 1, it can be seen that the total number of pixels is 150,100, the average brightness is 64.72788141, and the mode brightness is 63.00.

Table 2 below is a table showing the number of pixels for each luminance obtained from the image data of the distribution state of FIG.

Luminance The number of pixels 0, ..., 9 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 10, ..., 19 0, 0, 1, 2, 3, 3, 8, 3, 9, 9 20, ..., 29 13, 6, 11, 21, 25, 19, 43, 72, 89, 80 30, ..., 39 65, 60, 69, 81, 35, 83, 85, 77, 48, 79 40, ..., 49 77, 65, 99, 46, 111, 146, 137, 110, 161, 244 50, ..., 59 272, 225, 233, 446, 547, 618, 571, 741, 867, 1051 60, ..., 69 557, 1263, 1418, 1791, 2206, 2008, 3350, 3913, 4164, 2196 70, ..., 79 5132, 5886, 6096, 4499, 6235, 6151, 5852, 5354, 2542, 4779 80, ..., 89 4448, 4510, 3192, 3817, 3453, 2913, 1972, 1875, 2485, 2363 90, ..., 99 2163, 1455, 1812, 1576, 1444, 674, 1074, 941, 870, 733 100, ..., 109 489, 566, 559, 514, 239, 441, 386, 327, 236, 281 110, ..., 119 302, 267, 219, 125, 240, 192, 193, 151, 178, 186 120, ..., 129 178, 124, 132, 177, 173, 168, 97, 150, 151, 153 130, ..., 139 67, 140, 113, 123, 130, 89, 101, 106, 103, 55 140, ..., 149 83, 95, 98, 72, 96, 112, 84, 87, 50, 83 150, ..., 159 62, 75, 59, 95, 72, 91, 65, 61, 97, 87 160, ..., 169 87, 71, 81, 95, 78, 33, 83, 95, 80, 82 170, ..., 179 71, 93, 98, 92, 29, 81, 103, 337, 965, 2489 180, ..., 189 666, 201, 120, 49, 102, 81, 86, 59, 65, 58 190, ..., 199 37, 36, 13, 24, 17, 16, 4, 2, 2, 0 200, ..., 209 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 210, ..., 219 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 220, ..., 229 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 230, ..., 239 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 240, ..., 249 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 250, ..., 255 0, 0, 0, 0, 0, 0

Referring to Table 2, it can be seen that the total number of pixels is 144,054, the average brightness is 83.41825288, and the mode brightness is 74.00.

In the data of Tables 1 and 2, the variation in average brightness and mode brightness is calculated for the object with no gloss on the surface thereof, as shown in Table 3 below.

A matt object A shiny object Change Average brightness 64.728 83.418 18.690 Mode brightness 63.000 74.000 11.000

Referring to Table 3, it can be found that the average brightness has a large change rate depending on whether the background or the surface of the object is glossy, but the mode brightness does not. Therefore, by setting the boundary luminance between the divided regions using the most frequent luminance in each divided region, the object can be identified by maintaining the uniform boundary luminance regardless of whether the background or the surface of the object is glossy. have.

As described above, according to the image processing method according to the present invention, the object can be identified by maintaining a uniform boundary luminance regardless of whether the background or the surface of the object is glossy, so that the accuracy of object identification is stable. do.

The present invention is not limited to the above embodiments, and modifications and improvements are possible at the level of those skilled in the art.

Claims (2)

(a) receiving light reflected from an image area including an object; (b) processing the received light and converting the received light into image data representing luminance of each point of the image area; (c) calculating the mode luminance in the bright region that is greater than or equal to the predetermined boundary luminance and the mode luminance in the dark region that is less than the boundary luminance; (d) determining new boundary luminances by reflecting the most frequent luminance in the bright region and the dark region in a predetermined division ratio; And and (e) converting the image data so that the object is identified from the image area by applying the new boundary luminance. The method of claim 1, wherein step (d) thn = r ⋅ fd + (1-r) ⋅ Image processing method characterized in that performed according to the formula of fb. Where thn is the new boundary luminance, r is the division ratio, fd is the mode luminance in the dark region, and fb is the mode luminance in the bright region.

Images