JP3906221B2 - Image processing method and image processing apparatus - Google Patents

Description

  The present invention relates to an image processing method and an image processing method for performing binarization processing for changing image data having multiple gradations obtained by photographing with a normal digital camera into binary image data. It is about measures.

  For example, as a pre-process of OCR (Optical Character Reader), a character written on paper is read as image data having multiple gradations of 3 gradations or more by a camera or a scanner, that is, a grayscale image, It is necessary to perform binarization processing that separates the background area. In such a case, with reference to a certain threshold value, a gradation value larger than the threshold value, that is, a portion having a bright value is converted to a value of 1 as a background paper region, and a gradation value equal to or lower than that value That is, there is a method of converting a portion having a dark value into a value of 0 as a character area. However, this method has a problem that, due to uneven illumination, the background portion is partially recognized as a character area and cannot be read accurately.

  In addition, the method of separating the target image from the background region by binarizing the multi-gradation image in this manner separates the shooting target such as bacteria from the microscope image, and the position and number of the shooting target. It is also used to find FIG. 2 shows an original image 201 having multiple gradations obtained by photographing a microscope image with a digital camera, a target binary image 205 to be obtained by binarizing the original image 201, and the original image FIG. 6 is a diagram showing a binarized image 206 obtained as a result of binarizing 201 by a conventional method using a fixed threshold.

  The original image 201 is a multi-value gradation image having 256 gradations per pixel. In this image, the right side of the paper is dark due to uneven illumination. The original image 201 includes bacteria 202 to 204, and an image from which these positions and numbers can be obtained, that is, a binary image in which bacteria and a background are separated as shown in a binary image 205, The purpose is to obtain. However, when the original image 201 is binarized using a fixed threshold value, for example, using “128” as the threshold value, among the pixels of the original image 201, pixels larger than “128” are white, “ When the image of 128 ″ or less is replaced with black, the obtained image is as shown in the binarized image 206.

  In the binarized image 206, the bacteria 203 have a threshold value of 128 or less, and the periphery thereof is larger than the threshold value 128, so that they can be detected normally. However, since the right part of the image on the right side of the image is dark, on the right side of the image, the background becomes a value equal to or less than the threshold value, the background becomes black together with the bacteria, and the bacteria 204 are buried in the background. On the contrary, since the bacteria 202 are strongly illuminated, the value becomes larger than the threshold value and becomes white. As a result, it is difficult to accurately determine the positions and numbers of the imaging objects such as bacteria from the background.

As a method for solving such a problem, a method has been proposed in which an image is divided into local regions, and a threshold value for each local region is individually determined based on an average value of the local regions (for example, Patent Documents). 1).
JP-A-5-225392

  The conventional image processing method is configured as described above, and a method of individually determining a threshold value for each local region is adopted in order to eliminate a problem caused by using a single threshold value in binarization. However, since the threshold value is determined based on the average luminance value in the local area, for example, even when the same threshold value must be used with the same brightness, the density of the object to be separated from the background is high. Since the average luminance value is small in the portion, the threshold value is set to a low value. For example, an appropriate threshold value cannot be obtained, and the target binarized image may not be obtained. Had.

  The present invention has been made to solve the above-described problems. An appropriate threshold value in a local region of a multi-tone image is obtained, and an object can be accurately separated from the background using this threshold value. An object of the present invention is to provide an image processing method and an image processing apparatus capable of obtaining a binary image.

  An image processing method according to claim 1 of the present invention is an image processing method for generating a binarized image from a grayscale image that is an original image, and generates a gradation histogram in the whole or a partial region of the original image. And a step of obtaining a first reference luminance value and a reference threshold value which is a threshold value for binarization based on the histogram, and storing a ratio between the first reference luminance value and the reference threshold value as a luminance ratio Generating a local histogram in a local region of the original image; obtaining a second reference luminance value based on the local histogram; and calculating a threshold value for binarization by calculating the second reference luminance value and the luminance ratio And binarizing an image of an area in the vicinity of the local area of the original image based on the threshold value.

  Further, in the image processing method according to claim 2 of the present invention, in the image processing method, the second reference luminance value obtained based on the local histogram is obtained by the same processing as the first reference luminance value. It is what I did.

  In the image processing method according to claim 3 of the present invention, in the image processing method, the first reference luminance value and the second reference luminance value are luminance values at a point indicating the maximum value of each histogram. It is what I did.

  In the image processing method according to claim 4 of the present invention, in the image processing method, the first reference luminance value and the second reference luminance value are maximum values obtained after applying a low pass filter to each histogram. The luminance value at a point indicating the is indicated.

  The image processing method according to claim 5 of the present invention is the image processing method, wherein the reference threshold value is a luminance value in a concave portion adjacent to a maximum value of a histogram in the whole or a partial area of the original image. It is what you do.

  In the image processing method according to claim 6 of the present invention, in the image processing method, when there is no concave portion adjacent to the maximum value of the histogram, the luminance value at the inflection point of the histogram is set as the reference threshold value. It is made to do.

  Further, in the image processing method according to claim 7 of the present invention, in the image processing method, as the reference threshold value, a luminance value in a local convex portion adjacent to a maximum value of a histogram in the whole or a partial region of the original image. And the luminance value at the point where the histogram shows the maximum value, and the value included between the two luminance values is used.

  The image processing method according to claim 8 of the present invention is the image processing method according to claim 8, wherein the reference threshold is set to the maximum value in the vicinity of the maximum value of the histogram in the whole or part of the original image. Thus, the luminance value at a position having a predetermined ratio value is used.

  The image processing method according to claim 9 of the present invention is the image processing method according to claim 9, wherein the reference threshold is set to the maximum value in the vicinity of the maximum value of the histogram in the whole or a partial area of the original image. Thus, the luminance value at the position where the ratio value is determined in advance is obtained, and the luminance value at the position symmetrical to the obtained luminance value is centered on the luminance value at the maximum value.

  The image processing method according to claim 10 of the present invention is the image processing method, wherein the entire original image is divided into a plurality of local regions, and a threshold value is individually provided for each local region to binarize. It is what I did.

  An image processing method according to an eleventh aspect of the present invention is the image processing method according to the image processing method, wherein the entire image is divided into a plurality of local regions, a threshold value is individually provided for each local region, and the threshold value is close to the local region. If there is a local region that is significantly different from a predetermined value compared to the threshold value of the region, the threshold value of this local region is reset by calculation using the threshold values of other local regions that are close to each other. It is a thing.

  According to a twelfth aspect of the present invention, in the image processing method, when the luminance ratio is obtained, unnecessary values based on background characteristics are subtracted from the first reference luminance value and the reference threshold value, respectively. It is what you do.

  An image processing apparatus according to a thirteenth aspect of the present invention is an image processing apparatus that generates a binarized image from a grayscale image that is an original image, and stores multi-tone image data of the grayscale image, Storage means for outputting the entire or specific area of the stored image data as target image data, histogram generation means for generating a histogram of gradations in the target image data, and a first reference luminance based on the histogram A reference luminance ratio generating means for obtaining a value and a reference threshold value that is a threshold value in binarization, and outputting a ratio of the reference threshold value, the first reference luminance value, and the reference threshold value as a reference luminance ratio; The target image data is divided into a plurality of local regions for output, and image dividing means for outputting division information for the target image data, and a local histogram is generated in each local region Local histogram generation means, local area reference luminance value generation means for obtaining a second reference luminance value based on the local histogram, and a binarization threshold value by calculation of the second reference luminance value and the reference luminance ratio And binarization for generating a binarized image of the target image data by binarizing the target image data for each of the local regions based on the division information by a threshold generation unit for calculating Means.

  An image processing apparatus according to a fourteenth aspect of the present invention is the image processing apparatus according to the thirteenth aspect, wherein the histogram generating means and the local histogram generating means are configured to include the first reference luminance value and the second reference luminance value. The reference luminance value is obtained by the same process.

  An image processing apparatus according to claim 15 of the present invention is the image processing apparatus according to claim 14, wherein the first reference luminance value and the second reference luminance value indicate a maximum value of each histogram. It is a luminance value at a point.

  The image processing apparatus according to claim 16 of the present invention is the image processing apparatus according to claim 14, wherein the first reference luminance value and the second reference luminance value are subjected to a low-pass filter on each histogram. It is a luminance value at a point indicating the maximum value obtained after the operation.

  The image processing device according to claim 17 of the present invention is the image processing device according to claim 13, wherein the reference luminance ratio generation means is a luminance value in a concave portion adjacent to a maximum value of a histogram in the target image data. Is the reference threshold value.

  An image processing apparatus according to claim 18 of the present invention is the image processing apparatus according to claim 17, wherein when there is no concave portion in the maximum value of the histogram, the luminance value at the inflection point of the histogram is set. The reference threshold value is used.

  The image processing device according to claim 19 of the present invention is the image processing device according to claim 13, wherein the reference luminance ratio generation means is a luminance in a local convex portion adjacent to a maximum value of a histogram in the target image data. A value and a luminance value at a point where the histogram shows the maximum value are obtained, and a value included between the two luminance values is used as the reference threshold value.

  An image processing apparatus according to claim 20 of the present invention is the image processing apparatus according to claim 13, wherein the reference luminance ratio generation means sets the maximum value in the vicinity of the maximum value of the histogram in the target image data. On the other hand, a luminance value at a position having a predetermined ratio value is used as the reference threshold value.

  The image processing apparatus according to claim 21 of the present invention is the image processing apparatus according to claim 13, wherein the reference luminance ratio generation means sets the maximum value in the vicinity of the maximum value of the histogram in the target image data. On the other hand, the luminance value at a position where the ratio is a predetermined ratio is set as a reference luminance value, and the luminance value at a position symmetrical to the obtained reference luminance value with the luminance value at the maximum value as the center is set as the reference threshold value. It is characterized by that.

  The image processing apparatus according to claim 22 of the present invention is the image processing apparatus according to claim 13, wherein the threshold value generation means stores a threshold value memory for storing a threshold value of each local region, and a specification stored in the threshold value memory. A threshold value comparison unit that compares a threshold value of the local region with a threshold value of the surrounding local region, and a threshold value of the specific local region compared in the threshold value comparison unit is determined in advance compared to a threshold value of the surrounding local region When there is a difference greater than or equal to a value, a threshold value determination changing unit is provided that replaces the threshold value of the specific local region with an average value of the threshold values of the surrounding local regions.

  An image processing apparatus according to claim 23 of the present invention is the image processing apparatus according to claim 13, wherein the image processing apparatus is used for a business card reading device of a camera built-in mobile phone. .

  According to the image processing method of the first aspect of the present invention, there is provided an image processing method for generating a binarized image from a grayscale image that is an original image, and a histogram of gradations in the whole or a partial region of the original image And a reference threshold value corresponding to a binarization threshold value and a first reference luminance value are obtained based on the histogram, and a ratio between the first reference luminance value and the reference threshold value is stored as a luminance ratio. A step of generating a local histogram in a local region of the original image, obtaining a second reference luminance value based on the local histogram, and performing binarization by calculating the second reference luminance value and the luminance ratio And a step of binarizing the image of the area near the local area of the original image based on the threshold value. Therefore, a histogram is created for the entire image, and the reference luminance value and reference at the maximum frequency are generated. After determining the value ratio, divide the image into local regions, create a histogram for each local region, determine the luminance value at the maximum frequency, and multiply this value by the ratio between the reference luminance value and the reference threshold value. The threshold value in the local area is obtained by using this threshold value, and the binarization in the local area is performed using this threshold value, thereby obtaining an appropriate threshold value in the local area of the multi-tone image, and using this threshold value, the object is accurately detected from the background. There is an effect that a well-separable binary image can be obtained.

  According to an image processing method of claim 2 of the present invention, in the image processing method of claim 1, the second reference luminance value obtained based on the local histogram is the first reference luminance. Since it is obtained by the same processing as the value, it is possible to obtain an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background.

  According to an image processing method according to claim 3 of the present invention, in the image processing method according to claim 2, the first reference luminance value and the second reference luminance value are maximum values of the respective histograms. Since the luminance value at the indicated point is set, an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background can be obtained.

  According to an image processing method according to claim 4 of the present invention, in the image processing method according to claim 2, the first reference luminance value and the second reference luminance value are obtained by applying a low pass filter to each histogram. Since the luminance value at the point indicating the maximum value obtained after being applied is set to an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background Can be requested.

  According to an image processing method of claim 5 of the present invention, in the image processing method according to claim 1, the reference threshold value is adjacent to a maximum value of a histogram in the whole or a partial area of the original image. Since the luminance value in the concave portion is set, an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background can be obtained.

  According to the image processing method of claim 6 of the present invention, in the image processing method of claim 5, when there is no concave portion adjacent to the maximum value of the histogram, the inflection point of the histogram Since the luminance value is set as the reference threshold value, an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background can be obtained.

  According to an image processing method of claim 7 of the present invention, in the image processing method of claim 1, the reference threshold value is adjacent to a maximum value of a histogram in the whole or a partial area of the original image. Since the luminance value at the local convex portion and the luminance value at the point where this histogram shows the maximum value are obtained and the value included between the two luminance values is used, the object can be accurately separated from the background 2 An appropriate threshold value in a local region of a multi-tone image for obtaining a valued image can be obtained.

  According to an image processing method according to claim 8 of the present invention, in the image processing method according to claim 1, the reference threshold is set in the vicinity of the maximum value of the histogram in the whole or part of the original image. Since the luminance value is set at a position having a predetermined ratio with respect to the maximum value, a multi-tone image for obtaining a binarized image capable of accurately separating the object from the background is provided. An appropriate threshold value in the local region can be obtained.

  Further, according to an image processing method according to claim 9 of the present invention, in the image processing method according to claim 1, the reference threshold is set in the vicinity of the maximum value of the histogram in the whole or a partial region of the original image. The luminance value at a position where a predetermined ratio is obtained with respect to the maximum value is obtained, and the luminance value at a position symmetrical to the obtained luminance value is centered on the luminance value at the maximum value. Therefore, it is possible to obtain an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background.

  According to an image processing method according to claim 10 of the present invention, in the image processing method according to claim 1, the entire original image is divided into a plurality of local regions, and a threshold value is individually set for each local region. Since it is provided and binarized, it is possible to obtain an appropriate threshold value in a local region of a multi-tone image for obtaining a binarized image capable of accurately separating an object from the background.

  The image processing method according to claim 11 of the present invention is the image processing method according to claim 10, wherein the entire image is divided into a plurality of local regions, and a threshold is individually set for each local region. If there is a local region whose threshold value is significantly different from a predetermined value compared to the threshold value of the adjacent local region, the threshold value of this local region is calculated by using the threshold value of another local region. Since resetting is performed, it is possible to obtain an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background.

  According to an image processing method of claim 12 of the present invention, in the image processing method of claim 1, when obtaining the luminance ratio, the background characteristics are respectively obtained from the first reference luminance value and the reference threshold value. Since an unnecessary value based on the subtraction is subtracted, an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background can be obtained.

  According to the image processing apparatus of the thirteenth aspect of the present invention, there is provided an image processing apparatus that generates a binarized image from a grayscale image that is an original image, and stores multi-tone image data of the grayscale image. Storage means for outputting the entire stored image data or a specific region as target image data, histogram generation means for generating a histogram of gradations in the target image data, and a reference threshold based on the histogram A reference luminance ratio generating unit that obtains a first reference luminance value and outputs a ratio between the reference threshold value and the first reference luminance value as a reference luminance ratio; and divides the target image data into a plurality of local regions. Image division means for outputting division information for the target image data, local histogram generation means for generating a local histogram in each local region, Local area reference luminance value generation means for obtaining a second reference luminance value based on the local histogram; threshold generation means for obtaining a binarization threshold value by calculating the reference luminance ratio and the second reference luminance value; Binarizing means for binarizing the target image data for each of the local regions based on the division information by using the binarization threshold, and generating a binarized image of the target image data. Thus, there is an effect that a proper threshold value in a local region of a multi-tone image is obtained, and a binary image capable of accurately separating an object from the background can be obtained using this threshold value.

  According to an image processing apparatus of claim 14 of the present invention, in the image processing apparatus according to claim 13, the histogram generation means and the local histogram generation means include the first reference luminance value, and the Since the second reference luminance value is obtained by the same process, an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background is obtained. be able to.

  According to an image processing device of claim 15 of the present invention, in the image processing device according to claim 14, the first reference luminance value and the second reference luminance value are maximum values of respective histograms. Therefore, it is possible to obtain an appropriate threshold value in a local region of a multi-tone image for obtaining a binarized image capable of accurately separating an object from the background.

  According to an image processing device of the sixteenth aspect of the present invention, in the image processing device according to the fourteenth aspect, the first reference luminance value and the second reference luminance value are low-pass filters in each histogram. Since the luminance value at the point indicating the maximum value obtained after the image processing is performed, an appropriate value in the local region of the multi-tone image is obtained in order to obtain a binarized image capable of accurately separating the object from the background. A threshold can be determined.

  According to an image processing device of claim 17 of the present invention, in the image processing device according to claim 13, the reference luminance ratio generation means is provided for a concave portion adjacent to a maximum value of a histogram in the target image data. Since the luminance value is set as the reference threshold value, an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background can be obtained.

  According to the image processing device of claim 18 of the present invention, in the image processing device according to claim 17, when there is no concave portion in the maximum value of the histogram, the luminance at the inflection point of the histogram. Since the value is set as the reference threshold value, an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background can be obtained.

  According to an image processing device of claim 19 of the present invention, in the image processing device according to claim 13, the reference luminance ratio generation means is a local convex portion adjacent to a maximum value of a histogram in the target image data. Since the luminance value at the point where the histogram shows the maximum value is obtained and the value included between the two luminance values is used as the reference threshold, the object can be accurately separated from the background. An appropriate threshold value in a local region of a multi-tone image for obtaining a binarized image can be obtained.

  Further, according to an image processing device of claim 20 of the present invention, in the image processing device according to claim 13, the reference luminance ratio generation means is configured to detect the maximum in the vicinity of a maximum value of a histogram in the target image data. Since the luminance value at a position where the ratio is a predetermined ratio with respect to the value is used as the reference threshold value, a multi-tone image for obtaining a binarized image capable of accurately separating the object from the background An appropriate threshold value in the local region can be obtained.

  According to an image processing apparatus of claim 21 of the present invention, in the image processing apparatus according to claim 13, the reference luminance ratio generation means is configured to increase the maximum in the vicinity of the maximum value of the histogram in the target image data. The luminance value at a position where the ratio is a predetermined ratio to the value is set as a reference luminance value, and the luminance value at a position symmetrical to the reference luminance value with the luminance value at the maximum value as the center is set as the reference threshold value. Therefore, it is possible to obtain an appropriate threshold value in the local region of the multi-tone image for obtaining a binarized image capable of accurately separating the object from the background.

  The image processing apparatus according to claim 22 of the present invention is the image processing apparatus according to claim 13, wherein the threshold value generation means stores a threshold value memory for storing a threshold value of each local region, and a specification stored in the threshold value memory. A threshold value comparison unit that compares a threshold value of the local region with a threshold value of the surrounding local region, and a threshold value of the specific local region compared in the threshold value comparison unit is determined in advance compared to a threshold value of the surrounding local region When there is a difference greater than or equal to the value, the threshold judgment changing unit that replaces the threshold value of the specific local area with the average value of the threshold values in the surrounding local area is provided, so that the object is accurately separated from the background. An appropriate threshold value in a local region of the multi-tone image for obtaining a possible binarized image can be obtained.

  The image processing apparatus according to claim 23 of the present invention is the image processing apparatus according to claim 13, wherein the image processing apparatus is used for a business card reading device of a camera built-in mobile phone. It is possible to achieve the function of taking a picture, reading the characters of the business card, and making it into a database.

(Embodiment 1)
FIG. 3 is a diagram for explaining the principle of the image processing method according to the first embodiment of the present invention, and is a multi-gradation original image 300 having three or more gradations, that is, an original image having light and shade and its respective regions. Is a histogram. As the original image, an image of bacteria is used. A histogram 303 is a histogram of luminance values in the local area 301 of the original image 300.

  This histogram 303 represents the luminance values of all the pixels included in the local region 301 by the frequency, and takes a dark pixel on the left side of the paper, that is, a luminance value of 0, and a bright pixel on the right side of the paper, that is, a luminance value of 255. The number of pixels indicating each luminance value is plotted on the vertical axis. In the local region 301, the background portion has the largest area, and therefore, in the histogram 303, the number of pixels near the background luminance A is the largest. In addition, the bacteria portion having the brightness B appears as a small peak. In the local region 301, it is possible to separate the bacteria and the background by using the luminance C as a threshold value.

  Similarly, the histogram 304 is a histogram of luminance values in the local region 302 of the original image 300. Since the entire local area 302 is darker than the local area 301, the histogram 304 has a shape that is jammed to the left of the page as compared with the histogram 303. Also in the local region 302, since the background portion has the largest area, the number of pixels in the vicinity of the luminance D of the background is the largest in the histogram 304. In addition, the portion of bacteria having brightness E appears as a small peak. In the local region 302, it is possible to separate the bacteria and the background by using the brightness F as a threshold value.

  A histogram 305 is a histogram of luminance values of the entire original image 300. Since this histogram 305 is a sum of histograms of all local regions including the local regions 301 and 302, each peak is gentle, but the luminance G portion with the largest number of pixels in the background portion, A peak appears in each of the luminance H portions having the largest number of pixels in the bacterial portion. The luminance I is a luminance that separates the background and the bacterial part. However, when this value is binarized as the threshold value of the entire image, the value I is equal to or lower than the luminance B shown in the histogram 303. Cannot be processed normally because it determines the bacterial part as the background.

  For this reason, binarization is performed using threshold values suitable for each local region, thereby obtaining a binarized image capable of accurately separating an image of an object, here bacteria, from a background image. In determining the threshold, the physical properties of each local region are analyzed. For example, it is assumed that the background portion has a reflectance of 50% and the bacterial portion has a reflectance of 10% with respect to the applied light. Then, assuming that the intensity of light given to the local region 301 is X, the luminance A has a relationship of A = 0.5X and the luminance B has a relationship of B = 0.1X. When the threshold value is the median value C of the luminance A and the luminance B, the luminance C has a value of C = (0.5X + 0.1X) /2=0.3X. If X is eliminated from the relationship between A and C, the relationship C = 0.6A is established. Similarly, the relationship between the luminance D and the threshold value F in the local region 302 is F = 0.6D, and the threshold values in all the local regions are given as the luminance of the background portion multiplied by the same coefficient. Note that the threshold value in the local area can be obtained by the same method by multiplying the brightness of the bacterial part by the same coefficient, but if the background part area is large as in this image, the background part Therefore, the threshold value is obtained by multiplying the luminance of the background portion by a coefficient.

  In addition, the coefficient to be multiplied can be stably obtained by taking the ratio of the peak portion G serving as the background and the threshold value I in the histogram 305 of the entire original image 300 or a histogram of a large area equivalent thereto.

  Based on the above, the procedure for binarizing the original image by the image processing method according to the first embodiment will be described below with reference to FIG. FIG. 1 is a processing flowchart of the image processing method according to the first embodiment. Here, in order to obtain a binary image 205 by binarizing a microscopic image (original image) 201 (see FIG. 2) of bacteria having a size of 1000 × 1000 pixels and each pixel having 256 gray scales. Each step when performing the above process will be described.

  First, in step S101, a luminance histogram for the entire image is created. Here, the entire image is shown, but when the peripheral portion of the image is dark, a luminance histogram in an image obtained by cutting out only the central portion may be used. Since these histograms vary in distribution due to the influence of noise or the like, a low pass filter is applied to the histograms in the process of step S102. In the actual processing, it is realized by taking the average value of the frequencies for the front and rear luminances.

  In step S103, a luminance value indicating the maximum frequency is obtained in the histogram subjected to the low-pass filter. This is hereinafter referred to as a reference luminance value. This value means the luminance value G of the background portion in FIG. In step S104, a search is made for a luminance indicating a minimum value in the direction of decreasing luminance from the reference luminance value obtained in step S103. This portion corresponds to the luminance I in FIG. 3 and means a threshold value for binarization. Hereinafter, this is referred to as a reference threshold.

  In step S105, the ratio I / G between the reference threshold value and the reference luminance value is obtained and stored as J. In step S106, the original image 201 is divided into local areas of 100 × 100 pixels. There are 10 × 10 local areas in the vertical and horizontal directions. In step S107, a luminance histogram is created in one local area. Here, for example, if the local region of interest is considered to be the local region 301 in FIG. 3, a histogram 303 is generated. When the local region is divided as a very small region, a histogram may be created in a wider range centering on the local region in order to improve accuracy against noise.

  In step S108, a low pass filter is applied to the histogram of each local region. In step S109, the luminance value indicating the maximum frequency is obtained from the histogram subjected to the low-pass filter. This value corresponds to the luminance value A of the background portion in FIG. In step S110, the threshold value in the local region is obtained by multiplying the luminance value indicating the maximum frequency of the local region by J described above. For example, considering the local region 301, the threshold value in the local region 301 can be obtained by multiplying the luminance value A by J. In step S111, the processing from step S107 to step S110 is repeated for all local regions. In step S112, binarization is performed on all the pixels of the original image 300 using the threshold values obtained for each local region for each local region, and a final binary image 205 is obtained.

  In step S104, the luminance indicating the minimum value is searched from the reference luminance value obtained in step S103 in the direction of decreasing luminance, and the reference threshold value is obtained from the minimum value. When the value is searched for, the obtained maximum value indicates the brightness of the bacterium, so that the median value of this brightness value and the background brightness value obtained in step S103 can be used as the reference threshold value.

  In the above method, in the processing step 104, the reference threshold value is obtained using the local minimum value or the local maximum value. However, depending on the image, the local maximum value or the local minimum value may not exist. In addition, in the threshold value calculation method in the processing step S110 described above, an accurate threshold value may not be calculated due to the uneven distribution of bacteria or the like. Hereinafter, a modified example of the reference threshold value and the threshold value calculation method in such a case will be described.

First, a modification of the reference threshold calculation process will be described.
FIG. 4 is a diagram for explaining a modification of the process for obtaining the reference threshold in the image processing method according to the first embodiment, and shows a histogram 400 of an image. In this example, since the illuminance greatly changes in the original image, the luminance distribution of the background and the luminance distribution of the bacteria overlap, and there is no maximum value or minimum value between them. Therefore, in the processing step S104, after searching for the minimum value from the reference luminance value 402 indicating the maximum frequency in the direction of decreasing luminance, if the minimum value does not exist within the search range, refer to the luminance of the inflection point. The threshold value 403 is used.

  Further, since the inflection point may not be obtained stably, another method for obtaining the reference threshold in such a case will be described with reference to FIG. In FIG. 5, a luminance value indicating the maximum frequency 501 (peak value on the histogram) of the histogram 500 is set as a reference luminance value 504. A brightness having a frequency 502 equal to or less than a ratio of a value given in advance to the frequency 501 at this time, for example, 40%, is searched for in the direction of decreasing brightness. The luminance obtained as a result is set as a reference threshold value 503. In this method, when the shape of the histogram is known to some extent, the reference threshold value can be obtained reliably without being affected by noise.

  In the above-described method, the luminance with the frequency 502 having a value equal to or less than 40% of the maximum frequency 501 is obtained on the low luminance side, but this point is affected by the number of bacteria, so the high luminance side is used. Thus, the influence can be eliminated by obtaining the reference threshold. For example, in FIG. 6, a luminance value indicating the maximum frequency 604 (peak value on the histogram) of the histogram 600 is set as the reference luminance value 603. A brightness having a frequency 605 equal to or less than a ratio of a value given in advance to the frequency 604 at this time, for example, 40%, is searched for in a higher brightness direction. With respect to the luminance 601 obtained as a result, the luminance at a position symmetric with respect to the reference luminance 603 is set as a reference threshold value 602. For example, when the luminance 603 has a value of 150 and the luminance value 601 has a value of 180, the reference threshold 602 is 120.

  Next, a modified example of the threshold value calculation method in the local region will be described. In processing step S110, since the threshold value of the local region is obtained using only the maximum value of the histogram, the threshold value can be obtained accurately even when bacteria are not included in the local region. However, when the bacteria are concentrated in a specific local area and the maximum value of the histogram indicates the area of bacteria, a normal threshold value cannot be obtained. In order to avoid this, after the processing step S111 shown in FIG. 1, the threshold value for each local region is compared with the threshold value in the surrounding local region, and in the local region having a difference greater than a predetermined value, An appropriate threshold value can be obtained by rejecting the threshold value and interpolating surrounding threshold values to obtain a new threshold value.

  Further, in the processing step S110, the threshold value in the local region is obtained by multiplying the luminance value A of the background portion by J which is the ratio I / G of the reference threshold value and the reference luminance value calculated in the processing step S105 of FIG. However, when the original image is taken with a device such as a CCD, the dark current value based on the characteristics of the CCD is superimposed on the brightness value. Need to be removed. In such a case, if the luminance component superimposed by the dark current is Z, the ratio between the reference threshold value and the reference luminance value can be obtained by J = (I−Z) / (G−Z). By calculating by J × (A−Z) + Z, a more accurate threshold can be obtained.

  As described above, according to the image processing method according to the first embodiment, a gradation histogram is generated in the whole or a partial area of the original image, and the reference corresponding to the binarization threshold is based on the histogram. A threshold value and a first reference luminance value are obtained, a ratio between the two is stored as a luminance ratio, a local histogram is generated in a local region of the original image, and a second reference luminance value is obtained based on the local histogram. And a threshold value for binarization by calculating the luminance ratio, and binarizing an image in a region near the local region of the original image using the threshold value. In other words, a histogram is created for the entire image, the ratio between the reference luminance value and the reference threshold value at the maximum frequency is obtained, the image is divided into local regions, a histogram is created for each local region, and the luminance at the maximum frequency is obtained. A threshold value in the local region is obtained by determining the value and multiplying this value by the ratio between the reference luminance value and the reference threshold value, and binarization in the local region is performed using this threshold value. As a result, an appropriate threshold value in the local region of the multi-tone image can be obtained, and a binary image capable of accurately separating the object from the background can be obtained using this threshold value.

(Embodiment 2)
FIG. 7 is a block diagram of an image processing apparatus according to the second embodiment for realizing the image processing method according to the first embodiment of the present invention.

  In FIG. 7, an image taken by the camera 700 is stored in the frame memory 701. Here, it is assumed that the photographed image is 1000 × 1000 pixels and each pixel is a black and white image having 256 gradations. The frame memory 701 outputs the entire area of the stored captured image or a part of the captured image as image data. Here, it is assumed that the entire captured image area of the frame memory 701 is output.

  Next, the overall histogram creation unit 702 creates a brightness histogram of the entire image data output from the frame memory 701, as shown in steps S101 to S102 shown in the first embodiment.

  The reference luminance ratio creating unit 703 obtains a luminance value indicating the maximum frequency of the luminance histogram generated by the overall histogram creating unit 702 as shown in steps S103 to S105 of the first embodiment, and uses this as the reference luminance value G. At the same time, a luminance indicating a minimum value from the reference luminance value G toward a lower luminance direction is searched and used as a reference threshold value I, and a value J obtained by dividing the reference threshold value I by the reference luminance value G is obtained as a reference luminance ratio. Output.

  On the other hand, the image dividing unit 704 divides the image data output from the frame memory 701 into, for example, 10 local areas and 10 local areas as shown in step S106 of the first embodiment. At this time, one small region has a size of 100 × 100 pixels. Then, division information indicating the size of the image data output from the frame memory 701 is output to the binarization unit 708.

  The local histogram creation unit 705 creates a local histogram for each local region of the divided image data output from the image division unit 704, as shown in steps S107 to S108 of the first embodiment.

  As shown in step S109 of the first embodiment, the peak luminance detection unit 706 obtains a luminance value indicating the maximum frequency of the local histogram generated by the local histogram creation unit 705, and uses this as the background luminance value A of the local region. Output.

  As shown in step S110 of the first embodiment, the threshold calculation unit 707 multiplies the reference luminance ratio J output from the reference luminance ratio generation unit 703 by the background luminance value A output from the peak luminance detection unit 706, A binarization threshold value in the local region is obtained.

  As shown in step S112, the binarizing unit 708 performs threshold processing on the image data output from the frame memory 701 for each local region based on the division information from the image dividing unit 704. The binarization process using is repeatedly performed, and finally the entire image data output from the frame memory 701 is binarized to output a binarized image of the image data.

  The threshold value calculation unit 707 compares the threshold value memory for storing the threshold value of each local region, the threshold value comparison unit for comparing the threshold value of a certain local region stored in the threshold value memory with the threshold value in the surrounding local region, and the threshold value comparison unit. In the case where the threshold value has a difference greater than or equal to a predetermined value, a configuration may be provided that includes a threshold value determination changing unit that replaces the threshold value of the local region with the average value of the surrounding threshold values.

  Hereinafter, the threshold judgment changing unit having the above-described configuration will be described with reference to FIG. FIG. 9 shows threshold values for each region stored in the threshold memory. Reference numerals 901 to 935 denote areas in the threshold memory, and the values of the respective thresholds are indicated by bold characters. It is assumed that a threshold value “21”, which is different from the original value, is stored in the shaded area 913 because dust has entered. In the threshold judgment changing unit, for all the areas, the threshold value of the attention area and the threshold value of the surrounding area are compared, and when these are greatly different, for example, when they are different by “20” or more, The threshold value of the attention area is reset by interpolation of the threshold values of the surrounding area. For example, when the region 902 is set as the attention region, the threshold value of this region is 65. Further, regions 901, 911, 912, 913, and 903 that are in contact with the region 902 have threshold values of 61, 64, 67, 21, and 73, respectively. Therefore, 65 is compared with 61, 64, 76, 21, 73. The differences in this case are 4, 1, 2, 44 and 8, respectively, and only one exceeds “20”. In this case, since only one of the five values compared is less than “20”, the threshold value 65 of the area 902 is used as it is.

  On the other hand, when the region 913 is the region of interest, this threshold is 21. In addition, the areas in contact with the area 913 are 902, 903, 904, 914, 924, 923, 922, and 912, and their threshold values are 65, 73, 82, 82, 85, 79, 75, and 67, respectively. . Therefore, “21” is compared with these values. The differences in this case are 44, 52, 61, 61, 64, 58, 54, and 46, respectively, and there are eight that exceed “20”. In this case, eight of the eight compared values, which are half or more, exceed “20”, so the threshold “21” of the region 913 is reset by interpolation. Interpolation is performed by obtaining an average value of surrounding threshold values. In this case, (65 + 73 + 82 + 82 + 85 + 79 + 75 + 67) / 8 = 76, and the threshold value of the area 913 is reset to 76.

  With such a configuration, an appropriate threshold value can be obtained even when the threshold value of a certain local region becomes an abnormal value due to deformation of the local histogram due to contamination of dust or the like.

  Next, an example in which the image processing apparatus of the present invention is applied to a business card reading function of a camera built-in mobile phone will be described. FIG. 8 is a configuration diagram of a mobile phone with a business card reading function, which incorporates an image processing apparatus according to Embodiment 2 of the present invention. This mobile phone has a function of taking a business card with a built-in camera, reading the characters of the business card, and making it into a database.

  In FIG. 8, reference numeral 801 denotes a mobile phone, which takes a business card 802, recognizes characters, and stores it in a database. A business card 802 is taken by a camera 803 built in the mobile phone 801 and stored as image data. This image is binarized by the image processing apparatus 804, the background area is represented by “1”, and the character area is represented by “0”. The image binarized by the image processing device 804 is subjected to pattern matching of the character area in the character recognition unit 805, and character text is extracted. The extracted character text is stored in the business card database 806.

  Here, the image captured by the camera 803 often has uneven brightness of the background area and the character area, such as part of the background is dark due to uneven illuminance at the time of shooting. Even in such a case, the image processing apparatus 804 of the present invention can accurately separate the background area and the character area, and the character recognition unit 805 can perform stable character recognition processing.

  Here, the camera built-in mobile phone has been described as an example, but the same effect can be obtained by installing the image processing apparatus of the present invention as long as the device can capture a digital image such as a scanner or a digital still camera. Can be obtained.

  As described above, according to the image processing apparatus according to the second embodiment, an apparatus that performs the image processing method according to the first embodiment can be simply configured.

  INDUSTRIAL APPLICABILITY The image processing apparatus according to the present invention is useful as a binarization processing method performed for identifying a target image from a background image with respect to a multi-tone image, and in particular, a preprocessing for character reading of a printed matter The binarization method can be applied to uses such as a binarization method for preprocessing when automatically counting observation objects in a microscope image.

It is a figure which shows the flowchart of the image processing method in Embodiment 1 of this invention. It is a figure for demonstrating the conventional image processing method. It is a figure which shows the histogram shape in the original image and each part for demonstrating the image processing method in Embodiment 1 of this invention. It is a figure for demonstrating the modification of the image processing method in Embodiment 1 of this invention. It is a figure for demonstrating the modification of the image processing method in Embodiment 1 of this invention. It is a figure for demonstrating the modification of the image processing method in Embodiment 1 of this invention. It is a block diagram of the image processing apparatus in Embodiment 2 of this invention. It is a block diagram which shows the business card reading function of the mobile phone with a built-in camera in Embodiment 2 of this invention. It is a figure for demonstrating operation | movement of the threshold value judgment change part in Embodiment 2 of this invention.

Explanation of symbols

201 Original image 202 to 204 Bacteria part in original image 205 Target binary image 206 Binarized image with fixed threshold 300 Original image 301 to 302 Local region in original image 303 Histogram in local region 301 304 In local region 302 Histogram 305 Histogram of the entire original image 400 400 Histogram of the original image 402 Reference luminance 403 Reference threshold value 500 Histogram of the original image 501 Peak value on the histogram 502 40% level of the peak value 503 Reference threshold value 504 Reference luminance value 600 Histogram of the original image 601 Luminance value at 40% level of peak value 602 Reference threshold 603 Luminance value at peak value on histogram 604 Peak value on histogram 605 40% level of peak value 700 Camera 701 Frame Mori 702 Overall histogram creation unit 703 Reference brightness ratio creation unit 704 Image segmentation unit 705 Local histogram creation unit 706 Peak brightness detection unit 707 Threshold calculation unit 708 Binarization unit 801 Mobile phone with camera 802 Business card 803 Camera 804 Image processing device 805 Character Recognition unit 806 Business card database

Claims (23)

An image processing method for generating a binarized image from a grayscale image as an original image,
Generating a histogram of tones within the entire or partial area of the original image;
Obtaining a first reference luminance value based on the histogram and a reference threshold value that is a threshold value for binarization, and storing a ratio between the first reference luminance value and the reference threshold as a luminance ratio;
Generating a local histogram in a local region of the original image;
A second reference luminance value is obtained based on the local histogram, a threshold value for binarization is obtained by calculating the second reference luminance value and the luminance ratio, and the threshold value of the region near the local region of the original image is obtained by the threshold value. And an image processing method comprising: binarizing the image. The image processing method according to claim 1,
The image processing method according to claim 1, wherein the second reference luminance value obtained based on the local histogram is obtained by the same process as the first reference luminance value. The image processing method according to claim 2.
The image processing method according to claim 1, wherein the first reference luminance value and the second reference luminance value are luminance values at a point indicating the maximum value of each histogram. The image processing method according to claim 2.
The image processing method according to claim 1, wherein the first reference luminance value and the second reference luminance value are luminance values at points indicating maximum values obtained after applying a low-pass filter to each histogram. The image processing method according to claim 1,
An image processing method, wherein the reference threshold value is a luminance value in a concave portion adjacent to a maximum value of a histogram in the whole or a partial area of the original image. The image processing method according to claim 5.
An image processing method, wherein a luminance value at an inflection point of the histogram is used as the reference threshold when there is no concave portion adjacent to the maximum value of the histogram. The image processing method according to claim 1,
As the reference threshold value, a luminance value at a local convex portion adjacent to the maximum value of the histogram in the whole or a partial area of the original image and a luminance value at a point where the histogram shows the maximum value are obtained, and both luminance values are obtained. An image processing method characterized by using a value included in between. The image processing method according to claim 1,
The reference threshold value is a luminance value at a position that is a value of a predetermined ratio with respect to the maximum value in the vicinity of the maximum value of the histogram in the whole or a partial region of the original image. Image processing method. The image processing method according to claim 1,
The reference threshold is obtained as a reference luminance value at a position where the ratio is a predetermined ratio with respect to the maximum value in the vicinity of the maximum value of the histogram in the whole or part of the original image, An image processing method characterized in that a luminance value at a position symmetrical to the obtained reference luminance value is centered on a luminance value at a maximum value. The image processing method according to claim 1,
An image processing method characterized in that the entire original image is divided into a plurality of local areas, and a threshold value is individually provided for each local area to binarize. The image processing method according to claim 10.
The entire image is divided into a plurality of local regions, and a threshold value is individually set for each local region, and there is a local region whose threshold value is significantly different from a predetermined value compared to a threshold value of a local region in the vicinity. In this case, an image processing method is characterized in that the threshold value of the local region is reset by calculation using a threshold value of another local region adjacent thereto. The image processing method according to claim 1,
An image processing method characterized by subtracting unnecessary values based on background characteristics from the first reference luminance value and the reference threshold value when obtaining the luminance ratio. An image processing apparatus that generates a binarized image from a grayscale image that is an original image,
Storage means for storing multi-tone image data of the grayscale image, and outputting the entire stored image data or a specific area as target image data;
A histogram generating means for generating a gradation histogram in the target image data;
Based on the histogram, a first reference luminance value and a reference threshold value that is a threshold value for binarization are obtained, and a ratio between the reference threshold value, the first reference luminance value, and the reference threshold value is used as a reference luminance ratio. A reference luminance ratio generating means for outputting;
The target image data is divided into a plurality of local regions for output, and image division means for outputting division information for the target image data;
Local histogram generating means for generating a local histogram in each local region;
Local area reference luminance value generating means for obtaining a second reference luminance value based on the local histogram;
Threshold generation means for obtaining a binarization threshold by calculating the second reference luminance value and the reference luminance ratio;
Binarization means for binarizing the target image data for each of the local regions based on the division information by using the binarization threshold, and generating a binarized image of the target image data;
An image processing apparatus comprising: The image processing apparatus according to claim 13.
The histogram generating means and the local histogram generating means are:
The image processing apparatus, wherein the first reference luminance value and the second reference luminance value are obtained by the same process. The image processing apparatus according to claim 14.
The image processing apparatus according to claim 1, wherein the first reference luminance value and the second reference luminance value are luminance values at a point indicating a maximum value of each histogram. The image processing apparatus according to claim 14.
The image processing apparatus according to claim 1, wherein the first reference luminance value and the second reference luminance value are luminance values at points indicating maximum values obtained after applying a low-pass filter to each histogram. The image processing apparatus according to claim 13.
The image processing apparatus according to claim 1, wherein the reference luminance ratio generation unit uses a luminance value in a concave portion adjacent to a maximum value of a histogram in the target image data as the reference threshold value. The image processing apparatus according to claim 17.
The image processing apparatus according to claim 1, wherein when there is no concave portion in the maximum value of the histogram, a luminance value at an inflection point of the histogram is used as the reference threshold value. The image processing apparatus according to claim 13.
The reference luminance ratio generation unit obtains a luminance value at a local convex portion adjacent to the maximum value of the histogram in the target image data and a luminance value at a point where the histogram shows the maximum value, and is included between the two luminance values. An image processing apparatus characterized in that a value to be used is the reference threshold value. The image processing apparatus according to claim 13.
The reference luminance ratio generation means uses, as the reference threshold value, a luminance value at a position that is a value of a predetermined ratio with respect to the maximum value in the vicinity of the maximum value of the histogram in the target image data. An image processing apparatus. The image processing apparatus according to claim 13.
The reference luminance ratio generation unit uses a luminance value at a position that is a value of a predetermined ratio with respect to the maximum value in the vicinity of the maximum value of the histogram in the target image data as a reference luminance value, and the luminance at the maximum value An image processing apparatus characterized in that a luminance value at a position symmetrical to the obtained reference luminance value with the value as a center is used as the reference threshold value. The image processing apparatus according to claim 13.
The threshold generation means includes
A threshold memory for storing threshold values for each local region;
A threshold comparing unit that compares a threshold of a specific local area stored in the threshold memory with a threshold in a surrounding local area;
When the threshold value of the specific local region compared in the threshold value comparison unit has a difference greater than or equal to a predetermined value compared to the threshold value of the surrounding local region, the threshold value of the specific local region is An image processing apparatus comprising: a threshold value determination changing unit that replaces the average value of the threshold values in the local region. The image processing apparatus according to claim 13.
An image processing device used for a business card reading device of a camera-equipped mobile phone.

Images