JP5047126B2 - Image processing apparatus and background removal method - Google Patents

Description

  The present invention relates to an image processing apparatus, and more particularly, to an image processing apparatus and background removal method for setting a removal threshold value when removing a background of a document from image data obtained by reading a document.

  In a copying process in which an image reading device reads an original and performs printing, various types of paper are used as the original. Some paper used as a manuscript has a relatively high background density, such as newspapers, recycled paper, and colored paper. In the case of reading a document having such a high background density, processing for removing the background has been conventionally performed. Thus, in addition to making it easy to read the characters of the copied document clearly, it is possible to reduce the consumption of ink and toner during printing.

When performing background removal, image processing is performed to estimate the background density of the document, and pixels below the removal threshold corresponding to the density are removed, but in general, the background density of the document is estimated, It is not easy to determine the background removal threshold. As a technique for determining the background removal threshold, for example, in Patent Document 1, a histogram within the background density range of a document is created from input image data obtained by reading the document during pre-scanning, and the background removal threshold is determined. In addition, it describes that background removal is performed on input image data at the time of main scanning. Patent Document 2 describes that a background removal threshold is determined for each page. Further, Patent Document 3 describes, as a method for determining the background removal threshold using a histogram, how far the peaks of the histogram are separated by a dispersion calculation and determining the background removal threshold therefrom. Has been.
JP-A-4-313774 JP 2000-101839 A JP-A-7-303188

  The techniques described in Patent Document 1 and Patent Document 2 determine a background removal threshold value in page units. For this reason, the background removal threshold is determined after pre-scanning or after reading one page of data, and there is a problem that the background removal processing is delayed. In addition, the technique described in Patent Document 3 has a problem that the variance has to be calculated based on a histogram, and the processing becomes complicated.

  The present invention has been made in view of such circumstances, and an object of the present invention is to provide an image processing apparatus and a background removal method that can determine a background removal threshold at high speed with simple processing.

  In order to solve the above problems, the image processing apparatus according to the first aspect of the present invention creates a histogram that reads image data obtained by reading a document in a predetermined unit and creates a histogram indicating the number of pixels for each density. Means, background density estimation means for estimating the background density of the document based on the maximum value on the highlight side of the histogram, and the ratio of the number of pixels corresponding to the estimated background density to the total number of pixels is larger. The background removal reference value setting means for setting the background removal reference value with a tendency to decrease, and the read image data of the predetermined unit based on the estimated background density and the set background removal reference value. A ground removal means for removing the ground.

  According to the present invention, since the background density of the original is estimated based on the maximum value on the highlight side of the histogram, the background removal threshold value determination process can be performed at high speed with a simple process. In general, the greater the ratio of the number of pixels corresponding to the estimated background density to the total number of pixels, the higher the sharpness of the peak of the histogram and the less the background density blur. The smaller this ratio, the higher the peak of the histogram. It is considered that the sharpness of the film becomes lower and the blur of the background density increases. For this reason, by setting the background removal reference value such that the background removal reference value tends to decrease as the ratio of the number of pixels corresponding to the estimated background density to the total number of pixels increases, the accuracy of background removal can be improved. it can.

  Here, the background density estimating means can estimate the background density of the document based on the first local maximum value on the highlight side exceeding a predetermined threshold. As a result, it is possible to prevent the maximum value having a small number of pixels from being erroneously recognized as the background density of the document.

  As for the predetermined threshold, it is desirable that the first threshold is set in a predetermined range on the highlight side, and a second threshold smaller than the first threshold is set in other ranges. Accordingly, it is possible to prevent erroneous recognition of the background density when the image data includes the image of the platen cover.

  The predetermined unit may be a line unit. By setting the background removal threshold value for each line, the background removal threshold value determination process can be performed at higher speed.

  In this case, the background removal means weights the total value of the estimated background density and the set background removal reference value and the background removal threshold value of the previous line using a coefficient that determines the specific gravity of both. By adding them together, it is possible to determine a background removal threshold value in the read line, and to remove pixels having a density from white to the background removal threshold value. Thereby, it is possible to prevent the background removal threshold from fluctuating greatly for each line.

  In order to solve the above-described problem, the background removal method according to the second aspect of the present invention is a histogram creation that reads image data obtained by reading a document in a predetermined unit and creates a histogram indicating the number of pixels for each density. A background density estimating step for estimating the background density of the document based on the step, a maximum value on the highlight side of the histogram, and a ratio of the number of pixels corresponding to the estimated background density to the total number of pixels is larger. The background removal reference value setting step for setting the background removal reference value with a tendency to become smaller, and based on the estimated background density and the set background removal reference value, A ground removal step for removing the ground.

  According to the present invention, an image processing apparatus and a background removal method that can determine a background removal threshold at high speed with simple processing are provided.

  Embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a hardware configuration of an image reading apparatus 10 which is an embodiment of an image processing apparatus according to the present invention. As shown in the figure, the image reading apparatus 10 includes a controller 20 and an image reading mechanism 30, and removes the background of the original from the image data of the original read by the image reading mechanism 30 and outputs it to an external device. To do.

  The controller 20 includes a CPU 21, a ROM 22, a RAM 23, and an interface 24. The CPU 21 performs processing according to the program recorded in the ROM 22 to control the document reading operation of the image reading mechanism 30 and to perform image processing for removing the background of the document from the read image data. The RAM 23 is used as a work area at this time. The interface 24 communicates with an external device using a predetermined protocol such as USB.

  The image reading mechanism 30 includes a CCD as a light receiving element, a light source, a document table, a document table cover, a drive motor that scans the CCD, an AD converter that converts a CCD output signal into a digital signal, and the like. Image data of the original is obtained by irradiating the placed original with light and reading the reflected light with a CCD. The document table cover serves to hold down the document placed on the document table and block external light, and to reduce the influence on the image data when the document is smaller than the reading range of the CCD. The side is white. The document table cover may be provided with an auto seed feeder function.

  FIG. 2 is a block diagram illustrating processing function units related to background removal processing of the image reading apparatus 10 according to the present embodiment. As shown in the figure, the image reading apparatus 10 includes an image reading unit 110, an image data storage unit 120, a background removal threshold setting unit 130, a background removal unit 140, and an image output unit 150. These processing function units are realized in software or hardware by causing each unit of the controller 20 and the image reading mechanism 30 shown in FIG. 1 to perform processing alone or cooperatively.

  The image reading unit 110 reads a document placed on a document table and generates image data. The image data storage unit 120 stores image data read by the image reading unit 110. It is also used as a work area when removing the background from the image data. The background removal threshold value setting unit 130 reads image data line by line from the image data storage unit 120 and sets a background removal threshold value for each line. In other words, in the present embodiment, the background removal threshold is determined in units of lines, and background removal is performed in units of lines. For this reason, it is possible to remove the base at high speed. The background removal unit 140 performs background removal on a line basis from the image data stored in the image data storage unit 120 based on the background removal threshold set by the background removal threshold setting unit 130. The image output unit 150 outputs the image data with the background removed to an external device.

  FIG. 3 is a block diagram illustrating a detailed configuration of the background removal threshold value setting unit 130. As shown in the figure, the background removal threshold setting unit 130 includes a histogram creation unit 131, a histogram smoothing unit 132, a background density D estimation unit 133, a background density D pixel number normalization unit 134, a background removal reference value α setting unit, a background A removal threshold determination unit 136, a background removal threshold initial value storage unit 137, and a previous line background removal threshold storage unit 138 are provided.

  The histogram creation unit 131 reads the image data stored in the image data storage unit 120 for each line and creates a histogram. That is, a histogram for pixel values for one line instead of one page is created.

  For example, the histogram may be created by converting each pixel value of the read image data into luminance data, but here, the histogram is created with G (green) data values close to the luminance data among the RGB data. It shall be. If G is represented by 8 bits for each pixel, the horizontal axis represents a density of 0 to 255, and the vertical axis represents a histogram representing the number of pixels. The total number of pixels is equal to the number of pixels constituting one line. Here, 0 indicates the lightest state (highlight), and 255 indicates the darkest state (shadow). FIG. 4A shows an example of an 8-bit input histogram.

  The histogram smoothing unit 132 smoothes the peaks of the 8-bit input histogram created by the histogram creating unit 131. Here, as shown in FIG. 4B, smoothing is performed by reconstructing a histogram using the upper 5 bits of 8-bit data. As will be described later, in this embodiment, the background density is estimated by detecting the highest peak (maximum value) of the histogram. For this reason, if the up / down change of the histogram is sensitive, the background density may be erroneously detected. In addition, the 5 bits are set to 32 steps of accuracy in 8 steps with respect to 255 steps of the input 8 bits, so that practically sufficient smoothing can be achieved and an error of about 8 with respect to 255 steps is practically no problem. It is possible. On the other hand, if the upper 6 bits are set to 4 increments, sufficient smoothing cannot be performed, and if the upper 4 bits are incremented by 16 increments, the background density is considered to increase.

  The background density D estimation unit 133 estimates the background density D based on the smoothed histogram. Basically, the density corresponding to the first peak (maximum value) from the highlight side of the histogram is estimated as the background density D. For this reason, the background density D can be estimated by simple processing. For example, as shown in FIG. 5A, in a histogram obtained from a document with a low background density and close to white, the density corresponding to the first peak Pk1 is estimated as the background density D. Also, in the histogram obtained from a document with a dark background such as a newspaper as shown in FIG. 5B, the density corresponding to the first peak Pk2 is estimated as the background density D.

  However, when a mountain (Pk1) having a small number of pixels is formed as shown in FIG. 5C, the density corresponding to Pk1 is erroneously determined as the background density D. Therefore, as shown in FIG. 5D, a threshold value Th for the number of pixels is determined, and erroneous determination is prevented by not recognizing the number of pixels equal to or less than the threshold value Th as a mountain. In this case, Pk2 that first exceeds the threshold Th from the highlight side is detected as a mountain.

  By the way, when the document is smaller than the reading range of the CCD, the image data includes the image on the document table cover. When the image of the platen cover is included in the image data, the number of pixels having a density corresponding to the color of the platen cover increases in the histogram. Since the document table cover is white in order to reduce the influence on the read image data, a peak exceeding the threshold Th may be formed on the highlight side as shown in FIG. Since this mountain has nothing to do with the background of the document, it must be removed from the detection target.

  Therefore, as shown in FIG. 5 (f), two threshold values Th1 and Th2 (Th1> Th2) are prepared as threshold values for the number of pixels, and the range of Wt from the highlight side corresponding to the color of the document table cover is The threshold value Th1 is applied, and the threshold value Th2 having a value smaller than the threshold value Th1 is applied to other ranges. As a result, Pk1 based on the color of the platen cover is not recognized as a mountain, and Pk2 that first exceeds the threshold Th2 from the highlight side is detected as a mountain.

  Here, the threshold Th can be specified as a percentage of the total number of pixels in the histogram. For example, when the threshold value Th is designated as n% with respect to all pixels, the threshold value Th = (number of pixels in one line × n) / 100 can be obtained. However, what n is used is determined experimentally for each of the threshold Th1 and the threshold Th2. The width Wt to which the threshold value Th1 is applied is also set in advance based on the color of the document table cover.

  Note that the background density D estimation unit 133 outputs to the background removal threshold value determination unit 136 that the background density D cannot be detected when a peak exceeding the threshold value cannot be detected as shown in FIG. As will be described later, for a line in which a peak exceeding the threshold value cannot be detected, background removal is performed using the same background removal value as the previous line (see step S108 in FIG. 10). For this reason, for a line for which the background density D cannot be detected, a message to that effect is output to the background removal threshold value determination unit 136.

  The background density D pixel number normalization unit 134 normalizes the number of pixels corresponding to the background density D estimated by the background density D estimation unit 133. Since the number of pixels included in one line varies depending on the resolution and the document width, the number of pixels corresponding to the background density D is normalized in determining the background deletion amount. Here, it is assumed that the number of pixels is normalized in the range of 0 to 1000 by the following formula. That is, the normalized background density D pixel count = the background density D pixel count × the number of pixels included in the 1000/1 line.

  The background removal reference value α setting unit 135 sets a background removal reference value α that is a reference for the background removal threshold for the background density D. Here, the background removal reference value α is an amount widened from the background density D to the shadow side, as shown in FIGS. 7 (a) and 7 (b). A value obtained by adding the background removal reference value α to the background density D (background density D + α) becomes the temporary background removal threshold of the line. The reason is that the final background removal threshold value is determined in relation to the background removal threshold value of the previous line, as will be described later. However, in order to further simplify the processing, a value obtained by adding the background removal reference value α to the background density D (background density D + α) may be used as the background removal threshold of the line.

  Here, the background removal reference value α is determined so as to decrease as the number of normalized background density D pixels increases and to increase as the number of normalized background density D pixels decreases. In general, as the number of normalized background density D pixels is larger, as shown in FIG. 7A, the sharpness of the mountain is higher and the background density is less blurred, and as the number of normalized background density D pixels is smaller, FIG. This is because, as shown in (b), it is considered that the sharpness of the mountain is lowered and the blur of the background density is increased.

  FIG. 8 is a diagram illustrating an example of a method for setting the background removal reference value α. In this example, a function is used in which the background removal reference value α is determined according to the normalized background density D pixel count. This function indicates that the background removal reference value α is y1 when the normalized background density D pixel number is less than x1, and the background removal reference value α is y2 when the normalized background density D pixel number is x2 or more ( However, when x1 <x2, y1> y2) and the normalized background density D pixel number is not less than x1 and less than x2, the background removal reference value α is ((y2−y1) × normalized background density D pixel number + y1 × x2). -Y2 * x1) / (x2-x1). Here, x1, x2, y1, and y2 are experimentally determined in the development stage.

  The background removal threshold value determination unit 136 determines the background removal threshold value Cut (N) of the line N from the temporary background removal threshold value (background density D + α) of the line N and the background removal threshold value Cut (N−1) of the previous line. To decide. The background removal threshold Cut (N) is determined by the density of 8-bit or 5-bit G data, and pixels having the density lower than the density are removed by the background removal unit 140. That is, a histogram before smoothing may be used, or a histogram after smoothing may be used.

  The previous line background removal threshold value Cut (N−1) is stored in the previous line background removal threshold value storage unit 138. However, in the case of the first line, since the background removal threshold value of the previous line does not exist, the background removal threshold value Cut is obtained using the initial value stored in the background removal threshold value initial value storage unit 137 as Cut (0). Determine (1). For this reason, the background removal threshold initial value storage unit 137 stores a standard initial value of the background removal threshold in advance. However, the background removal threshold value Cut (1) may be determined using only the temporary background removal threshold value of the first line without using the background removal threshold value initial value storage unit 137.

  Specifically, the background removal threshold Cut (N) is determined by the following equation. That is, the background removal threshold Cut (N) = (1−W) × (background density D + α) + W × Cut (N−1). Here, since (background density D + α) is the temporary background removal threshold of the line N, the background removal threshold Cut (N) is calculated using the weighting factor W and the temporary background removal threshold of the line and the previous line. Are the values obtained by weighting and adding (1-W) and W respectively.

  That is, the larger the W, the greater the influence of the background removal threshold of the previous line, and the smaller the W, the greater the influence of the temporary background removal threshold of that line. In this embodiment, it is assumed that W is set so that the influence of the line itself is relatively large as the document starts to be read, and thereafter, the influence of the lower removal threshold of the previous line becomes large. Thus, the background removal threshold value is prevented from changing abruptly for each line. Further, since the line at the beginning of reading is the edge of the document, there are many background portions, and it is expected that the estimation accuracy of the background density D is high, so the influence of the line itself is increased. Then, in the intermediate portion where the estimation accuracy of the background density D may be lowered, the influence of the reading start portion is increased.

  Specifically, W as shown in FIG. 9 is used. In the example of this figure, the first line at the start of reading is set to W = 0.5, W gradually increases so that the line in the B-th row becomes W = 1.0, and thereafter, W = 1.0. It is determined to be. The number of lines in the Base portion from the 0th row to the Bth row is determined experimentally in the design stage.

  In order to prevent the background removal threshold Cut (N) from becoming excessively large, a background removal threshold threshold ThC is determined. If the calculated Cut (N) exceeds ThC, the background removal threshold is set to ThC. To do.

  Next, the flow of the background removal threshold value setting process in the background removal threshold value setting unit 130 of the present embodiment will be described with reference to the flowchart of FIG. First, 1 is set as the line N to be processed (S101). Then, the Nth line of the image data is read from the image data storage unit 120 (S102). Then, a histogram is created based on the read image data (S103). As described above, in this embodiment, it is assumed that a histogram is created using the G value indicated by 8 bits.

  Next, the background density D is estimated (S104). FIG. 11 is a flowchart illustrating the background density D estimation process (S104). In the background density D estimation process (S104), the histogram is first smoothed (S1041). In the smoothing of the histogram, as described above, the histogram is reconstructed using the upper 5 bits of the 8 bits.

  A threshold having a different value for each density range as shown in FIG. 5F is applied to the smoothed histogram (S1042). Specifically, a threshold value Th1 is set on the highlight side, and a threshold value Th2 lower than Th1 is set.

  Then, the first peak on the highlight side exceeding the threshold is detected (S1043). If detected (S1043: Yes), the density corresponding to the mountain is estimated as the background density D (S1044), and if not detected (S1043: No), the background density D estimation failure (S1045).

  If the background density D is estimated as a result of the background density D estimation process (S104) (S105: Yes), a background removal reference value α is set (S106). In setting the background removal reference value α, the number of pixels corresponding to the background density D is normalized, and the background removal reference value α is set according to a predetermined function as shown in FIG. As described above, the background removal reference value α is determined so as to decrease as the number of normalized background density D pixels increases and to increase as the number of normalized background density D pixels decreases.

  When the background removal reference value α is set, the background removal threshold Cut (N) for the Nth line is calculated (S107). The background removal threshold Cut (N) of the Nth line is set to the temporary background removal threshold (background density D + α) of the Nth line and the background removal threshold Cut (N−1) of the N−1th line. Determined based on. The specific gravity at this time is determined by the weighting factor W. As for the coefficient W, the specific gravity of the temporary background removal threshold (background density D + α) of the Nth line is relatively large in the reading start portion of the document, and thereafter, the specific gravity of the background cut amount of the N−1th line becomes large. It is prescribed as follows. However, the background removal threshold Cut (N) is determined so as not to exceed the predetermined threshold ThC.

  On the other hand, if the background density D cannot be estimated as a result of the background density D estimation process (S105: No), the background removal threshold Cut (N-1) of the N-1th line is set to the Nth line background removal threshold. (S108).

  When the background removal threshold Cut (N) for the Nth line is determined, the previous line background removal threshold storage unit 138 is updated with the determined background removal threshold Cut (N) (S109). Then, the determined background removal threshold Cut (N) is notified to the background removal unit 140, and the background removal unit 140 executes the background removal of the Nth line (S110). In the background removal, pixels having a density equal to or lower than the background removal threshold Cut (N) are removed.

  If there is an unprocessed line (S111: No), N = N + 1 is set (S112), and the process for the next line is performed. When the background removal processing for all lines is completed (S111: Yes), this processing ends.

  As described above, according to the present embodiment, an image processing apparatus that can determine the background removal threshold at high speed with simple processing is provided. Since the background removal threshold can be determined by simple processing, the load on the CPU 21 spent on the background removal processing is reduced, and the load can be assigned to other processing. In addition, since the background removal threshold can be determined at high speed, it is particularly effective in a case where high speed is required, such as when obtaining the background removal amount for each line as shown in the present embodiment. Furthermore, when the processing function unit related to the background removal processing is realized by hardware, a small-scale and simple circuit configuration can be realized, so that the development period and the evaluation man-hour can be reduced.

It is a block diagram which shows the hardware constitutions of an image reading apparatus. 3 is a block diagram illustrating a processing function unit related to background removal processing of the image reading apparatus. FIG. It is a block diagram which shows the detailed structure of a background removal threshold value setting part. It is a figure explaining the smoothing of a histogram. It is a figure explaining detection of background density D. It is a figure which shows the example which cannot detect background density | concentration D. FIG. It is a figure explaining the background removal reference value α. It is a figure explaining the setting method of background removal reference value (alpha). It is a figure explaining the weighting coefficient W. FIG. It is a flowchart explaining the background removal threshold value setting process of a background removal threshold value setting part. It is a flowchart explaining a background density | concentration D estimation process.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Image reading apparatus, 20 ... Controller, 21 ... CPU, 22 ... ROM, 23 ... RAM, 24 ... Interface, 30 ... Image reading mechanism, 110 ... Image reading part, 120 ... Image data storage part, 130 ... Background removal threshold value Setting unit 131 ... Histogram creation unit 132 ... Histogram smoothing unit 133 ... Background density D estimation unit 134 ... Background density D pixel number normalization unit 135 ... Background removal reference value α setting unit 136 ... Background removal threshold determination 137 ... Background removal threshold initial value storage section, 138 ... Previous line background removal threshold storage section, 140 ... Background removal section, 150 ... Image output section

Claims (6)

Histogram creation means for reading image data obtained by reading a document in a predetermined unit and creating a histogram indicating the number of pixels for each density;
A background density estimating means for estimating a background density of the document based on a maximum value on a highlight side of the histogram;
A background removal reference value setting means for setting a background removal reference value with a tendency to decrease as the ratio of the number of pixels corresponding to the estimated background density increases with respect to the total number of pixels;
An image processing apparatus comprising: a background removing unit that removes a background of the read image data of a predetermined unit based on the estimated background density and the set background removal reference value. The image processing apparatus according to claim 1,
The image processing apparatus according to claim 1, wherein the background density estimating means estimates the background density of the document based on an initial maximum value on a highlight side exceeding a predetermined threshold. The image processing apparatus according to claim 2,
As the predetermined threshold, a first threshold is defined in a predetermined range on the highlight side, and a second threshold smaller than the first threshold is defined in the other range. Image processing device. The image processing apparatus according to any one of claims 1 to 3,
The image processing apparatus according to claim 1, wherein the predetermined unit is a line unit. The image processing apparatus according to claim 4,
The background removal means adds the weight of the total value of the estimated background density and the set background removal reference value and the background removal threshold of the previous line using a coefficient that determines the specific gravity of both. An image processing apparatus characterized by determining a background removal threshold in the read line by combining the pixels, and removing pixels having a density from white to the background removal threshold. A histogram creating step of reading image data obtained by reading a document in a predetermined unit and creating a histogram indicating the number of pixels for each density;
A background density estimation step for estimating a background density of the document based on a maximum value on a highlight side of the histogram;
A background removal reference value setting step for setting a background removal reference value with a tendency to decrease as the ratio of the number of pixels corresponding to the estimated background density to the total number of pixels increases.
A background removal method comprising: a background removal step of removing a background of the read predetermined unit of image data based on the estimated background density and the set background removal reference value.