JP3260891B2 - Edge extraction method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for detecting an edge of a sheet of a document from color image data read by a color scanner, a color copier or the like.

[0002]

2. Description of the Related Art When a thick paper piece (document) is copied by a color copier, noise (shadow) peculiar to an optical system remains at an edge portion of the paper piece, which deteriorates the quality of the copy document. When a thick paper piece is read by an independent color scanner or a color scanner with a built-in facsimile, similar noise occurs in the read image.

Further, when copying a paper piece with a color copier, the paper sheet may be skewed due to a displacement of the paper piece when the cover is hung or a malfunction of the automatic sheet feeder. This skew causes image deterioration, and this deterioration is particularly remarkable in a digital copier that performs pixel-by-pixel processing.

If the edge of a sheet of paper (original document) is detected from the color image data and appropriate processing is performed on the edge portion, or if the edge is detected and the skew is obtained and appropriate processing is performed in accordance with the skew, It is possible to improve image deterioration due to edge shadows and skew.

However, with respect to such edge detection of a piece of paper, there is no suitable known technique for color images.

[0006]

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a means for detecting the edge of a sheet from color image data obtained by reading the sheet with a color scanner or the like.

More specifically, an object of the present invention is to provide a method for detecting the edge of a sheet from color image data when the background color of the sheet during reading is the same or almost the same as the background color of the sheet. It is to provide. Another object of the present invention is to provide a method for detecting the edge of a sheet from color image data when the background color of the sheet during reading is different from the background color of the sheet.

[0008]

According to a first aspect of the present invention, a luminance is obtained from color image data obtained by reading a sheet of paper, and a valley point of luminance satisfying a specific condition is determined .
It is extracted as an edge candidate point of a sheet of paper. The conditions as edge candidate points include the valley depth and valley width for the luminance valley point.
Including further including the chroma or chromaticity determined from the color image data is substantially flat at the location of the valley points of the luminance
You .

According to the second aspect of the present invention, the saturation is determined from the color image data obtained by reading the paper piece or the data after smoothing the color image data, and the saturation valley satisfying a specific condition is obtained.
Points and points that have moved their valley points are extracted as edge candidate points. The color image data is smoothed with different mask sizes, and the saturation can be obtained from the data after smoothing with each mask size. The condition as an edge candidate point can include a valley depth related to a saturation valley point, and can include a mutual relationship between the saturation valley points after smoothing by each mask size.

[0010]

According to the first aspect of the present invention, as will be described in detail later with reference to a first embodiment, it is possible to reliably detect the edge of a paper piece having the same or close to the background color at the time of reading. By including the saturation or chromaticity in the extraction condition, the discrimination performance between the edge of the paper piece and the edge of the color picture or the like on the paper piece is improved.

According to the second aspect of the present invention, as will be described in detail later with reference to a second embodiment, it is possible to reliably detect an edge of a paper piece having a background color different from a ground color at the time of reading. Become.

[0012]

Embodiment 1 FIG. 1 is a block diagram showing the configuration of this embodiment. In FIG. 1, a color image input device 100 is a device that optically scans and reads a color original and inputs color image data to an edge detection device 200 and a color image processing device 300, and is, for example, a general color scanner. . In this embodiment, the color image data is RGB data. The edge detection device 200 detects a candidate point of the edge of the paper piece from the RGB data, and inputs the coordinate data to the color image processing device 300.

Since the reading range of the color image input device 100 is wider than that of the color document, the background outside the color document is also read at the same time. The edge detection device 200 according to the present embodiment is configured such that the background color and the background color of the document read by the color image input device 100 are the same or substantially the same (for example, the background is white and the background color of the color document is Is white).

The color image processing device 300 extracts the edge of the paper piece as a straight line based on the coordinate data of the candidate edge point input from the edge detection device 200, and removes the shadow of the extracted edge portion from the input color image. Skew is obtained from the edge, and skew correction is performed on the input color image.

In the edge detecting device 200, the luminance calculating section 201 calculates the luminance L of each pixel from the RGB data according to the equation (1), and the R chromaticity calculating section 203 calculates the R (red) color of each pixel from the RGB data. The degree r is calculated by the equation (2), and the G chromaticity calculating unit 204 calculates the G of each pixel from the RGB data.
The chromaticity g of (green) is calculated by equation (3), and the B chromaticity calculating unit 205 calculates the chromaticity b of B (blue) of each pixel from the RGB data.
Is calculated by equation (4). Also, the saturation calculation unit 202
Calculates the saturation S of each pixel from the RGB data by the equation (5).

L = (R + G + B) / 3 (1) r = 255 R / (R + G + B) (2) g = 255 G / (R + G + B) (3) b = 255 B / (R + G + B) (4) S = 255-3 min (r) , G, b) = 255 max (G + B−2R, B + R−2G, R + G−2B) / (R + B + G) (5) The luminance valley point extraction threshold value calculation unit 206 calculates the luminance valley depth. A threshold value of the depth of the luminance valley is calculated from the distribution of the height.
The saturation threshold calculating unit 208 calculates a saturation valley depth threshold or a saturation gradient threshold from the saturation valley depth distribution or the saturation gradient distribution. I do. The R chromaticity threshold value calculation unit 209, the G chromaticity threshold value calculation unit 210, and the B chromaticity threshold value calculation unit 211 respectively determine the depth or inclination of the valley of R chromaticity, G chromaticity, and B chromaticity. , The threshold of the depth of each chromaticity valley or the threshold of the gradient of saturation is calculated. A method of calculating each threshold will be described later.

The luminance valley point extracting unit 207 extracts a luminance valley point that satisfies one of the following conditions, or a combination thereof, or all valley points (unconditional). Regarding the extraction condition, the edge detection device 200 can select various modes.

1. The height (the depth of the valley) from the bottom of the valley to one of the left and right peaks is larger than the threshold calculated by the threshold calculator 206. 2. The difference between the heights of the left and right peaks is equal to or less than a predetermined threshold. 3. The distance between the peaks (valley width) on both the left and right sides is equal to or less than a predetermined threshold. 4. The valley depth is greater than or equal to a predetermined threshold.

The edge extracting unit 212 includes a luminance valley point extracting unit 2
07, a candidate that satisfies the following conditions relating to one or more of saturation, R chromaticity, G chromaticity, and B chromaticity is extracted as an edge candidate point. Are extracted as edge candidate points (that is, unconditionally), and their coordinate data is output. Regarding the extraction conditions (including unconditional conditions), the edge detection device 200 can select various modes.

1. There is no saturation or chromaticity valley point in the luminance valley point. 2. There is no saturation or chromaticity valley point having a depth equal to or greater than the predetermined threshold value in the luminance valley point. 3. At the luminance valley point, there is no saturation or chromaticity valley having a depth equal to or greater than the threshold calculated by the threshold calculator 206. 4. There are no saturation or chromaticity peaks at the luminance valley points. 5. There are no peaks of saturation or chromaticity at a height equal to or higher than a predetermined threshold value at the luminance valley point. 6. There are no peaks of chroma or chromaticity having a height equal to or higher than the threshold calculated by the threshold calculator (208 to 211) at the luminance valley. 7. At the luminance valley point, the slope of the saturation or chromaticity is smaller than a predetermined threshold.

In the above-mentioned condition judgment, a color arbitrarily specified from the chroma and the RGB chromaticities is used. When a plurality of items are designated, for example, when saturation and R chromaticity are used, both of them must satisfy the above conditions.

Next, an operation example of the edge detection device 200 will be described. In the following description, the scanning line 2 of the color image 1 shown in the upper part of FIG. 2 will be used as an example. This color image 1
A sheet of paper (original) is read by the color image input device 100 with a background of the same color as the ground color.
5 indicates a background portion. The middle part of FIG. 2 shows the luminance on the scanning line 2, and the lower part shows the saturation on the scanning line 2. The luminance or saturation of the valley point or the peak of the luminance or saturation and the coordinates (x-coordinate) are added in parentheses in the vicinity thereof.

[Operation Example 1] In this operation example, the luminance calculation section 201, the luminance valley point extraction section 207, and the edge extraction section 212
Only involved.

The luminance calculator 201 calculates the luminance data shown in the middle part of FIG. 2 from the RGB data of the scanning line 2 input from the color image input device 100. The luminance valley point extracting unit 207 extracts all valley points from the luminance data.
That is, a, b, c, d, e, f, g,
9 points of h and i are extracted. The edge extracting unit 212 unconditionally sets the nine luminance valley points as edge candidate points, and inputs the respective coordinate data (Table 1) to the color image processing apparatus 300.

[0025]

[Table 1]

In the case of this operation example, a true edge point a
Since seven points other than the point and the h point are detected as edge candidate points, the detection accuracy of the edge points is poor. However, since true edge points are definitely detected, there is no particular problem if the color image processing device 300 can correctly extract edges as straight lines from such edge candidate points. That is, depending on the edge extraction algorithm in the color image processing device 300, the object can be sufficiently achieved with the edge detection accuracy as in this operation example.

[Operation Example 2] In this operation example, the luminance calculation unit 201, the luminance valley point extraction unit 207, and the edge extraction unit 212
Only involved. The luminance valley point extraction unit 207 determines that the difference between the height of the left mountain and the height of the right mountain among the valley points a to i of the luminance data shown in the middle part of FIG. )
Only the following valley points are extracted as effective luminance valley points.

Point a has a height from the bottom to the left mountain of 4 points.
8 (= 48-0), and the height to the right mountain is 52 (= 52−
0), and the difference is 4 or less and 5 or less. Therefore, the point a is extracted as a valid valley point. In the case of point d, the difference between the heights of the right and left mountains is | (26-0)-(1-0) | = 25
Is larger than 5, and is not extracted as a valid valley point. Thus, in scan line 2, a,
The five points b, f, h, and i are regarded as valley points, and
Sent to The edge extracting unit 212 unconditionally sets these valley points as edge candidate points and sends the coordinate data to the color image processing apparatus 300.

Also in this operation example, true edge points a and a
Three points b, f, and i are extracted in addition to the point h (the detection accuracy is higher than in the operation example 1). However, as described in the first embodiment, there is no particular problem in this operation example depending on the edge extraction processing algorithm on the color image processing apparatus 300 side.

[Operation Example 3] In this operation example, the luminance calculation unit 201, the luminance valley point extraction unit 207, and the edge extraction unit 212
Only involved. As a first step, the luminance valley point extracting unit 207 performs, as in the operation example 2, the height of the left mountain and the height of the right mountain among the valley points a to i of the luminance data shown in the middle part of FIG. Only the valley points whose difference is less than a predetermined threshold value (here, 5) are extracted. In the second step, only the valley points having the interval between the left and right peaks, that is, the valley width of which is equal to or less than a predetermined threshold (here, 100), among the valley points extracted in the first step, are finally set to the effective luminance. Extract as valley points.

In the case of scan line 2, a, b,
Five points f, h, and i are extracted. The width of the valley at point a is 4 (=
Since 78-74) is 100 or less, it is an effective valley point. The width of the valley at point b is 165 (= 243-78) and 100
Because it is larger, point b is not extracted. In this way, the four points a, f, h, and i are input to the edge extraction unit 212 as luminance valley points. The edge extraction unit 212 unconditionally sets these valley points as edge candidate points and outputs coordinate data thereof.

Also in this operation example, true edge points a and a
In addition to the h point, two points f and i are extracted (the operation example 2 described above).
Detection accuracy is higher). However, as described in the above embodiment, there is no particular problem in this operation example depending on the edge extraction processing algorithm on the color image processing apparatus 300 side.

[Operation Example 4] In this operation example, the luminance calculation section 201, the luminance valley point extraction section 207, and the edge extraction section 212
Only involved.

The luminance valley point extracting unit 207 determines, from the luminance valley points, that the difference between the heights of the left peak and the right peak is equal to or smaller than a predetermined threshold value (here, 5), as in the operation example 3. And a valley point whose valley width is equal to or less than a predetermined threshold value (here, 100) is extracted. In the case of the scanning line 2, four points a, f, h, and i are extracted.

The luminance valley point extracting unit 207 further includes a, f,
Among the four points h and i, a valley whose valley depth (here, the larger value of the height to the left and right peaks is the valley depth) is equal to or greater than a predetermined threshold value (here, 20). The point is finally extracted as a valid luminance valley point. For point a, the depth max of the valley
Since ((48-0), (52-0)) = 52> 20, it is extracted as a valley point. In the case of point i, the valley depth max
Since ((48-45), (46-45)) = 3 <20, it is not extracted as a valley point.

Thus, finally, a, f, and h of 3
Points are extracted and input to the edge extraction unit 212. The edge extraction unit 212 outputs these three points as edge candidate points as they are, and outputs their coordinate data.

As described above, in this operation example, only the point f is detected other than the true edge points a and h, so that the detection accuracy is higher than that in the operation example.

[Operation Example 5] In this operation example, in addition to the luminance calculating section 201, the luminance valley point extracting section 207, and the edge extracting section 212, a luminance valley point extracting threshold value calculating section 206 is involved.

The luminance valley point extraction threshold value calculation unit 206
A histogram of the depth of the valley of the luminance (the larger value of the height to the left and right peaks) is created from the luminance data, and the depth of the valley corresponding to the valley of the histogram is determined as the valley for extracting the luminance valley point. Is determined as the threshold value of the depth. When there are a plurality of valleys in the histogram, the depth of the valley corresponding to the deepest valley is set as the threshold. In the case of the scanning line 2, a histogram as shown in FIG. 3 is obtained. Since the deepest valley of this histogram has a valley depth of 10 to 19,
The representative value 15 is set as a threshold.

As another method for determining the threshold of the valley depth from the histogram of the valley depth, a dissertation “Binarization Method of Drawing Utilizing Shape of Luminance Distribution” (Toyama / Ito,
IEICE D-II Vol. J74 No.
5, pp. 648-654) can be employed.

As in the operation example 3, the luminance valley point extracting unit 207 determines that the difference between the heights of the right and left peaks is less than or equal to a predetermined threshold value (here, 5) from among the luminance valley points a to i. , And only valley points whose valley width is less than or equal to a predetermined threshold value (here, 100) are extracted. That is, at this stage, only the four points a, f, h, and i remain. Next, among these four points, only those having a valley depth equal to or greater than the valley depth threshold (15 in this case) calculated by the threshold calculator 206 are finally set to the effective luminance. Extract as valley points.

In the case of the point a, max ((48-0), (5
2-0)) = 52> 15, which is an effective valley point.
In the case of point i, max ((48-45), (46-4
5)) = 3 <15, so it is not extracted. In this way, only the three points a, f, and h are extracted as valid luminance valley points and input to the edge extraction unit 212.

The edge extracting section 212 uses those luminance valley points as edge candidate points as they are, and inputs the coordinate data to the color image processing apparatus 300. Therefore, according to this operation example, the same detection accuracy as that of the operation example 4 can be achieved.

[Operation Example 6] In this operation example, the luminance calculating section 201, the luminance valley point extracting section 207, the edge extracting section 212,
Luminance valley point extraction threshold value calculation unit 206 and saturation calculation unit 20
8 are involved.

The luminance valley point extraction threshold value calculation unit 206 calculates
The threshold value of the depth of the luminance valley point is obtained in the same manner as in the fifth embodiment. Using this threshold value, the difference between the heights of the right and left peaks is equal to or smaller than a predetermined threshold value (here, 5) and the width of the valley is smaller than the threshold value in the luminance valley point extraction unit 207 as in the operation example 5. A luminance valley point whose value is equal to or less than a predetermined threshold value (here, 100) and whose valley depth is equal to or more than the threshold value (here, 15) calculated by the threshold value calculation unit 206 is extracted. Thus, in the case of the scanning line 2, three points a, f, and h are extracted. Further, the saturation is calculated by the above-described equation (5) in the saturation calculation unit 202, and the saturation data is input to the edge extraction unit 212.

The edge extracting unit 212 extracts a valley point of saturation from the saturation data. Since the scanning line 2 has saturation as shown in the middle part of FIG. 2, a, j, and
Five points d, e, and f are extracted. Then, among the luminance valley points extracted by the luminance valley point extraction unit 207, only those which are not the valley points of the saturation are determined as the edge candidate points, and the coordinate data is output. Therefore, only the point h is extracted in the scanning line 2.

[Operation Example 7] In this operation example, in the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, and the saturation calculation section 20
8 are involved.

A luminance valley point is extracted in the same manner as in the operation example 5. The edge extraction unit 212 extracts the saturation peaks, and among the luminance valley points, those that are not the saturation peaks are set as edge candidate points. Except for this, the operation is the same as the operation example 6. Since the point h is the peak of the saturation in the scanning line 2, only the point a among the luminance valleys a, j, h is an edge candidate point.

[Operation Example 8] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, the R, G, and B chromaticity calculations are performed. The units 203, 204, 205 are involved.

A luminance valley point is extracted in the same manner as in the operation example 5. The R, G, and B chromaticity calculation units 203, 204, and 205 calculate the chromaticity of R, G, and B, and the edge extraction unit 212 extracts a valley point or a peak point of the chromaticity from the chromaticity data. Then, in the edge extracting unit 212, only the valley points or the hill points of any chromaticity among the luminance valley points are set as edge candidate points. Note that valley points or peak points may be extracted for one or two chromaticities of R, G, and B.

[Operation Example 9] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, and the saturation calculation section 20
8 are involved.

A luminance valley point is extracted in the same manner as in the fifth embodiment. The edge extracting unit 212 calculates the gradient of the saturation at the luminance valley point, and sets the luminance valley point having the saturation gradient smaller than a predetermined threshold value as an edge candidate point.

[Operation Example 10] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, the R, G, and B chromaticity calculations are performed. The units 203, 204, 205 are involved.

A luminance valley point is extracted in the same manner as in the fifth embodiment. In the edge extraction unit 212, R,
The gradients of the chromaticity of G and B are calculated, and a luminance valley point where the gradient of the chromaticity is smaller than a predetermined threshold value is set as an edge candidate point. In addition,
A gradient for one or two chromaticities of R, G, and B may be extracted, and a luminance valley point whose gradient is equal to or less than a predetermined threshold may be set as an edge candidate point.

[Operation Example 11] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, and the saturation calculation section 2
08 are involved.

A luminance valley point is extracted in the same manner as in the operation example 5. The edge extraction unit 212 extracts a valley point having a saturation whose valley depth is equal to or greater than a predetermined threshold value (here, 20).
The depth of the saturation valley is the greater of the height to the left mountain and the height to the right mountain of the valley point. For point a,
Since max ((50-49), (50-49) = 1 <20, it is not extracted as a saturation valley point, and at point f, the valley depth is 42, so it is extracted as a saturation valley point. You.
Similarly, d, e, and f are extracted as saturation valley points.

The edge extracting unit 212 calculates
Only the non-saturation valley points are extracted as edge candidate points, and their coordinate data is output. For scan line 2,
Among the luminance valley points a, f, h, a and h are extracted as edge candidate points. That is, both true edge points are extracted.

[Operation Example 12] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, and the saturation calculation section 2
08 are involved.

In this operation example, the luminance calculating section 201, the luminance valley point extracting section 207, the edge extracting section 212, the luminance valley point extracting threshold value calculating section 206, and the saturation calculating section 208 are involved.

A luminance valley point is extracted in the same manner as in the operation example 5. The edge extracting unit 212 extracts a mountain peak having a saturation higher than a predetermined threshold. The height of the mountain is the larger of the height from the left valley point and the height from the right valley point. Then, among the luminance valley points, those that are not the peak points of the saturation are extracted as edge candidate points.

[Operation Example 13] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, the R, G, and B chromaticity calculations are performed. The units 203, 204, 205 are involved.

A luminance valley point is extracted in the same manner as in the operation example 5. In the edge extraction unit 212, the valley points of the chromaticity of R, G, and B whose valley depths are equal to or more than a predetermined threshold value, or the R, G, and C heights whose peak heights are equal to or more than a predetermined threshold value. The peak of the chromaticity of B is extracted. Then, a luminance valley point which is not a valley point or a peak point of chromaticity is extracted as an edge candidate point. Note that R, G,
For one or two chromaticities of B, valley points or peak points may be extracted.

[Operation Example 14] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, and the saturation calculation section 2
08, the saturation threshold value calculation unit 208 is involved.

A luminance valley point is extracted in the same manner as in the operation example 5. In the saturation threshold value calculation unit 208, the saturation calculation unit 2
From the saturation data obtained in step 02, a histogram of the depth of the saturation valley (the larger value of the height to the left and right peaks) is created, and the depth of the valley corresponding to the valley of this histogram is calculated as It is obtained as a threshold value of a valley depth for extracting a saturation valley point.
When there are a plurality of valleys in the histogram, the depth of the valley corresponding to the deepest valley is set as the threshold. In the case of the scanning line 2, a histogram as shown in FIG. 4 is obtained.
Since the deepest valley of this histogram has a valley depth in the range of 10 to 19, the representative value 15 is used as the threshold value of the valley depth of saturation.

As another method for determining the threshold value of the valley depth from the histogram of the valley depth of the saturation, the above-mentioned paper "Binarization Method of Drawing Utilizing Shape of Luminance Distribution" is described. It is also possible to adopt a method or the like that has been used.

The edge extraction unit 212 extracts a valley point having a saturation whose valley depth is greater than or equal to the threshold value (here, 15) obtained by the saturation threshold value calculation unit 208 from the saturation data. The depth of the saturation valley point is the larger value of the heights to the left and right peaks. In the case of the scanning line 2, the point a is a valley point of the saturation, and the depth max of the valley ((50−
49), (50-49)) = 1 <15, so that the point a is not extracted as a saturation valley point. f point is max
Since ((50-8), (49-8)) = 42> 15, it is extracted as a saturation valley point. Similarly, d, e,
j is also extracted as a saturation valley point.

The edge extracting unit 212 calculates
Those that are not the valley points of the saturation extracted as described above are regarded as edge candidate points, and their coordinate data are output. In the case of the scanning line 2, a and h are accurately extracted as edge candidate points.

[Operation Example 15] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, and the saturation calculation section 2
08, the saturation threshold value calculation unit 208 is involved.

A luminance valley point is extracted in the same manner as in the operation example 5. In the saturation threshold value calculation unit 208, the saturation calculation unit 2
From the saturation data obtained in step S02, a histogram of the height of the saturation peak (the larger value of the height to the left and right valley points) is created, and the height of the peak corresponding to the valley of this histogram ( (Representative value) is obtained as a threshold value of a peak height for extracting a chroma peak. If there are a plurality of valleys in the histogram, the height (representative value) of the valley corresponding to the deepest valley is set as the threshold.

As another method for determining the threshold value of the peak height from the histogram of the peak heights of the saturation, the above-mentioned paper, “Binarization Method of Drawing Utilizing Shape of Luminance Distribution”, is described. It is also possible to adopt a method or the like that has been used.

The edge extraction unit 212 extracts, from the saturation data, a mountain point having a saturation whose height is equal to or higher than the threshold value obtained by the saturation threshold value calculation unit 208. Then, among the luminance valley points, those which are not the peak points of the chroma extracted as described above are set as edge candidate points, and their coordinate data are output.

[Operation Example 16] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, the R, G, and B chromaticity calculations are performed. The units 203, 204, 205 and the R, G, B chromaticity threshold value calculation units 209, 210, 211 are involved.

A luminance valley point is extracted in the same manner as in the operation example 5. R, G, B chromaticity threshold value calculation units 209, 210, 2
11, the R, G, B chromaticity calculation units 203, 204,
From the R, G, and B chromaticity data obtained in 205, a histogram of the depth of the valley of each chromaticity (the larger value of the height to the left and right peaks) is created, and a histogram corresponding to the valley of the histogram is created. The valley depth (representative value) is obtained as a valley depth threshold value for valley point extraction for each chromaticity. When there are a plurality of valleys in the histogram, the depth (representative value) of the valley corresponding to the deepest valley is set as the threshold value.

As another method for determining the threshold value of the valley depth from the histogram of the chromaticity valley depth, the above-mentioned paper, “Binarization Method of Drawing Utilizing Shape of Luminance Distribution”, is described. It is also possible to adopt a method or the like that has been used.

In the edge extracting unit 212, R, G, B
From the chromaticity data, valley points whose chromaticity is equal to or greater than the threshold value obtained by the R, G, B chromaticity threshold value calculation units 209, 210, 211 are extracted. Then, among the luminance valley points, those that are not valley points of the chromaticity extracted as described above are set as edge candidate points, and their coordinate data are output.

[Operation Example 17] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, the R, G, B chromaticity calculation The units 203, 204, 205 and the R, G, B chromaticity threshold value calculation units 209, 210, 211 are involved.

A luminance valley point is extracted in the same manner as in the operation example 5. R, G, B chromaticity threshold value calculation units 209, 210, 2
11, the R, G, B chromaticity calculation units 203, 204,
From the R, G, and B chromaticity data obtained in 205, a histogram of the peak height of each chromaticity (the larger value of the height to the left and right valley points) is created, and the histogram corresponds to the valley of the histogram. The peak height (representative value) is obtained as a peak height threshold value for peak point extraction of each chromaticity. If there are a plurality of valleys in the histogram, the height (representative value) of the valley corresponding to the deepest valley is set as the threshold.

As another method for determining the threshold value of the peak height from the histogram of the peak height of the chromaticity, the above-mentioned paper “Binarization method of drawing utilizing shape of luminance distribution” is described. It is also possible to adopt a method or the like that has been used.

In the edge extracting unit 212, R, G, B
From the chromaticity data, mountain points whose chromaticity is equal to or greater than the threshold value obtained by the R, G, B chromaticity threshold value calculation units 209, 210, 211 are extracted. Then, among the luminance valley points, those which are not the peak points of the chromaticity extracted as described above are set as edge candidate points, and their coordinate data are output.

[Operation Example 18] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, and the saturation calculation section 2
08, the saturation threshold value calculation unit 208 is involved.

A luminance valley point is extracted in the same manner as in the operation example 5. In the saturation threshold value calculation unit 208, the saturation calculation unit 2
A histogram of the saturation gradient is created from the saturation data obtained in step 02, and the gradient (representative value) corresponding to the valley of the histogram is obtained as a threshold value of the saturation gradient. When there are a plurality of valleys in the histogram, the slope (representative value) corresponding to the deepest valley is set as the threshold.

The edge extracting unit 212 calculates the gradient of the saturation at the luminance valley from the chroma data, and among the luminance valleys, the one whose gradient of the saturation is smaller than the threshold value is determined as an edge candidate point. And outputs the coordinate data.

[Operation Example 19] In this operation example, the luminance calculation section 201, the luminance valley point extraction threshold value calculation section 206, the luminance valley point extraction section 207, the edge extraction section 212, the R, G, and B chromaticity calculations are performed. The units 203, 204, 205 and the R, G, B chromaticity threshold value calculation units 209, 210, 211 are involved.

A luminance valley point is extracted in the same manner as in the operation example 5. R, G, B chromaticity threshold value calculation units 209, 210, 2
In step 11, a histogram of the slope of each chromaticity is created from the R, G, and B chromaticity data, and the slope (representative value) corresponding to the valley of the histogram is obtained as a threshold value of the slope of each chromaticity. When there are a plurality of valleys in the histogram, the slope (representative value) corresponding to the deepest valley is set as the threshold.

In the edge extracting unit 212, R, G, B
From the chromaticity data, the gradient of each chromaticity at the luminance valley point is determined, and among the luminance valley points, those whose chromaticity gradient is smaller than the threshold value are set as edge candidate points, and their coordinate data are output.

Embodiment 2 FIG. 5 is a block diagram showing the configuration of this embodiment. The edge detection device 400 according to the present embodiment includes the color image input device 10.
This configuration is suitable when the background color and background color of the document read at 0 are different.

In the edge detecting device 200, the smoothing unit 401 performs a smoothing process on the color image data, and includes three mask processing units for R, G, and B data. Each of the mask processing units determines the pixel value g after the smoothing represented by the equation (6), where f (x) is the input pixel value of the corresponding color and W is the size of the mask for smoothing.
(X) is output.

[0088]

(Equation 1)

The other smoothing unit 402 also performs a smoothing process on the color image data. Similarly to the above-mentioned smoothing unit 401, the three mask processing units for the R, G, and B data are used. Become. The relationship between the input pixel value and the output pixel value of each mask processing unit is represented by the above (6), but the size W of the mask is smaller than the mask size of the mask processing unit in the smoothing unit 401.

The saturation 0 calculator 403 calculates the saturation from the RGB data input from the color image input device 100 by the above equation (5), and calculates the calculated saturation as “saturation 0”. Call. The saturation 1 calculation unit 404 calculates the saturation by the above equation (5) from the RGB data smoothed by the smoothing unit 401, and the calculated saturation is referred to as “saturation 1”. The saturation 2 calculation unit 404 calculates
From the RGB data smoothed by the smoothing unit 402,
The saturation is calculated by the above equation (5). The saturation is referred to as “saturation 2”.

The saturation 0 threshold value calculation unit 406 calculates the threshold value of the saturation 0 valley depth from the distribution of the saturation 0 valley depth. Similarly, the saturation 1 threshold value calculation unit 40
7 and the saturation 2 threshold value calculation unit 408
From the distribution of the depths of the valleys with saturation 2, the threshold values of the depths of the valleys with saturation 1 and saturation 2 are calculated. A specific example of such a threshold value calculation method will be described later. The edge extraction unit 409 extracts a candidate point of an edge of a paper piece, and a specific example of the extraction condition or algorithm will be described later.

Next, an operation example of the edge detection device 400 will be described. In the following description, the scanning line 12 of the color image 11 shown in the upper part of FIG. 6 will be used as an example. This color image 1
Reference numeral 1 denotes a paper sheet (original) read by the color image input device 100 with a background different from the ground color, 14 denotes an edge of the paper sheet (original), 15 denotes a background portion, and 16 denotes a background portion.
Indicates a character on a piece of paper. The second row in FIG.
, The third row shows the saturation 2 in the scanning line 12, and the lower row shows the saturation 1. The saturation and coordinates (x coordinate) of the valley point or peak point of each saturation are added in parentheses in the vicinity thereof.

[Operation Example 20] In this operation example, only the saturation 0 calculation section 403 and the edge extraction section 409 are involved.

R input from the color image input device 100
From the GB data, the saturation 0 calculation unit 403 calculates
Data having a saturation of 0 as shown in the lower row is obtained. The edge extraction unit 409 detects a valley point of saturation 0 from the saturation 0 data, and outputs coordinate data of each valley point as it is (unconditionally) as an edge candidate point.

In the case of the scanning line 12 shown in FIG.
Edge candidate points are extracted from ten points of c, i, k, l, d, m, e, and n.

In the case of this operation example, c which is a true edge point
Since eight points other than the point and the point e are detected as edge candidate points, the detection accuracy of the edge points is poor. However, as described in the operation example 1 of the first embodiment, since the true edge point is definitely detected, it is possible to correctly extract the edge as a straight line by the color image processing device 300 from such an edge candidate point. If possible, there is no particular problem.

[Operation Example 21] In this operation example, only the saturation 0 calculation section 403 and the edge extraction section 409 are involved.

The edge extraction unit 409 determines that the depth of the valley (the larger of the height to the left peak and the height to the left peak) among the valley points of saturation 0 is a predetermined threshold value. (Here 2
The above valley points are extracted as edge candidate points.

In the case of the scanning line 12 shown in FIG.
There are ten valley points c, i, k, l, d, m, e, n. The depth max of the valley at point a ((24-5), (6-5))
Since = 91 is greater than 20, point a is an edge candidate point. Since the depth of the valley at the point h is 2 and smaller than 20, the point h is not an edge candidate point. Thus, c, d, e
Only three points are edge candidate points.

Point d is the edge of the character and is not the edge point of the target piece of paper, but points c and e are true edge points. As described above, this operation example corresponds to the operation example 20 described above.
Edge detection accuracy is higher than.

[Operation Example 22] In this operation example, only the saturation 0 calculation section 403, the saturation 0 threshold value calculation section 406, and the edge extraction section 409 are involved.

Based on the chroma 0 data obtained by the chroma 0 calculator 403, a chroma 0 threshold calculator 406 creates a histogram of the depth of the chroma 0 valley. The corresponding valley depth (representative value) is set as the threshold value of the valley depth of saturation 0. When the histogram has a plurality of valleys, the depth of the deepest valley in the plurality of valleys is set as the threshold.

In the case of the scanning line 12 in FIG. 6, a histogram as shown in FIG. 7 is obtained. Since the valley depths 30 to 89 correspond to the threshold value, the representative value 60 is set as the threshold value of the valley depth of 0 saturation.

As another method for determining the threshold value of the valley depth from the histogram of the valley depth of saturation 0, the above-mentioned paper, “Binarization Method of Drawing Utilizing Shape of Luminance Distribution”, is used. It is also possible to adopt the described method and the like.

The edge extraction unit 409 selects only the valley points having a valley depth equal to or larger than the threshold value calculated by the saturation 0 threshold value calculation unit 406 among the valley points of the saturation 0 as edge candidate. It is extracted as a point and its coordinate data is output. In the case of the scanning line 12 in FIG.
Among them, only three points c, d, and e are the edge candidate points. The depth of the valley is the larger of the height to the left mountain and the height to the right mountain of the valley point.

[Operation Example 23] In this operation example, the smoothing section 401, the saturation 1 calculation section 404, and the saturation 1 threshold value calculation section 4
07, only the edge extractor 409 is involved.

The smoothing unit 401 individually performs a mask process on each of the RGB data input from the color image input device 100 and outputs the smoothed RGB data. The size of the mask used for this mask processing is selected so that the valley points of saturation 1 other than the edge of the paper piece become small. Here, a mask having a width of 39 dots is used.

The saturation 1 calculation unit 404 uses the RGB data smoothed by the smoothing unit 401 to calculate
To calculate the saturation 1 and output the data. In the case of the scanning line 12 in FIG. 6, the saturation 1 is as shown in the lower part of FIG.

In the saturation 1 threshold value calculation unit 407, the saturation 1
In the same manner as in the saturation 0 threshold value calculation unit 406, a histogram of the valley depth is created, and the threshold value of the valley depth of saturation 1 is obtained. In the case of the scanning line 12, since the highest valley appears in the range of 10 to 19 in the histogram of the saturation 1, the representative value 15 is set as the threshold value.

The edge extraction unit 409 selects only the valley points having a valley depth equal to or larger than the threshold value calculated by the saturation 1 threshold value calculation unit 407 from among the valley points of the saturation 1 as edge candidates. It is extracted as a point and its coordinate data is output.

In the case of the scanning line 12, the valley points of the saturation 1 are four points a, o, d, and g. The depth of the valley at point a is max
Since ((22-0), (95-0)) = 95> 15, point a is an edge candidate point. Similarly, point g is also an edge candidate point. However, at point d, the depth of the valley is max
Since ((95-89), (93-89)) = 6 and less than 15, it is not extracted as an edge candidate point. The point o is not similarly extracted. The points a and g are the true edge points c and e
Although it is a point shifted slightly outward (in FIG. 6, the shift is emphasized more than actually), even if it is regarded as an edge point for the purpose of skew detection / correction, etc., there is usually no particular problem. Absent.

[Operation Example 24] In this operation example, the smoothing section 401, the saturation 1 calculation section 407, and the saturation 1 threshold value calculation section 4
07, only the edge extractor 409 is involved.

The operations up to the smoothing process, the calculation of the saturation 1 and the determination of the threshold value of the valley depth of the saturation 1 are the same as those of the twenty-third embodiment.

The edge extraction unit 409 performs the first step
As in the operation example 23, among the valley points of the saturation 1, the saturation 1
A valley point having a valley depth greater than or equal to the threshold value calculated by the threshold value calculation unit 407 is extracted as an edge candidate point. That is,
In the case of the scanning line 12 in FIG. 6, two points a and g are extracted.

The edge extraction unit 409 sets a point shifted by a predetermined distance toward a higher one of the mountains on both sides of the edge candidate point extracted in the first stage as a final edge candidate point. And output its coordinates. The above predetermined distance,
Here, the value is set to a half value of the size of the mask used in the mask processing of the smoothing unit 401, that is, 19 dots.

Therefore, in the case of the scanning line 12 shown in FIG. 6, two points a and g are extracted in the first stage, but the point a is moved to the right (inward) by 19 dots (89, 83).
9) The point (674, 839) obtained by moving the point g to the left (inward) by 19 dots is the final edge candidate point. By such a coordinate correction, the operation example 23 is performed.
, The edge candidate point and the true edge point c (100, 83
9), position errors with respect to e (655, 839) can be reduced.

[Operation Example 25] In this operation example, the smoothing section 401, the saturation 1 calculation section 404, and the saturation 1 threshold value calculation section 4
07, only the edge extractor 409 is involved.

In this operation example, in the edge extraction unit 409, as in the operation example 24, the valley points of the saturation 1 are equal to or larger than the threshold value calculated by the saturation 1 threshold value calculation unit 407. A valley point having a valley depth is extracted (first stage), and the coordinates of the extracted valley point are moved to the higher mountain side on both sides thereof and corrected (second stage). However, unlike the case of the operation example 24, the moving amount (offset) of the valley point coordinates is not fixed, but is adaptively set by the following equation (7) according to the height to the peaks on both sides.

Offset = min (LH, RH) · M / max (LH, RH) (7) where LH is the height to the mountain on the left side of the valley point, and RH is the height to the mountain on the right side of the valley point. , M are the sizes of the masks used for the mask processing of the smoothing unit 401.

In the case of the scanning line 12 in FIG. 6, two points a and g are extracted in the first stage. The offset for point a is 9 and the offset for point g is 10. Therefore, a point (79, 839) obtained by moving point a by 9 dots to the right and a point (683, 839) obtained by moving point g by 10 dots to the left become edge candidate points.

[Operation Example 26] In this operation example, the saturation 0 calculation section 403, the saturation 0 threshold value calculation section 406, and the smoothing section 4
01, saturation 1 calculator 404, saturation 1 threshold calculator 40
7. The edge extraction unit 409 is used.

As in the twenty-second embodiment, the saturation 0 calculating section 4
03 to obtain saturation 0, and a saturation 0 threshold value calculation unit 406
Find the threshold with. Further, similarly to the embodiment 23,
Smoothing is performed by the smoothing unit 401, and the saturation 1 calculation unit 404 is performed.
, And a saturation 1 threshold value calculation unit 407 determines a threshold value.

In the edge extracting unit 409, a valley point having a valley depth equal to or larger than the threshold value of the valley depth of the saturation 0 is selected from among the valley points of the saturation 0 (c, c, 3 of d and e
Points), and a valley point having a valley depth equal to or greater than the threshold value of the valley depth of the saturation 1 (a, g 2 points). Next, for each of the detected valley points of saturation 1, the closest valley point is selected from the detected valley points of saturation 0, and the coordinates are output as edge candidate points.

In the case of the scanning line 12 in FIG.
The valley point of saturation 0 closest to the point is point c, and the valley point of saturation 0 closest to point g of saturation 1 is point e. Therefore, the coordinates of point c and point e (these are true edge points) are obtained as the coordinates of the edge candidate points.

[Operation Example 27] In this operation example, the saturation 0 calculation section 403, the saturation 0 threshold value calculation section 406, and the smoothing section 4
01, saturation 1 calculator 404, saturation 1 threshold calculator 40
7. The edge extraction unit 409 is used.

This operation example is the same as the operation example 26 except for a part of the operation of the edge extraction unit 409. This difference is as follows.

In the edge extraction unit 409, as in the operation example 26, the saturation 1 having a valley depth equal to or greater than the threshold value is set.
(The two points a and g in the case of the scanning line 12) and the valley point of saturation 0 (the three points c, d and e in the case of the scanning line 12)
Is detected. Next, in the operation example 26, the valley point of saturation 0 closest to the valley point of saturation 1 is set as an edge candidate point. However, in this operation example, among the left and right peaks of the valley point of saturation 1, The valley point of saturation 0 closest in the direction of the peak is extracted as an edge candidate point.

In the case of the scanning line 12 in FIG. 6, since the right peak is higher at the valley point a of the saturation 1, the nearest valley c of the saturation 0 (true edge point) in the right direction is an edge candidate. Points. Also, at the valley point g of saturation 1, since the ridge on the left side is higher, the hill point e (true edge point) of saturation 0 closest to the left direction is an edge candidate point.

[Operation Example 28] In this operation example, the saturation 0 calculation section 403, the saturation 0 threshold value calculation section 406, and the smoothing section 4
01, saturation 1 calculator 404, saturation 1 threshold calculator 40
7. The edge extraction unit 409 is used.

In the same manner as in the operation example 26, the threshold values of the valley depths of saturation 0 and saturation 1 and saturation 0 and saturation 1 are obtained.

[0131] The edge extraction unit 409 performs the following steps:
A saturation 0 valley point with a valley depth greater than or equal to the saturation 0 valley depth threshold, and a saturation with a valley depth greater than or equal to the saturation 1 valley depth threshold The valley point of 1 is detected. Scan line 1 in FIG.
In the case of 2, the threshold value of the depth of the valley having a saturation of 0 is 60, and the saturation is 1
Since the threshold value of the depth of the valley becomes 15, three points c, d, and e are detected as the valley points of saturation 0, and two points a and g are detected as the valley points of saturation 1. You.

The edge extraction unit 409 performs the following steps:
A valley point as an edge candidate point is extracted from the valley points detected in the first stage, and its coordinates are output. In other words, from the valley point of saturation 1 to the higher side of the peak points on both sides,
The nearest valley point of saturation 0 higher than the valley point of saturation 1 is selected, and this valley point of saturation 0 is set as an edge candidate point. When the height of the valley point of saturation 1 is B1 and the height of the valley point of saturation 0 is B0, when the condition of B1-B0 <judgment threshold (here, 5) (8) is satisfied The valley point of the saturation 0 is the saturation 1
Is determined to be higher than the valley point.

In the case of the scanning line 12 in FIG. 6, the valley point a of the saturation 1 is higher in the right-hand mountain, so that attention is paid to the nearest valley point c of the saturation 0 in the right direction. -5 =
Since -5 <5, point c is extracted as an edge candidate point corresponding to point a. Since the valley point g of the saturation 1 is higher on the left mountain, the valley point e of the nearest saturation 0 in the left direction is B1-B0 = 0-1 = -1 <5.
The valley point e is extracted as an edge candidate point.

[Operation Example 29] In this operation example, the saturation 0 calculation section 403, the saturation 0 threshold value calculation section 406, and the smoothing section 4
01, saturation 1 calculator 404, saturation 1 threshold calculator 40
7. The edge extraction unit 409 is used.

In the same manner as in the operation example 26, the threshold values of the valley depths of saturation 0 and saturation 1 and saturation 0 and saturation 1 are obtained.

The edge extraction unit 409 performs the following steps:
A saturation 0 valley point with a valley depth greater than or equal to the saturation 0 valley depth threshold, and a saturation with a valley depth greater than or equal to the saturation 1 valley depth threshold The valley point of 1 is detected. Scan line 1 in FIG.
In the case of 2, the threshold value of the depth of the valley having a saturation of 0 is 60, and the saturation is 1
Since the threshold value of the depth of the valley becomes 15, three points c, d, and e are detected as the valley points of saturation 0, and two points a and g are detected as the valley points of saturation 1. You.

The edge extraction unit 409 performs the following steps:
A valley point as an edge candidate point is extracted from the valley points detected in the first stage, and its coordinates are output. That is, from the valley point of saturation 1 to the valley point of saturation 0 determined to be higher than the valley point of saturation 1,
An edge candidate point whose distance from the focused valley point of saturation 1 is smaller than a predetermined threshold value (here, 50) is determined. The determination that the valley point of saturation 0 is higher than the valley point of saturation 1 is based on the above equation (8).

In the case of the scanning line 12 in FIG. 6, there are two c and e valley points to the right of point a, which are judged to be higher than the point a. Point c is 30 distance from point a,
Although smaller than the threshold value of 50, the point e is at a distance of 585 from the point a and is larger than the threshold value of 50. Therefore, only point c is extracted as an edge candidate point corresponding to point a. Similarly, point e on the left side is extracted as an edge candidate point corresponding to point g.

[Operation Example 30] In this operation example, all the functional units of the edge detection device 400 are used.

In the same manner as in the operation example 26, the threshold value of the valley depth of the saturation 0 and the saturation 1 is obtained.

The smoothing section 402 smoothes the input RGB data. In this R, G, B mask processing for smoothing, a mask smaller than the mask size used in the smoothing unit 401 is used. Here, it is assumed that a mask having a width of 19 dots is used. The saturation 2 calculator 405 calculates the saturation 2 from the smoothed RGB data. Scan line 1 in FIG.
The saturation 2 of 2 is as shown in the third row of FIG. Saturation 2
The threshold value calculation unit 408 includes a saturation 0 threshold value calculation unit 406
Alternatively, a threshold value of the depth of the valley of saturation 2 is calculated based on the saturation 2 data by a method similar to that of the saturation 1 threshold value calculation unit 407. For scan line 6, this threshold is 1
It becomes 5.

The edge extraction unit 409 performs a first step
A saturation 0 valley point with a valley depth greater than or equal to the saturation 0 valley depth threshold, and a saturation 1 with a valley depth greater than or equal to the saturation 1 valley depth threshold. And a valley point of saturation 2 having a valley depth greater than or equal to the threshold value of the valley depth of saturation 2 is detected. In the case of the scanning line 12 in FIG.
Points, two points a and g are detected as valley points of saturation 1, and b, d, and f are detected as valley points of saturation 2.

The edge extraction unit 409 performs the following steps:
For each valley point of saturation 1, if there is a valley point of saturation 2 that is determined to be higher than the valley point within a range of 50 dots on the left and right sides, it is determined that a corresponding edge exists. When the saturation of the valley point of saturation 1 of interest is B1, the saturation of the valley point of saturation 2 is B2, and the condition of B1-B2 <predetermined threshold value (here, 5) (9) is satisfied. The valley point of the saturation 2 is the saturation 1
Is determined to be higher than the valley point.

The valley point a of saturation 1 has a valley point b of saturation 2 which is 18 dots away, and the condition of equation (9) is satisfied because 0-0 = 0 <5. Therefore, it is determined that a corresponding edge exists at the valley points a and b. Similarly, it is determined that the corresponding edge exists also in the valley points g and f.

The edge extraction unit 409 performs the third step
Attention is paid to the valley point of saturation 2 determined to have a corresponding edge in the second stage, and if the valley point of saturation 0 is higher than the valley point of saturation 2 among the valley points of saturation 0 in the high direction of the left and right mountains. The closest one determined is extracted as an edge candidate point, and its coordinates are output. The height of the focused valley point of saturation 2 is B2, and the saturation is 0
When the height of the valley point of B is defined as B1, when the following condition is satisfied: B2−B0 <predetermined threshold value (here, 5) (10)
Is determined to be higher than the valley point.

Saturation 2 determined to have a corresponding edge
In the case of the valley point b, since the mountain on the right side is higher, the valley point of the saturation 0 in the right direction is examined. Then, c
Since the point and the point e are higher than the point b, the closest point c is extracted as an edge candidate point. Similarly, a point e is extracted as a candidate edge point by performing the processing on the valley point f of the saturation 2.

[Operation Example 31] In this operation example, all the functional units of the edge detection device 400 are used.

In the same manner as in the operation example 30, threshold values of the valley depths of saturation 0, saturation 1, and saturation 2 are obtained.

In the edge extraction unit 409, in the first stage, the valley points of the saturation 0, the saturation 1, and the saturation 2 where the depth of the valley is equal to or larger than the threshold value are determined in the same manner as in the first stage of the operation example 30. To detect. In the case of the scanning line 12 in FIG. 6, three valley points of saturation 0 are c, d, and e, and two valley points of saturation 1 are a and g.
B, d, and f are detected as the valley points of the point and the saturation 2.

In the second stage, it is determined that the valley point of saturation 1 is higher than the valley point of saturation 1 within a range of 50 dots on the left and right (this determination method is the same as in the operation example 30). If a valley point of degree 2 exists, it is determined that a corresponding edge exists. In the case of the scanning line 12, it is determined that there are edges corresponding to points a and b and points g and f.

In the third stage, among the valley points of saturation 2 for which it is determined that there is a corresponding edge, among the valley points of saturation 0 in the opposite direction to the corresponding valley point of saturation 1, It is determined that the valley point is higher than the valley point of degree 2.
The closest one (same as 0) is extracted as an edge candidate point, and its coordinates are output. In the scanning line 12 in FIG. 6, the valley point of saturation 1 corresponding to the point b is the point a on the left side. Upon examination, it is determined that points c and e are higher than point b.
As a result, the closest point c is an edge candidate point. By performing the same processing for point f, point e is extracted as an edge candidate point.

[0152]

As described above in detail, according to the present invention, it is possible to detect the edge of a paper piece having the same or different background color as the background color at the time of reading from color image data. For example, in a color copying machine, it is possible to achieve effects such as improving the quality of a copied document by correcting the shadow and skew at the edge of the document.

[Brief description of the drawings]

FIG. 1 shows a block configuration of a first embodiment of the present invention.

FIG. 2 shows an example of a color image for explaining Embodiment 1 and luminance and chroma in one scanning line thereof.

FIG. 3 shows an example of a histogram of luminance valley depths.

FIG. 4 shows an example of a histogram of saturation valley depths.

FIG. 5 shows a block configuration of a second embodiment of the present invention.

FIG. 6 shows an example of a color image for explaining Embodiment 2 and saturations 0, 1, and 2 in one scanning line thereof.

FIG. 7 shows an example of a histogram of the depth of a valley having a saturation of 0;

[Explanation of symbols]

 REFERENCE SIGNS LIST 100 color image input device 200 edge detection device 201 luminance calculation unit 202 saturation calculation unit 203 R chromaticity calculation unit 204 G chromaticity calculation unit 205 B chromaticity calculation unit 206 threshold calculation unit for luminance valley point extraction 207 luminance valley Point extraction unit 208 Saturation threshold value calculation unit 209 R chromaticity threshold value calculation unit 210 G chromaticity threshold value calculation unit 211 B chromaticity threshold value calculation unit 212 Edge extraction unit 300 Color image processing device 400 Edge detection Apparatus 401 Smoothing unit (1) 402 Smoothing unit (2) 403 Saturation 0 calculation unit 404 Saturation 1 calculation unit 405 Saturation 2 calculation unit 406 Saturation 0 threshold value calculation unit 407 Saturation 1 threshold value calculation Section 408 saturation 2 threshold value calculation section 409 edge extraction section

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06T ⁷ /00-7/60 G06T 1/00 H04N 1/04 H04N 1/46

Claims (21)

(57) [Claims] 1. A color obtained by reading a piece of paper
Calculate the brightness from the image data, and find the edge
Those satisfying the supplementary point extraction conditions are considered as edge candidate points of the paper chip.
Edge extraction method, which calculates saturation or chromaticity from color image data and
The valley point of luminance is the peak of saturation or chromaticity in the complement point extraction condition.
Or not a valley point.
Extraction method. 2. A color obtained by reading a piece of paper.
Calculate the brightness from the image data, and find the edge
Those satisfying the supplementary point extraction conditions are considered as edge candidate points of the paper chip.
Edge extraction method, which calculates saturation or chromaticity from color image data and
The valley point of luminance is higher than a certain threshold
Not a peak of saturation or chromaticity, or
Values that are not deeper than the saturation or chromaticity valley
An edge extraction method characterized in that 3. An edge extracting method according to claim 2, wherein
The height of the peaks or valleys of saturation or chromaticity
Set the threshold to the peak height or valley depth of saturation or chromaticity
Edge extraction method characterized in that it is determined from the distribution of the edge. 4. A color obtained by reading a sheet of paper.
Calculate the brightness from the image data, and find the edge
Those satisfying the supplementary point extraction conditions are considered as edge candidate points of the paper chip.
Edge extraction method, which calculates saturation or chromaticity from color image data and
Complement point extraction conditions include the saturation or chromaticity slope at the luminance valley point.
Is smaller than a certain threshold value.
Edge extraction method. 5. An edge extracting method according to claim 4, wherein
The saturation or chromaticity slope threshold to
Edge extraction characterized by being determined from the gradient distribution of degrees
How to get out. 6. The method according to claim 1, wherein
In the edge extraction method, the edge candidate
The difference between the peak heights on both sides of the degree valley point is smaller than a certain threshold
An edge extraction method characterized by including: 7. The method according to claim 1, wherein
In the edge extraction method, the edge candidate point extraction conditions include luminance
That the width of the valley is less than a certain threshold
Edge extraction method to be featured. 8. The method according to claim 1, wherein
In the edge extraction method, the edge candidate point extraction conditions include luminance
Valley depth greater than a certain threshold
And an edge extracting method. 9. An edge extracting method according to claim 8, wherein
The threshold value of the luminance valley depth
An edge extraction method characterized by being further determined. 10. A color obtained by reading a sheet of paper.
-Calculate the saturation from the image data.
An edge extraction method for extracting as a candidate edge point, wherein both of the saturation valley points extracted as the edge candidate point are determined.
A certain offset towards the higher peak of the side saturation peak
Extract points that have moved by minutes as edge candidate points.
Edge extraction method to be featured. 11. The edge extracting method according to claim 10, wherein
And the depth of the saturation valley point is greater than a certain threshold
A valley point of saturation is extracted as a candidate point for an edge.
An edge extraction method characterized by: 12. An edge extracting method according to claim 11, wherein
And the threshold value of the saturation valley point depth is
Edge extraction method characterized by determining from distribution of height
Law. 13. The method according to claim 10, wherein:
In the edge extraction method described above, color image data is smoothed
After the smoothing process, the color image data
An edge extraction method characterized by determining a degree. 14. A saturation valley point extracted as an edge candidate point by the edge extraction method according to claim 13 in the direction of a higher peak of the saturation peaks on both sides thereof. An edge extraction method characterized in that a point moved by an offset determined from a mountain height and a size of a mask used for smoothing processing is extracted as an edge candidate point. 15. Saturation before performing smoothing processing and saturation after performing smoothing processing (for the distinction, the former saturation is defined as 0 saturation from the color image data obtained by reading the paper piece). , The latter saturation is referred to as saturation 1), and a saturation 0 valley point having a valley depth greater than or equal to a certain threshold, and a saturation having a valley depth greater than or equal to a certain threshold. An edge extraction method characterized by extracting a valley point having a saturation of 0 closest to a valley point of a saturation of 1 as a candidate point of an edge of a sheet of paper. 16. Saturation before smoothing processing and saturation after smoothing processing are performed based on color image data obtained by reading a sheet of paper. , The latter saturation is referred to as saturation 1), and a saturation 0 valley point having a valley depth greater than or equal to a certain threshold, and a saturation having a valley depth greater than or equal to a certain threshold. Regarding the valley point of 1, the valley point of saturation 0 closest to the valley point of saturation 1 is extracted as an edge candidate point of the sheet of paper in the direction of the highest ridge on both sides of the valley point of saturation 1. An edge extraction method characterized in that : 17. Saturation before smoothing processing and saturation after smoothing processing (for the sake of distinction, the former saturation is defined as 0 saturation from color image data obtained by reading a piece of paper). , The latter saturation is referred to as saturation 1), and a saturation 0 valley point having a valley depth greater than or equal to a certain threshold, and a saturation having a valley depth greater than or equal to a certain threshold. With respect to the valley point of 1, the valley of the saturation 0 that is higher than the valley point of the saturation 1 and is closest to the valley point of the saturation 1 in the direction of the higher ridge on both sides of the valley point of the saturation 1 this to extract a point as the edge candidate points of a piece of paper
And an edge extracting method. 18. Saturation before performing smoothing processing and saturation after performing smoothing processing (for the sake of distinction, the former saturation is defined as 0 saturation from color image data obtained by reading a sheet of paper). , The latter saturation is referred to as saturation 1), and a saturation 0 valley point having a valley depth greater than or equal to a certain threshold, and a saturation having a valley depth greater than or equal to a certain threshold. With respect to the valley point of 1, a color higher than the valley point of the saturation 1 within a certain distance range from the valley point of the saturation 1 in the direction of the highest peak on both sides of the valley point of the saturation 1 An edge extraction method characterized by extracting a valley point having a degree of 0 as an edge candidate point of a paper piece. 19. Saturation before performing a smoothing process, saturation after performing a smoothing process using a first size mask, based on color image data obtained by reading a sheet of paper,
And the saturation after smoothing using a mask of a second size smaller than the first mask size (the first saturation is saturation 0, and the second saturation is saturation 1, 3 The second saturation is referred to as saturation 2), and a saturation 0 valley point having a valley depth greater than or equal to a certain threshold, and a saturation 1 valley having a valley depth greater than or equal to a certain threshold value. A point and a valley point of saturation 2 having a valley depth greater than or equal to a certain threshold value are determined to be higher than the valley point of saturation 1 within a certain distance range of the valley point of saturation 1. When a valley point of saturation 2 exists, the valley point of saturation 2 is determined to be a valley point having an edge corresponding to the valley point of saturation 1 and the valley point of saturation 2, and it is determined that an edge exists. Among the peaks on both sides of the point, in the direction of the higher peak, the valley point of saturation 0 determined to be higher than the valley point of saturation 2, the one closest to the valley point of saturation 2 is defined as the edge of the sheet of paper. Candidate points and Be characterized in that the extraction Te
Edge extraction method that. 20. Saturation before performing a smoothing process, saturation after performing a smoothing process using a first size mask, based on color image data obtained by reading a piece of paper,
And the saturation after smoothing using a mask of a second size smaller than the first mask size (the first saturation is saturation 0, and the second saturation is saturation 1, 3 The second saturation is referred to as saturation 2), and a saturation 0 valley point having a valley depth greater than or equal to a certain threshold, and a saturation 1 valley having a valley depth greater than or equal to a certain threshold value. A point and a valley point of saturation 2 having a valley depth greater than or equal to a certain threshold value are determined to be higher than the valley point of saturation 1 within a certain distance range of the valley point of saturation 1. When a valley point of saturation 2 exists, it is determined that there is an edge corresponding to the valley point of saturation 1 and the valley point of saturation 2, and from the valley point of saturation 2 that is determined to have an edge, in the opposite direction to the trough point of saturation 1 corresponding thereto, extracting the valley point of closest saturation 0 is judged to be higher than the valley point of該彩degree 2 as the edge candidate points paper Edge extraction method characterized. 21. The edge extraction method according to claim 15, wherein the threshold value of the depth of the saturation valley point is determined from the distribution of the saturation valley point depth. An edge extraction method characterized by the following.