JPH05181966A - Image output method - Google Patents

Abstract

PURPOSE:To obtain an image output method which has small deterioration in picture quality even when an image is repeatedly copied. CONSTITUTION:Image data are represented as specific density, contour restoration information regarding the contour of an image which is previously set are represented as density different from the density of the image data, and the contour restoration information is added to the image data and outputted. When a hard copy of the output data is taken, the contour of the image can be restored with the contour restoration information. In one way, the image data are represented as plural density bands Wa and Wb, the contour restoration information regarding the image of the image which is previously set is represented as plural density bands Wa and Wb different from the density of the image data, and the contour restoration information is added to the image data and outputted. In another, the image density is varied avoiding the density of the contour restoration information when the image density is consecutive.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital image output device such as a digital printer.

[0002]

2. Description of the Related Art In a digital printer or the like, digital image data is read by an image input section, and the image processing section corrects the input image data according to the printing characteristics of the printing section or an edit mode designated by an operating section. Edit the image corresponding to. After processing this image data,
A hard copy is made on paper at the printing section by electrophotography.

[0003]

The image input section reads an image at a predetermined resolution. Therefore, detailed information on the original that exceeds the resolution is lost during reading. Further, even in the printing section, the output image is not faithfully hard-copied. Therefore, in an image input device that digitally reads an image, image deterioration occurs when a hard copy of image data is taken. Degradation of an image does not pose a problem even if the original is copied once. But,
In the case of hard copy, the hard copy may be used as a document and further copying may be repeated, and the image quality may be significantly deteriorated. There is no problem if you copy it as it is with a digital source such as a floppy disk of a word processor, but due to compatibility problems and ease of use, hard copies are often copied. Therefore, there is a need to improve hard copy quality.

In Japanese Patent Laid-Open No. 2-112966, the original shape is estimated and smoothing is performed to improve the contour line. However, this method does not correctly restore the contour deformation at the time of hard copy or reading.

An object of the present invention is to provide an image output method in which the deterioration of image quality is small even if copying is repeated.

[0006]

According to a first image output method of the present invention, image data is represented by a predetermined density, and contour restoration information relating to the contour of a preset image is different from the density of the image data. The feature is that the contour restoration information is added to the image data and output, and when the hard copy of this image data is copied, the contour of the image can be restored from the contour restoration information. According to a second image output method of the present invention, image data is represented by a plurality of predetermined density bands, and contour restoration information regarding a preset image contour is represented by a plurality of density bands different from the density of image data. It is characterized in that contour restoration information is added to image data and output, and when a hard copy of this image data is copied, the contour of the image can be restored from the contour restoration information. In a third image output method according to the present invention, image data is represented by a plurality of predetermined density bands, and contour restoration information regarding a preset image contour is represented by a plurality of density bands different from the density of the image data, When the determination means determines that the change in the image density is continuous near the density band of the contour restoration information, the contour restoration of the density band is performed. It is characterized in that it is not assigned to information but to image data, and when a hard copy of this image data is copied, the contour of the image can be restored from the contour restoration information.

[0007]

When the copying is repeated, the image quality deteriorates because the shape of the figure cannot be read correctly.
It is due to not being able to write correctly. However, as long as reading is performed at a predetermined resolution, overwriting occurs in continuous graphics. Therefore, digital information (contour restoration information) representing the shape of the graphic is added to the analog graphic of the original. This digital information is information relating to the contour of a graphic that is set in advance. In the first image output method according to the present invention, the predetermined density in the output image is used as the contour restoration information of the image (see FIGS. 4 and 5). When the contour restoration information is added to a part of the original document by changing the density, when the halftone is included, it is necessary to add it so that the influence on the density of each halftone is minimized. Therefore, in the second image output method according to the present invention, the contour restoration information is represented by a plurality of density bands different from the image data (see FIG. 32).
By expressing the contour restoration information with an appropriate density, the influence on the halftone image can be reduced. Further, preferably, when there is image data corresponding to the density band assigned to the contour restoration information, the image data is converted to a density different from that density band (see FIG. 35). Further, in the third image output method according to the present invention, the image data is represented by a plurality of predetermined density bands, and the contour restoration information regarding the preset image contour is represented by a plurality of density bands different from the density of the image data. The determination means for determining the continuity of the change in the image density is provided, and when the determiner determines that the change in the image density is continuous near the density band of the contour restoration information, the density band It is characterized in that it is assigned to the image data, not to the contour restoration information, and the contour of the image is restored from the contour restoration information (see FIG. 34). At the time of writing, digital information (contour restoration information) is added to the analog data graphic as described above to create a hard copy. At this time, even if a faithful copy cannot be obtained, correction can be performed after reading as long as there is no loss of digital information. In this way, even if the hard copy is used as the original and the copy is repeated, the image quality does not deteriorate. At the time of reading, both analog (image data) and digital (contour restoration information) information is read, digital information is extracted by the extraction means, analog image reading errors are corrected using the digital information, and image data is restored. To do. Analog image data is unnecessary for restoration.

[0008]

Embodiments of the present invention will be described below in the following order with reference to the drawings. (A) Addition of feature points and contour restoration A. First Embodiment (b) Configuration of Image Processing Unit (c) Image Analysis (d) Image Restoration (e) Feature Point Adding Unit B. Second embodiment (f) Feature point generation / addition during hard copy C. Third embodiment (g) Density reference pattern (h) Addition of feature points at low density in halftone image (i) Feature point extraction corresponding to continuity of density change (j) Density changes continuously Addition of feature points to halftone image (k) Addition of feature points when creating original document D. Fourth Embodiment (l) Error processing

(A) Addition of feature points and contour restoration In the present invention, contour restoration information, which is digital information, is embedded in an analog figure of a document. Since the hard copy is viewed by the user, it is necessary to add digital information so as not to obstruct the original figure. Specifically, for example, as shown in FIG. 1, a point (feature point) representing the contour is added so as to be in contact with the contour of the original figure (or image). The feature points are expressed in different densities and colors than the original figure. The size of the characteristic points is inconspicuous by visual inspection of the hard copy, but should be such that there is no omission when reading the hard copy. FIG. 1 shows an example in which the density of the feature point 1 is made higher than that of the graphic 2. Figure 2 is the Roman letter "a",
It can be seen that the feature point 1 is added to the diagram on the right side, which is an enlarged view of a part thereof. FIG. 2 schematically shows the processing on the image of FIG. When the document shown in (a) is read,
The feature points are extracted as shown in (b). Next, a contour line is calculated so as to circumscribe this feature point as shown in (c). When outputting the restored image data to another device, both the restored image data (d) and the extracted feature points (b) are processed.

FIGS. 3A to 3D show an example of the feature point 1. The feature point 1 is arranged at a position inscribed in the contour of the figure 2. Further, the number of characteristic points indicates the type of contour curve.
Here, 3 represents a restored contour. In the examples of FIGS. 1 and 2, the shape of the feature point 1 is a quadrangle, but in the following, it is represented by a circle for simplicity. Basically, as shown in FIG.
Place one feature point at the corner of. (B)-(d)
Shows an application example for performing the irregular processing described later. In the present embodiment, if there are three feature points within a predetermined minute distance, it is indicated that the figure is to be subjected to irregular processing. As a result, the number of feature points can be reduced. (B) ~
In the case of (d), it is clear that some feature points do not exist at the edge of the contour of the image. In the case of (b), there are three special feature points 1 ′ that form a right angle in close proximity to the convex corner, but the restored figure 3 after the irregular processing includes these feature points 1 ′ inside. It is a sharp-angled figure. (C)
In the case of, the inside includes a special feature point 1 ′ that forms three adjacent straight lines, and the deformed contour is obtained by connecting the front and rear points with a circle. In the case of this processing, there is no effect when the curve is approximated by an arc. In the case of (d),
Similar to the case of (b), there are three special feature points 1'that form a right angle at the corner, but (b)
It exists in the concave corner opposite to. In this case, the deformed contour is a curve that smoothly connects the front and rear feature points 1.
As in the case of (b), the number of feature points can be reduced.

Here, the density of the feature point 1 is set higher than the density of the image data. FIG. 4 shows the determination output in the case of a multi-valued original, and the density of the characteristic point 1 is set to Ia or more. Therefore, when the read document density is equal to or higher than Ia, it is determined to be a feature point. However, when the density of the original image is low, the characteristic point 1 may appear as noise on the image, and at this time, the characteristic point cannot be added. FIG. 5 shows a determination output in the case of a binary document including only characters and figures, and is determined as a feature point when the read document density is Ia or more.

First Embodiment (b) Structure of Image Processing Section In a digital copying machine, after image data is read by an image input section such as an image reading device, various processing of the image data is performed in the image processing section, In a printing unit such as an electrophotographic printer, a hard copy on paper is output. In the image processing unit, correction is performed according to the printing characteristics of the printing unit, and image editing is performed according to the editing mode designated by the operation unit. FIG. 6 is a block diagram of the image processing unit according to the first embodiment of the present invention. In this image processing unit, the image analysis unit 11 performs feature point extraction and region determination processing on digital image data input from the image input unit via the bus B1. That is, the characteristic point 1 is extracted from the input image data, and the extracted characteristic point 1 is used for the area discrimination. That is, the input image data is separated into a restorable area that includes the feature points and a non-restorable area that does not include the feature points. In the non-restorable area, output the input image data as it is,
Used in later processing. In the restorable area, the input image data is not used, and the image restoration unit 13 restores the image data using the extracted feature points 1 and outputs it. These image data are combined by the feature point addition unit 14, and when the processing mode is set from the operation unit 16, the extracted feature points are added. If multiple density of feature points can be set,
The feature point with the density closest to the density near the feature point is added.
If necessary, the editing unit 12 independently edits (rotates or enlarges) the restored data and the data in the unrestorable area. Here, the editing is performed for both data in order to match the positions of both data. Then, the edited data is combined in the feature point addition unit 14. The feature point 1 can be output without being added depending on the processing mode set by the operation unit 16. Finally, the optimizing unit 15 corrects the image density by smoothing processing or the like according to the processing mode input from the operating unit 16 and outputs the corrected image density to the bus B2.

(C) Image Analysis FIG. 7 is a block diagram of the image analysis unit 11 which performs area discrimination and feature point extraction. The image analysis unit 11 includes an image memory 111 that stores input digital image data, a feature point memory 122 that stores position data of feature points extracted by comparison with density levels, and an image analysis processor 123 that performs image analysis. , A program memory 128 for storing a program executed by the image analysis processor, a block management memory 124 for storing management data of blocks obtained by checking a continuous area of an image, a working memory 127 used for image analysis, and a feature point. It is composed of an interface 126 for outputting the image data of the non-restored area not included to the bus B12 and an interface 125 for outputting the coordinates of the extracted feature points to the bus B11. Image processor 123
Starts processing when data is loaded into the image memory 111, examines a continuous region of an image, forms a block, extracts a feature point, and determines an image block that does not include the feature point. The block management data is stored in the block management memory 124, and the extracted position data of the feature points is
It is stored in the feature point memory 122. When the predetermined process is completed, the two interfaces 125 and 126 are respectively provided.
To output the processing result.

FIG. 8 shows a processing flow of the image processor 123. When the power is turned on (S100), first, initialization is performed (S101). Next, waiting for the input of image data (S102), if there is an image input,
Blocking process (S103), process for searching feature points (S1)
04), vectorization processing of peripheral data (S105), vectorization processing of irregular points (S106), and output processing (S107). Then, the process returns to S102 and the above processing is repeated.

(C-1) Blocking The image processor 123 examines continuous areas stored in the image memory 111 and creates blocks (see the flow in FIG. 11). Since the blocking process is vulnerable to noise, it is desirable to perform it after the smoothing process. The block information can also be used in a block identification process for identifying a non-restored image (see FIG. 31). The condition of the non-restoring block is
The attribute is slope, and the feature point flag is not set (present) in the block. For blocks satisfying this condition, the original image data in the image memory 111 is used instead of the restored image. In blocking, first, the image data stored in the image memory 111 is scanned line by line. Then, when there is a change of a predetermined density or more in a predetermined minute section during scanning, it is considered as an edge of the block. The start and end points of the line are also considered as edges. In the example shown in FIG. 9, the result of blocking is shown in (a). Here, both boundaries of the block are indicated by "<" and ">". Further, "a", "b", etc. represent the names of the restoration identification numbers given in order from the beginning of the line. The restoration identification number will be described later. Then, the maximum value, the minimum value, and the average value are obtained for the density data of one block from one edge to the next edge. Next, an attribute is given to the block (section) from the maximum value and the minimum value. If the difference between the maximum value and the minimum value is within a predetermined range, it is set as a uniform density section (attribute: flat), and if not, it is set as a continuously changing section (attribute: slope).
The following information (that is, the restoration identification number,
The feature point flag (presence / absence), the start point, the end point, the average density, the maximum density, the minimum density, and the attribute (flat, slope) are stored in the block management memory 124. Here, the feature point flag indicates whether or not a feature point is included in the block, and has no initial value. When the determination of the section is completed for one line, the same section is connected and divided into blocks as compared with the line immediately before (the head side). Then, when all the following conditions are satisfied, they are regarded as the same block.
(1) The sections are adjacent to each other (there are overlapping sections between the start point and the end point of the two sections.) (2) The difference in average density is within a predetermined value. (3) The difference between the maximum density and the minimum density in the two sections is within a predetermined value. (4) The attributes are the same.
However, if there are slopes, (2) and (3) may be ignored. If they are not regarded as the same block, a unique restoration identification number is added as a new block. 1
All lines are regarded as new blocks. In the example of FIG. 9, all the first lines are new blocks.
Restoration identification numbers of a "," b "," c ", ... Are given. The second line is" a "," 1 ", respectively.
It is regarded as the same block as b "," c ", ... The same applies to the third line. The restoration identification number appearing on the fourth line"
The block of d "is regarded as the same block as" a "of the third line, and then the block of" c "in the first to third lines is regarded as the same block as" a ".

When the blocking is completed for all lines, feature points are removed from the image data.
(Actually, the start point or end point of the adjacent block is changed, and the area is combined.) A block with a flat attribute,
If the average in the block is within a predetermined range corresponding to the density of the feature points, it is considered as a feature point. The feature point is considered to be the same as the block having the longest contact length among other contacting blocks. If the lengths are the same, it is considered to be the same as the closer average density. The feature point flag of the block that fetches the feature point block is set to "Yes". The processed feature point block information is discarded. In the example of FIG. 10A, a block including a feature point (represented by "m") is found. Here, as shown in (b), the block including the feature point is regarded as the same as the block of “b”, and the starting point of the block of “b” is changed to combine the regions. At the same time when the feature points are removed, the feature point coordinates and the density at the time of restoration can be obtained. As the feature point coordinates, the center of the feature point block or a point on the outer circumference outside the absorbed block (the center of the outer circumference in the above figure) is used. Which point is adopted is predetermined for each system including addition of feature points.

FIG. 11 shows a flow of the blocking process (S103). In this flow, blocking, feature point extraction, and block identification are performed at the same time. First, for one line, the edge within the line is obtained (S12
1). Next, the maximum value and the minimum value in the block (section) obtained from the edge are obtained (S122), and the section attribute is determined (S123). Next, it is searched whether the previous line has the same section (S124), and if there is the same section (S125).
YES), copy the restoration identification number of the previous section (S1
26), the maximum value and the minimum value are updated (S127). If there is no same section (NO in S125), a new restoration identification number is given to that section (S128). Next, it is determined whether or not this processing has been completed for all sections in the line (S129). If not finished (N in S129,
O), the process returns to S124 to process the next section. If the processing of all the sections has been completed for one line (YES in S129), then it is determined whether the processing of all the lines has been completed (S130). If not completed, the process returns to S121 to process the next one line. If the processing for all lines has been completed, the blocking processing is completed (S131).

(C-2) Feature point extraction processing The feature point extraction processing performed by the image analysis unit 11 (FIG. 8, S10).
In 4), the feature points are extracted from the block management memory 124, vectorized in contour order in contour units, and the feature point memory 1
22 is stored. The information of each characteristic point is as follows. -Corresponding restoration identification number (block number that absorbed the feature point) -Coordinates (x, y) -Density at restoration-Regular flag (representing a point indicating irregular contour processing) -Previous feature Pointer indicating a point (for example, a counterclockwise point on the contour) -Pointer indicating a next feature point (for example, a clockwise point on the contour) The specific process (FIG. 12) is the following three. It is classified into steps. 1: Obtain feature point coordinates, peripheral density, and restoration identification number (block ID). At this time, the characteristic points of the pattern memory are deleted. 2: A point on (or near) the contour of the pattern is searched for along the contour and vectorized. 3: Find anomalous points inside the pattern and connect to the relevant contours.

FIG. 12 shows a feature point extraction process (S10 in FIG. 8).
The flow of 4) is shown. First, a block having a flat attribute is searched (S141), and an average density in the block is calculated (S142). Then, the average density is compared with a predetermined density to determine whether or not it is a feature point (S143). For example, the judgment output is obtained as shown in FIGS. 4 and 5. If it is a feature point (YES in S143), then the length of the tangent to the peripheral block is obtained (S144). Then, if there are a plurality of longest peripheral blocks (YES in S145), the average densities are compared and the closest one is selected (S146). Next, a block to be absorbed is determined (S147), a feature point area is added to the determined block, a feature point flag is set (S148), and feature point coordinates and density are obtained (S14).
9). Then, these pieces of information are stored in the feature point memory 187 together with the restoration identification number (S150), and the information of the block including the corresponding feature point is discarded (S151). Next, it is determined whether the processing of all blocks is completed (S15).
2) If not completed, the process returns to S141 to process the next flat block. Also, if it is determined in S143 that it is not a feature point, the process proceeds to S152. If the processing of all blocks is completed (S152), the feature point extraction processing is completed (S153).

(C-3) Peripheral Data Vectorization Processing FIG. 13 shows a flow of peripheral data vectorization (S105 in FIG. 8). First, the restoration identification number (I
D) is determined (S221). Next, one feature point near the contour is extracted (S222). If there is a feature point (S223
If YES, the contour starting point is set (S224). Next, the next feature point is searched for along the contour of the corresponding block (S2).
25). Here, it is assumed that the error from the contour is within a predetermined distance. If there is a feature point (YES in S226), the previous feature point is connected (S227), and the process returns to S225. On the other hand, if there is no feature point (NO in S226), the last feature point is connected to the start point (S228), the process returns to S222, and the processing of the next contour is continued. If it is determined in S223 that there is no feature point, it is then determined whether the processing of all restoration identification numbers has been completed (S230). If it is determined that the processing has not been completed, the processing returns to S221, and the next Process the restoration identification number. When it is determined that the process is finished (Y in S230
S), the vectorization process is terminated (S231).

(C-4) Vectorization of irregular points FIGS. 14 and 15 show vectorization of irregular points (S10 in FIG. 8).
The flow of 6) is shown. First, the restoration identification number (block ID) of the target block is determined (S161). Next, a vectorized feature point is searched (S162). Next, the connected feature points are searched (S163). If it is determined that the three feature points are within the predetermined minute distance (YE in S164).
S), the three points are the characteristic points to be subjected to the irregular processing, and the characteristic point flag is set (S165). Next, it is determined whether the processing for one round of the contour is completed (S166). If not completed, the process returns to S163, and the processing for the next feature point in the same contour is performed. If the processing has been completed, it is then determined whether or not the processing for all contours has ended (S167). If not completed, one block is composed of a plurality of contours, so the processing returns to S162 and the next Performs contour processing. When the processing of all contours is completed, it is next determined whether or not the processing of all restoration identification numbers has been completed (S168). If not finished, the process returns to S161 and the block of the next restoration identification number is processed. to continue. When the processing for all restoration identification numbers is completed, the above processing is completed.

Next, contour connection processing at irregular points is performed.
(Note that FIG. 16 shows an example of connection.) First, the restoration identification number (block ID) of the target block is determined (S16).
9). Next, an unconnected feature point (point a in FIG. 16) is searched for (S
170). Next, it is determined whether or not there are two unconnected feature points in the vicinity (S171). If not, the noise is removed as noise (S179), and the process returns to S170. If there are two (points b and c in FIG. 16), since they are irregular points, the connected feature points within a predetermined distance in the predetermined feature point search direction (lines d and e in FIG. 16) are searched for ( S172). And the feature point is 2
If there are individual points (points f and g in FIG. 16), irregular points are connected between the two points (S176) to form a temporary contour (h and i in FIG. 16). (Note that the finally restored contour is the broken line j.) Then, it is determined whether or not there are still unconnected points (S17).
7) If there is, return to S170 and continue the irregular point processing.
If there is no unconnected point, then it is determined whether the processing for all restoration identification numbers has been completed (S178). If not completed, the process returns to S170 to continue the process for the next restoration identification number, and if completed, the vectorization process is completed.
If it is determined in S173 that there are no two feature points, the angle is widened (S174), and it is determined whether there are two feature points (S175). If not, S17
In step 9, the noise is deleted, and if there is noise, in step S176, an irregular point is connected between the two points.

(C-5) Output Processing FIG. 17 shows the flow of the output processing (S106 in FIG. 8).
First, it is determined whether there is a non-restored data request for one line (S201). If there is no non-restored data request, it is then determined whether there is a feature point request for that line (S202). If there is a feature point request, the restoration identification number (block ID) on the corresponding coordinates is searched (S20).
3) Then, the data of the feature point related to the restoration identification number is output (S204), the process returns to S201, and the next process is performed. S
If there is no feature point request in 202, it is then determined whether or not all data has been output (S210). If there is data that has not been processed, the process returns to S201 and the next process is performed.
If there is a non-restored data request (YES in S201), there is data in the corresponding coordinates (YES in S205), and if the feature point flag is not set in the same block (S2).
If YES in 06), the data is output (S207). Then, the process returns to S201 to perform the next processing. When outputting to the feature point generator 22 in the second embodiment, the original data and the block information are output.

(D) Image Restoration The image restoration unit 13 develops a plurality of vectorized contours, scans the rasterized contours in a raster pattern, and applies odd-numbered to even-numbered points of the contours to predetermined density. Fill with. As the filling density, the average density of the blocks including the feature points is used. An example of a rule representing the contour is shown below (see FIG. 18). -A straight line having a length equal to or longer than a predetermined length is composed of a start point and an end point, and each point is composed of two feature points (total of 4 points) arranged at a minute distance.・ Short arcs are approximated by short straight lines. In the case of arranging within a minute distance, the total of the characteristic points is not judged to be an irregular point, and is therefore set to other than three points. -A long curve is approximated by an arc passing through a start point, a midpoint, and an end point that are separated by a predetermined distance or more. The midpoint is located exactly in the middle of the start point and the end point, and each point is a single feature point. The irregular points are composed of three points arranged within a predetermined minute distance, and the contour deformation method is determined by the positional relationship with the front and rear feature points.・ Points that do not fit in either case are considered to be straight lines. -One curve includes at least one straight line. FIG. 19 is a block diagram of the image restoration unit 13. The program executed by the image restoration processor 134 is stored in the program memory 135. The image restoration processor 134 executes this program, and when a restoration data request is generated through the bus B11, the feature points are input through the bus B1, restoration is performed on the block-filling work memory 137 for each block, and the result is stored in the output buffer 138. Unfold,
The restored image is output to the bus B11 for each line unit.

FIG. 20 shows a flow of image restoration processing executed by the image restoration processor 134. First, initialization is performed (S401) and the output buffer 138 is cleared (S
402). Next, if there is a restoration request from the bus B11,
(YES in S403), the feature point of the block starting from the corresponding line is input (S404). Next, the restoration identification number (block ID) is determined (S405), and the corresponding 1
The block is restored (S406, see FIG. 21) and output to the output buffer 138 (S407). Next, if all blocks have not been restored for that line (S408).
No), the process returns to S405 to continue the processing of the next block. When all blocks are restored (YES in S408),
The data of the one line is output (S409). next,
If it is determined that the data of all lines have not been restored (NO in S410), the process returns to S403 and waits for a restore request for the next line. If it is determined that the processing for all lines is completed (YES in S410), the process returns to S402 and the output buffer is cleared (S402).

FIG. 21 shows one block restoration (S40 in FIG. 20).
The flow of 6) is shown. First, a straight line portion is searched for in one contour to determine a starting point (S421). Figure 1 shows the starting point
It is performed as shown in FIG. The irregular point can be determined by the fact that there are three feature points within a predetermined minute distance. A straight line should have a number of feature points other than three located in a line within a short distance, and be connected to another similar set of feature points by a long distance. Can be determined by. Circular arcs with a short distance are represented by short straight lines. An arc is an intersection of a straight line perpendicular to the center of the straight line connecting the feature points A and B and a straight line perpendicular to the center of the straight line connecting the feature points B and C for the three feature points A, B and C. When the center is, the characteristic points A, B, C from this center
It can be determined that the distance to is within a predetermined error. Next, one feature point is extracted for one contour (S422). When it is determined that the extracted feature points are irregular points (YES in S423), the remaining two characteristic points are extracted (S424), the positional relationship between the irregular point and the characteristic points before and after the irregular point is checked (S425), and the deformation curve is worked. Drawing in the memory 136 (S426). If it is determined that the point is a straight line (YES in S432), a straight line segment is taken out (S
433), and draw a straight line in the working memory 136 (S43).
4). Furthermore, if it is determined that the point is an arc (S4
(YES at 35), the data for the arc is extracted (S43
6) Draw an arc in the work memory 136 (S43)
7). If the point is not an arc (NO in S435),
In step S434, a straight line is drawn. Next, it is determined whether the final position is the start point (S427). If it is not the start point,
Returning to S422, the processing of the next point is continued. If it is determined that the final position is the start point, the processing of one contour is completed, so it is next determined whether the processing of all contours (all vectors) is completed (S428). Is S
Returning to 421, the next contour restoration process is performed. If processing of all contours is completed, scanning is performed line by line (S42
9) Then, the space between the contour lines is painted at the designated density in the painting work memory 137 (S430). This filling is continued until it is completed for all lines (YES in S431).

(E) Feature Point Addition Unit FIG. 22 shows the mechanism of the feature point addition unit 14. The feature point addition unit 14 inputs the feature point coordinates and adds data of a predetermined density to the output image at the corresponding coordinates. Density correction unit 14
1 is not used in this embodiment. The density is shifted so that an image having the same density as the density used for the feature points included in the non-restored image is not mixed with the feature points. When the feature point coordinate b is input, the selector 143 selects the feature point density data instead of the image information. As a result, the image data and the feature point data are combined. Further, when the processing mode is set on the operation unit 16, addition of feature points by the selector 143 can be prohibited. As described above, the operation unit 1
When the edit mode is designated by 6, the editing unit 12 independently edits the extracted feature point data and the image data of the region not including the feature points output from the restoration unit 13 such as rotation and enlargement. Processing is done. 2 after editing
The seed data can be synthesized by the above-described feature point addition processing.

B. Second Embodiment (f) Generation / Addition of Feature Points During Hard Copy For documents with feature points added during document generation as described in the above embodiments, image deterioration is suppressed even when copying is repeated. Although it is possible, for a normal document that does not include feature points or a document that is cut and pasted on a part of the normal document,
Image deterioration due to copying cannot be suppressed. Therefore, by adding a feature point to such an original during hard copy, deterioration can be prevented in subsequent copies. FIG. 23 is a block diagram of an image processing circuit of a digital copying machine capable of generating additional feature points. In this image processing circuit, the image analysis unit 21 performs feature point extraction and area determination processing on digital image data input from an image reading device or the like via the bus B1. That is, the characteristic points are extracted from the input image data, and the extracted characteristic points are used for area discrimination. That is, the input image data is separated into a restorable area that includes the feature points and a non-restorable area that does not include the feature points. In the non-restorable area, the input image data is used as it is in the subsequent processing, and the feature point generation unit 22 generates feature points from the input image data. In the restorable area, the input image data is not used, and the image restoration unit 23 restores and outputs the image data using the extracted feature points. These image data are combined by the feature point adding unit 24, and the generated feature points are added. When multiple densities of feature points can be set, the feature point having the density closest to the densities near the feature points is added. Further, the feature points can be output without being added depending on the processing mode set by the operation unit 25. Finally, in the optimization unit 26, the density of the image is corrected by smoothing processing or the like according to the processing mode input from the operation unit 25,
Output to bus B2.

FIG. 24 shows a method of generating feature points. With respect to the input image (a) of the non-restored area determined to have no feature point added by the area determination, the contour line is first extracted (b). If the shape is simple, the contour feature points may be immediately picked up from this information.However, in the case of characters, noise such as when reading is included, so smoothing is performed once to make the contour straight. Approximate (c).
Next, the coordinates of the feature points are set at the intersections of the straight lines and inside the pattern (d).

FIG. 25 is a block diagram of the feature point generator 22. The feature point generation processor 181 uses the bus B15,
Processing is started in response to a request input from the B16 via the interfaces 185, 186, and information of a block not containing a feature point (non-restoration information) is input from the bus B12 via the interface 184. Correspondingly, feature points are generated and input to the feature point memory 187. The generated feature point information is output to the bus B15, and the original image is output to the bus B16. The data structure in the feature point memory 187 is the same as that of the image analysis processor 123 (see section (c-2)). Feature point generation processor 181
The contents of the process (see FIGS. 26 to 29) are as follows. Input the information of the block starting from the requested line (data of the block management memory 124).・ Trace the contour to approximate an arc or straight line. -If there is a contour whose restoration time can be shortened by the irregular processing, and if there is, transform the contour.・ When one line is completed, the feature point is stored in the feature point memory 187.
To the bus B15, the original image is input from the bus B12 (contents of the image memory 111), and is output to the bus B16.

FIG. 26 shows a control flow of the feature point generation processor 181. After initialization (S601), the process waits for a feature point generation request (S602). When requested, the information of the block starting from the corresponding line is input (S603),
Decide the target restoration identification number (block ID) (S6
04), feature point generation processing (S605, refer to FIG. 27) and irregular processing (S606, refer to FIG. 29). This process is repeated until all blocks are completed (YES in S607). Then, the original data of one line is input from the bus B12 (S608) and one line is output (S609). Then, the process returns to S602 and waits for the next request. FIG. 27 shows a flow of feature point generation (S605). A feature point having the designated restoration identification number (ID) is searched for (S621), and the contour is traced to a length corresponding to the minimum arc (S622). That is, the contour is traced while calculating the distance based on the block information. Next, an arc determination is performed (S623, see FIG. 28),
If it is determined to be an arc (YES in S624), the next feature point is determined by one pixel (S625), and the arc determination is performed (S626, see FIG. 28). And it is an arc (S6
If YES is determined in 27, the process returns to S625, and the arc of the next feature point is determined. Not an arc (N in S627
If it is determined to be O), the data of the characteristic points of the circular arc up to that point is output to the characteristic point memory 187 (S628). S62
If it is determined that the arc is not a circular arc in 4, then straight line determination is performed by the least squares method (S632). Then, if it is determined to be a straight line (YES in S633), the next feature point is determined for one pixel (S634), and the straight line determination is performed (S63).
6). When it is determined that the line is not a straight line (NO in S636), the data of the long straight line feature points is stored in the feature point memory 18
It outputs to 7 (S638). It is neither an arc nor a straight line (S
If it is determined to be NO in 633, the section is output to the feature point memory 187 as a minute straight line (S638). This determination and output are continued until the feature point becomes the same as the start point (YES in S639), that is, until one contour ends. And
The contour is connected to the starting point (S630) to form one contour. Further, the above processing is completed for all contours (Y in S631).
Continue until S).

FIG. 28 shows an arc determination (S623 in FIG. 27).
The flow of S626) is shown. First, the start point and the end point are determined (S651), and the midpoint is obtained (S652). If necessary
Perform noise processing. Next, it is determined whether the midpoint is inside the pattern (S653). If it is not internal, it is determined to be an error (S658), and this processing ends. If it is inside, then a provisional center is obtained (S654), and the distance from the center to each point is obtained (S655). If it is determined that the distance to each point is within a predetermined error (YES in S656), it is determined to be an "arc" (S657), otherwise, it is determined to be an error (S658), and the data is ignore.

FIG. 29 shows an irregular process (S606 in FIG. 26).
Shows the flow of. First, find a point that is not a straight line (S67
1), compare with a predetermined pattern (S672). If the error is less than the predetermined value (YES in S673), the transformed data is replaced (S674). This is repeated until each feature point ends (YES in S675).

C. Third Embodiment (g) Density Reference Pattern At the characteristic points described above, the density level of the image is important. Therefore, how to accurately read the density of feature points becomes an issue. As a method of optimizing the density, a method of reading a reference pattern and correcting the input data according to the density variation of the reference pattern, as used in automatic exposure of a copying machine, can be used. As a method of creating the reference pattern, there is a method of printing in advance in an area in the document that does not hinder the use. FIG. 30 shows an example of a layout on a document. The reference pattern is printed on the lower left edge of the document. The reference density (left side) is also printed together with the mark (M1, M2, ...) Of each feature point and additional information (right side) such as the number of used densities of the feature point. Similar to the automatic exposure, the reference pattern may be read by a temporary manuscript operation prior to the actual reading, or all the manuscript information may be read and the reference pattern may be taken out. Figure 31
Shows an example of the flow of processing in the latter case where all the document information is read and the reference pattern is taken out. That is, FIG.
In the image analysis unit 11 shown in FIG.
Reference numeral 23 performs feature point extraction and area discrimination, but in the present embodiment, reference density detection is performed before feature point extraction. Then, in the feature point extraction, the image data from the image memory 111 is compared with the detected reference density, and if the determination result matches the predetermined reference density, it is determined to be a feature point.
In this reference density reading, when feature points of a plurality of density levels are set as in the example described below, the density of the reference pattern corresponding to each feature point is read and used for the feature point determination. ..

(H) Addition of feature points at low density in a halftone image In a halftone image, if the feature points are added to a part of the original document by changing the density, the effect on the halftone image is reduced. It is necessary to add feature points like this. When feature points with a high density are added as in the first embodiment, the feature points may cause image noise. Therefore, in this embodiment, a plurality of densities are used as the densities of the characteristic points so as to cover the entire density distribution of the image. In this method, characteristic points are assigned to a plurality of density bands, and the density closest to the density of the image to which the characteristic points are to be added is assigned to the characteristic points. The density of the feature point is the density at the center of each density band. As a result, even for a document including a low density area, it is possible to add feature points of low density by minimizing the influence on the image. In the example shown in FIG. 32, the concentration In of four feature points In
(N = 1, 2, 3, 4) is used, and the concentration In (n =
2, 3, 4) are within the image density area. Except for the density In of the feature points, the input image data is directly captured as image data. The right side of FIG. 32 shows a situation in the vicinity of one feature point density I2. However, in the vicinity of the feature point, when a hard copy is generated, the feature point is written with a predetermined density In, but when read by a reading device, variations occur due to the quality and reading accuracy of the hard copy. Therefore, the width of the variation is set to Wa, and the image having the density in this range is regarded as the feature point. Further, since the image data also varies, the range of the width Wb on both sides of the image data is the same level Dn.
Image.

(I) Extraction of Feature Point Corresponding to Continuity of Density Change If a density band of feature points is assigned within the density range of image data as described above, there arises a problem of reduction in dynamic range. In other words, since the density range that can be expressed on the hard copy is determined, if the characteristic points are assigned, the range that can be used as an image will be reduced. Also,
Feature points can cause image noise. In a multi-value halftone original, there is an image whose density changes continuously (see FIG. 33).
(A)). In this case, since the density in the vicinity assigned to the feature point cannot be used on the hard copy, streak-like noise is generated along with the density. On the other hand, in the multi-value halftone image, there is an image in which the density is constant for each area ((b) of FIG. 33). In this case, if there is a density in the vicinity of the feature point, a change in the image can be avoided by forcibly moving the density of the feature point slightly. Specifically, the density correction unit 141 of the feature point addition unit 14 in FIG. 22 shifts the density so that an image having the same density as the density used for the feature points included in the non-restored image is not mixed with the feature points.

The two types of images can be discriminated as follows. Now, it is assumed that the characteristic point of the density I1 is at the position P1.
It is desirable to make the discrimination both vertically and horizontally. When the density is read in one direction and the density continuously changes in the vicinity of the feature point as shown in FIG.
It is not considered as a feature point but as a part of image data. on the other hand,
When the density becomes discontinuous in the vicinity of the feature point as shown in FIG. 34B, it is regarded as the feature point. Therefore, in the blocking process, the change in the density of the image is detected in one direction. Therefore, if the density change is similarly detected and it is determined that the image continuously changes at the density assigned to the feature point, , A characteristic point with a different density from that density is added, and if not, a characteristic point with that density is added.
Even when reading a hard copy, if it is determined that the image continuously changes at the density assigned to the characteristic point from the density change in one direction, it is determined that the image is not the characteristic point and it is regarded as a part of the image data. Good. If the density changes discontinuously, it is determined that the feature point is detected.

(J) Addition of feature points to a halftone image in which the densities are continuously changed In principle, a band of a predetermined density is assigned to the feature points. However, as shown in (a) of FIG. When the image density changes continuously, all bands are assigned to the image as in the first embodiment. If the density changes continuously, this does not cause a problem because it is not necessary to assume a contour line in the area where the density changes continuously. But,
In an image in which the density changes continuously, noise is generated when the characteristic points are assigned in the original density area, as described above. Therefore, in order to suppress the occurrence of noise in image data whose density changes continuously, in this embodiment, in the document density area, the sensitivity is continuously and linearly assigned to the document density. When the density change becomes discontinuous as shown in FIG. 34B, all internal data can be expressed even if the density band Wa assigned to the feature points is excluded. The sensitivity is partially changed, as shown enlarged. That is, in the band Wb following the band Wa assigned to the characteristic points, the inclination of the image data with respect to the document density is made different. As a result, the document density in the band Wa is assigned to the outside of the band Wa. Further, the width of the band for which the sensitivity is changed may be changed from the band for the next lower characteristic point to the current band. FIG. 36 shows the configuration of an image processing unit for identifying feature points from related images in this method. The difference from the processing of the image analysis processor 123 described above is the blocking processing 112 in the area discrimination processing. Here, a block is regarded as a different block not only in a pattern that is simply separated but also in an image in which adjacent density differences have a predetermined level or more. Then, the area identification 113 is performed on the block thus obtained. In addition, for the density in the vicinity of the feature point in the image of the area not including the feature point, the density correction processing 114 is performed to correct the internal data so as to be continuous. That is, the data within the width Wb of FIG. 35B is converted into the data within the range of the widths Wa and Wb.

C. Third Embodiment (k) Addition of Feature Points when Creating Original Document It is desirable that feature points be added when creating an original document. FIG. 37 shows the configuration of a digital copier with a printer function for adding feature points when an original document is created by a computer (word processor or the like) having an outline font. The outline font code information from the computer is input to the code processing unit 31, generates image data, is output to the image processing unit 33, and is printed by the printing unit 34. On the other hand, when the characteristic points are added to the original document, the image data read from the document is input to the image processing unit 33 via the image input unit 32. Further, from the code information, in the code processing unit 31,
The feature points are calculated from the outline font data and sent to the image processing unit 33. By setting the processing mode from the operation unit 35, the feature points are added in the image processing unit 33. The data to which the characteristic points are added by the image processing unit 33 is printed on paper by the printing unit 34, and a hard copy is obtained. FIG. 38 shows the image processor 33 and the code processor 31.
Shows the configuration of. In the code processing unit 31, the code information is divided by the code analysis unit 51 into a character code indicating a character and a format control code designating a print position of the character.
The drawing position control unit 52 determines the printing position of each character based on the character size information obtained from the format control and the character pattern generation unit 54. The character pattern generation unit 54 (see FIG. 39) generates a pattern of characters specified by the character code from the code analysis unit 51, and also generates feature points of the pattern from the outline font data. In the drawing processing unit 53, at the print position obtained by the drawing position control unit 52,
The image in which the pattern obtained by the character pattern generation unit 54 is drawn is edited. The edited image is added with the feature points obtained by the character pattern generation unit 54 by the feature point addition unit 44, corrected by the optimization unit 45 such as density, and then output. The image processing unit 33, similar to the first embodiment, has an image analysis unit (performs area discrimination and feature point extraction) 41 and an editing unit 4.
2, an image restoration unit 43, a feature point addition unit 44, and an optimization unit 45.

FIG. 39 shows the structure of the character pattern generator 54 for adding feature points to characters. The character pattern generator 54 uses an outline font. The outline font has coordinates (control points) necessary for generating the outline of the character. The control point calculator 541 calculates control points from the outline font. The outline font information section 542 sends the character size information to the drawing position control section 52. The contour generating unit 543 generates a contour from the coordinates of the control points, and the filling unit 545 fills the inside of the contour to obtain a character pattern. Further, the feature point calculation unit 544 calculates the feature point from the control point and outputs it to the feature point addition unit 44. FIG. 40 shows an example of the control points that represent the outline of the outline font. This is 4 points (P0-
This is an example of using a cubic Bezier curve having P3) as a parameter. The outline font information is mainly composed of parameters (control points in this case) for a predetermined function (cubic Bezier in this case). The outline information of the outline font can be placed outside the font pattern, but the feature points must be placed inside the character pattern. A simple method is linear approximation. The amount of data is large to use as a control point of the outline font, but there is no problem in the feature point. FIG. 41 shows a method of converting control points into feature points. The concept of conversion is the same as the case of calculating a contour line from control points. First, the midpoint P4 between the control points P0 and P1,
A midpoint P5 between control points P2 and P3 and a midpoint P6 between control points P2 and P3 are obtained. Next, the line segments connecting P4 and P5 and P5 and P6 are calculated, and P0-P4-P5-P is used as an approximation of the contour line.
6-P3 is calculated. When the approximation is insufficient, the midpoints of these line segments are further connected. The feature points use the intersections of the approximated straight lines. In the case of characters, it was calculated from the outline font information, but even when printing a figure such as a circle or a straight line, the calculation of the characteristic points can be done easily.

D. Fourth Embodiment (l) Error Processing Even when information for restoration such as feature points is added, when partial cutting and pasting of the pattern is performed or when the additional information cannot be read correctly due to contamination of the original, Not only the deterioration of image quality due to recopying cannot be suppressed, but it may not be possible to restore correctly. FIG. 42 shows a case where an error occurs in the restoration. It is assumed that there is a reading error of the feature point from the input image of (a) and the contour of (b) is obtained. From this contour,
The restoration pattern of (c) is obtained. This restoration pattern is compared with the input image (a), and if there is a difference of not less than a predetermined value as shown in (d), it is considered that there is an error in reading the feature point. In FIG. 42D, the hatched portion is a portion where the restoration pattern (c) and the input image (a) are different. When such an error is detected, the feature point is newly generated by the method described above. Figure 4
3 shows the configuration of an image processing unit having a reading error detection function and an error correction function. The image processing unit includes an image analysis unit 61, a feature point generation unit 62, an image restoration unit 63, a feature point addition unit 64, and an optimization unit 66, as in the case of the second embodiment (FIG. 23). The configuration of the image analysis unit 61 is different.
In the image analysis unit 61, after performing the block processing 612 on the image data read from the image memory 611,
The image block identifying unit 613 also cuts out a block including a feature point and sends it to the comparing unit 614. On the other hand, the restored data (c) is obtained in the image restoration unit 63 based on the feature points extracted by the feature point extraction unit 618. This restored data is also sent to the comparison unit 614. The comparison unit 614 compares the non-restored data (a) from the block identification unit 613 with the pattern restored from the feature points by the image restoration unit 63. If there are few errors, the switches 615 and 616 are closed and restoration is performed. The image data and the original feature points are passed through the feature point generation unit 62, and only the feature points generated by the feature point generation unit 62 for the non-restored image are newly added. When an error occurs, the comparison unit 1
It is desirable to send a signal from 9 to the operation unit 4 to display an error occurrence on the display unit. In the present embodiment, the points inscribed in the contour are used as the characteristic points, but they do not necessarily have to be inscribed.

[0042]

The information of the original is digitally added to prevent deterioration of the image quality due to copying, and
As before, it can also be used as a visible hard copy. Even for a multi-valued original including halftone, data for restoration can be added digitally without affecting the original. Even for a multi-valued original including halftone, even if there is a change in gradation, data for restoration can be added digitally without affecting the original. In the case where the document contains restoration information about important points on the contour in advance, it is possible to restore the contour that is missing due to the limit of reading resolution or an error during reading. Since the restoration information is added in addition to the image information to be output, the restoration can be performed even when the hard copy is generated or the hard copy is read as an original even if it is slightly deformed. Since restoration information is added (combined) to the image information in the image data output at the time of hard copy, there is little deterioration in image quality even after re-copying after processing.

[Brief description of drawings]

FIG. 1 is a diagram of an example of an image to which feature points are added.

FIG. 2 is a diagram schematically showing a process for the image of FIG.

FIG. 3 is a diagram showing examples (a) to (d) of contour points.

FIG. 4 is a diagram of determination output in the case of a multi-valued original.

FIG. 5 is a diagram showing a determination output in the case of a binary original.

FIG. 6 is a block diagram of an image processing circuit of a digital copying machine.

FIG. 7 is a block diagram of an image analysis unit.

FIG. 8 is a flowchart of processing of an image processing processor.

FIG. 9 is a diagram showing an example of blocking.

FIG. 10 is a diagram showing an example of block synthesis of feature points.

FIG. 11 is a flowchart of a blocking process.

FIG. 12 is a flowchart of a feature point extraction process.

FIG. 13 is a block diagram of an image restoration unit.

FIG. 14 is a flowchart showing part of vectorization of irregular points.

FIG. 15 is a flowchart showing a part of vectorization of irregular points.

FIG. 16 is a diagram showing an example of vectorization of irregular points.

FIG. 17 is a flowchart of output processing.

FIG. 18 is a diagram showing an example of a rule representing a contour.

FIG. 19 is a block diagram of an image restoration unit.

FIG. 20 is a flowchart of image restoration.

FIG. 21 is a flowchart of 1-block restoration.

FIG. 22 is a block diagram of a feature point addition unit.

FIG. 23 is a block diagram of an image processing circuit capable of generating additional feature points.

FIG. 24 is a diagram showing an example of generation of feature points.

FIG. 25 is a block diagram of a feature point generation unit.

FIG. 26 is a flowchart of processing of a feature point generation processor.

FIG. 27 is a flow chart of feature point generation.

FIG. 28 is a flowchart of circular arc determination.

FIG. 29 is a flowchart of irregular processing.

FIG. 30 is a diagram showing an example of a layout of a density reference pattern on a document.

FIG. 31 is a diagram showing a flow of reading a reference pattern.

FIG. 32 is a diagram showing the densities of feature points when a plurality of density bands are used.

FIG. 33 is a diagram illustrating an example of an image in which the density continuously changes and an image in which the density is constant in each region.

FIG. 34 is a diagram showing a change in density of the figure in FIG. 33.

FIG. 35 is a diagram of an example of partially changing the sensitivity.

FIG. 36 is a flowchart of a process of identifying a feature point from a related image.

FIG. 37 is a block diagram of adding feature points when creating an original.

FIG. 38 is a diagram showing a configuration of an image processing unit and a code processing unit.

FIG. 39 is a block diagram of a character pattern generation unit that adds a feature point to a character.

FIG. 40 is a diagram showing an example of control points representing the outline of an outline font.

FIG. 41 is a diagram showing an example of conversion from control points to feature points.

FIG. 42 is a diagram showing detection of a restoration error.

FIG. 43 is a block diagram of an image processing unit including a reading error detection unit and an error correction unit.

[Explanation of symbols]

21 ... Image analysis unit, 22 ... Feature point analysis unit, 23 ... Image restoration unit, 24 ... Feature point addition unit.

Claims (4)

[Claims] 1. The image data is represented by a predetermined density, the contour restoration information relating to the contour of a preset image is represented by a density different from the density of the image data, and the contour restoration information is added to the image data for output. And an image output method capable of restoring the contour of the image from the contour restoration information when a hard copy of the image data is copied. 2. The image data is represented by a plurality of predetermined density bands, the contour restoration information relating to the contour of a preset image is represented by a plurality of density bands different from the density of the image data, and the contour restoration information is added to the image data. An image output method capable of restoring the contour of an image from contour restoration information when a hard copy of this image data is copied. 3. The image output method according to claim 2, wherein when there is density information corresponding to the density of the contour restoration information in the image data of the data to be processed, the density of the image data is changed. An image output method characterized in that the density is changed to a density not included in the density band of the image restoration information. 4. Continuity of changes in image density, wherein image data is represented by a plurality of predetermined density bands, and contour restoration information regarding preset image contours is represented by a plurality of density bands different from the density of image data. When the determination unit determines that the change in the image density is continuous in the vicinity of the density band of the contour restoration information, the density band is not assigned to the contour restoration information, and the image data And an image output method capable of restoring the contour of the image from the contour restoration information when a hard copy of the image data is copied.