IMAGE PROCESSING APPARATUS, CONTROL METHOD FOR IMAGE PROCESSING APPARATUS, AND PROGRAM

Abstract

To provide an image processing device capable of properly detecting a sheet edge even when image noise is generated and, in particular, to provide an image processing device capable of accurately obtaining print position adjustment parameters.SOLUTION: An image processing device provided with a reading unit that reads an image from a sheet includes means for reading a single image from a single sheet using the image reading unit, means for performing edge detection processing on the single image, and means for determining at least one edge pair among a plurality of edges on the basis of the information of the plurality of edges detected in the edge detection processing and the size information of the single sheet.SELECTED DRAWING: Figure 17

Description

[Technical field]

[0001] The present invention relates to picture Image processing device, Method and program for controlling image processing device Regarding. [Background technology]

[0002] Conventionally, printing devices (image processing devices) that form images on sheets have a function for adjusting the position at which an image is formed on a sheet (hereinafter referred to as "print position adjustment"). Patent Document 1 discloses a technique for acquiring parameters for print position adjustment by printing adjustment marks on a sheet, reading the marks with a reader, and acquiring the positional relationship between the marks and the edge of the paper. Patent Document 1 also describes an automatic document feeder for sheets called an ADF (Auto Document Feeder) as an example of a reader. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2016-111628 A Summary of the Invention [Problem to be solved by the invention]

[0004] The device described in Patent Document 1 has room for improvement in terms of the accuracy of detecting the paper edge. This is because linear image noise called leader lines can occur when a sheet is read using a mechanism such as an ADF. These leader lines look similar to the shadow of the paper edge, so there is a risk that they will be mistaken for the paper edge. If the paper edge position cannot be obtained correctly, the relative position with respect to the mark cannot be obtained correctly either, and the parameters for adjusting the print position cannot be obtained with high accuracy.

[0005] In view of the above-mentioned problems, the present invention provides a method for detecting an image even when image noise occurs. Sheet edge pair can be properly detected structure The purpose is to provide. [Means for solving the problem]

[0006] The present invention is an image processing apparatus, comprising: a selection unit for selecting sheet information; and a sheet corresponding to the sheet information selected by the selection unit. having a predetermined pattern including at least one mark a printing means for printing an image, a reading means for reading the image from the sheet, a detection means for detecting a plurality of candidates for the edges of the sheet from the image read by the reading means, and a determination means for determining a set of edges of the sheet from among the plurality of candidates based on the plurality of candidates detected by the detection means and sheet size information included in the sheet information selected by the selection means; a means for obtaining a parameter based at least on position information of the at least one mark and position information corresponding to one edge of the set of edges determined by the determining means; an adjustment means for adjusting a printing position by the printing means based on the The set of edges is a set of edges among the plurality of edges, in which the absolute value of the difference between the angles is equal to or less than a threshold value, and in which the distance between the edges and the length of the sheet in a predetermined direction satisfy a predetermined relationship. It is characterized by the above. Effect of the Invention

[0007] According to the present invention 、 Even if image noise occurs, Sheet edge pair To properly detect but can 。 [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a schematic diagram showing a control block diagram of a system including an image processing device. [Diagram 2] FIG. 1 is a schematic diagram of an image forming apparatus including an image processing apparatus. [Diagram 3] FIG. 13 is an explanatory diagram showing a paper library editing screen. [Figure 4] FIG. 13 is an explanatory diagram illustrating a paper library editing interface. [Diagram 5] FIG. 2 is a schematic diagram illustrating a paper library. [Figure 6] FIG. 4 is a schematic diagram of a print position adjustment chart. [Figure 7] FIG. 2 is a sequence diagram showing a series of processes in the present embodiment. [Figure 8]13 is a flowchart showing an operation of a mark relative position measuring process. [Figure 9] FIG. 1 is a schematic diagram showing input data and output data for operation by machine learning; [Figure 10] FIG. 1 is a schematic diagram of an ADF reading device involving paper transport. [Figure 11] 13 is a flowchart showing the operation of a vertical paper edge detection process. [Figure 12] 1 is a schematic diagram showing the positional relationship of top, bottom, left, and right ROI images. [Figure 13] 13 is a flowchart showing the operation of a paper edge candidate extraction process. [Figure 14] FIG. 11 is an explanatory diagram illustrating an example of a Hough table. [Figure 15] Fig. 15(a) is a schematic diagram showing an ROI image and a plurality of paper edge detection results, and Fig. 15(b) is a schematic diagram showing a binary image of the upper right ROI image. [Figure 16] 13 is a flowchart showing the operation of an optimum edge pair determination process. [Figure 17] 13 is a schematic diagram showing an example of a process for determining an optimum pair of left and right paper sides; FIG. [Figure 18] 13 is a flowchart showing the operation of a horizontal paper edge detection process. [Figure 19] FIG. 13 is a schematic diagram showing an example of a process for determining an optimal pair of upper and lower paper sides; [Figure 20] 13 is a flowchart showing the operation of vertical paper edge detection processing in the second embodiment. [Figure 21] 13 is a flowchart showing the operation of a detailed edge estimation process. [Figure 22] 10 is a flowchart showing the operation of a process for measuring a relative position of a mark from a plurality of read images. [Figure 23] 10 is a flowchart showing the operation of a paper size estimation process. [Figure 24] 13 is a schematic diagram showing histogram information of paper size estimates; FIG. [Diagram 25] 11 is a schematic diagram illustrating an example of histogram information of estimated paper size values ​​visualized. FIG. [Figure 26]FIG. 11 is a sequence diagram showing a series of processes in the third embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Example 1 <Control block diagram that enables this configuration> 1 is a block diagram showing a schematic hardware configuration of an embodiment of a printing system (image processing system). As shown in the figure, the printing system according to this embodiment is composed of an image forming apparatus 100 and a host computer 101. The image forming apparatus 100 and the host computer 101 are connected to each other via a communication line 102. Note that the printing system may be connected to a plurality of host computers, printing apparatuses, etc.

[0010] The host computer 101 can obtain input information from a user through an input device (not shown), create a print job to be sent to the image forming apparatus 100, and send it to the image forming apparatus 100. The controller 110 performs various data processing and controls the operation of the image forming apparatus 100. The operation panel 120 accepts various operations from the user using a touch panel. As described later, the paper size information 121 obtains the size and position adjustment amount of the print paper from the paper library via the operation panel 120. The image reading unit 130 is a scanner that scans an original document using an optical sensor and obtains scanned image data. As described later, the print position measurement process 131 obtains the relative coordinates of the print position for the read image of the print result of the print position adjustment chart (image of a predetermined pattern). The paper feed unit 140 is a paper feed device composed of multiple paper feed stages. Each paper feed stage can store various types of print paper. Each paper feed stage can separate only the topmost sheet of stored paper and transport it to the image forming unit 150. The image forming unit 150 physically prints image data on a printing paper (sheet). In this embodiment, an electrophotographic image forming unit 150 is described, but an inkjet type may be used as the image forming unit 150. Note that 160 represents the printed result.

[0011] Next, the configuration of the controller 110 will be described. The I / O control unit 111 controls communication with an external network. The ROM 112 is a ROM that stores various control programs. The RAM 113 is a RAM that reads and records the control programs stored in the ROM 112. The CPU 114 is a CPU that executes the control programs read into the RAM 113 and controls image signals and various devices in an integrated manner. The HDD 115 is a HDD used for the purpose of temporarily or long-term storage of large amounts of data such as image data and print data. Each module is connected to each other via a system bus 116. Furthermore, the system bus 116 connects the controller 110 to each device in the image forming apparatus 100. The RAM 113 also functions as a main memory and a work memory for the CPU 114. Furthermore, the control programs and the operating system are stored in the HDD 115 in addition to the ROM 112. Furthermore, a non-volatile random access memory (NVRAM) (not shown) may be provided to store printer mode setting information from the operation panel 120.

[0012] <Image forming device> Next, the structure and operation of the image forming apparatus 100 will be described.

[0013] 2 is a cross-sectional view showing the structure of image forming apparatus 100 in this embodiment. As shown in the figure, image forming apparatus 100 includes a housing 201. Housing 201 houses each mechanism constituting an engine section, an engine control section that controls each printing process (e.g., paper feed process) performed by each mechanism, and a control board storage section that stores a printer controller.

[0014] The image reading unit 130 is composed of a document table 230, a pickup roller 231, a transport roller 232, a roller 233, a light source 234, a second reading unit 235, a paper discharge roller 238, a reading unit 236, and a document 237. The document 237 loaded on the document table 230 is sent to the reading path one by one by the pickup roller 231. The picked-up document 237 is transported in the path direction by the transport roller 232. The document 237 reaches the reading position through the path, and the image information added to the surface of the document 237 is read by the light source 234 and the reading unit 236 together with the image information at the end of the document. A white member is disposed at a position opposite the reading unit 236, and the reading unit 236 reads the document when it passes through this member. The reading unit 236 is a reader device that is also used for reading the pressure plate. Thereafter, when the original 237 reaches the reading position of the reading unit 235, the second reading unit 235 reads image information on the back side of the original 237. A white member is disposed at a position opposite the reading unit 235, and performs reading when the original passes through the reading unit 236. The reading unit 235 is, for example, a CIS. Thereafter, the original 237 is discharged by discharge rollers 239.

[0015] By repeating the above operation, the documents placed on the document table 237 have their front and back images read in one transport.

[0016] Although the most preferred example is one in which both sides of the document are read in one transport, the present invention is not limited to this. As long as both sides of the document can be read, a reading device equipped with a reversing transport mechanism may also be used.

[0017] The mechanisms that make up the printer engine section include the following: An optical processing mechanism for forming an electrostatic latent image on the photosensitive drum 205 by scanning with laser light, visualizing the electrostatic latent image, multiply transferring the visualized image onto the intermediate transfer body 252, and further transferring the multiply transferred color image onto the sheet P. A fixing processing mechanism for fixing the toner image transferred onto the sheet P, a sheet P feeding processing mechanism, and a sheet P transport processing mechanism.

[0018] The optical processing mechanism includes a laser driver in a laser scanner unit 207 that drives a laser beam emitted from a semiconductor laser (not shown) on and off in response to image data supplied from a controller 110. The laser beam emitted from the semiconductor laser is deflected in the scanning direction by a rotary polygon mirror 208. The laser beam deflected in the main scanning direction is guided to a photosensitive drum 205 via a reflecting polygon mirror 209, and exposes the surface of the photosensitive drum 205 in the main scanning direction. On the other hand, an electrostatic latent image formed on the photosensitive drum 205 by scanning exposure with a laser beam, which is charged by a primary charger 211, is visualized into a toner image by a toner supplied by a developing unit 212 described later. The visualized toner image on the photosensitive drum 205 is transferred (primary transfer) onto an intermediate transfer body 252 to which a voltage having the opposite characteristics to that of the toner image is applied (first transfer). When forming a color image, each color is sequentially formed on the intermediate transfer body 252 from a Y (yellow) station 220, an M (magenta) station 221, a C (cyan) station 222, and a K (black) station 223. As a result, a full-color visible image is formed on the intermediate transfer body 252.

[0019] Next, the sheet P fed from the transfer material storage 210 is transported, and the sheet P is pressed against the intermediate transfer body 252 by the transfer roller 251, and at the same time, a bias having the opposite characteristics to the toner is applied to the transfer roller 251. As a result, the visible image formed on the intermediate transfer body 252 is transferred (secondary transfer) onto the sheet P which is transported by the paper feed processing mechanism in synchronization with the transport direction (sub-scanning direction) of the sheet P.

[0020] After the secondary transfer, the sheet P passes through the fixing device 260, whereby the toner transferred onto the sheet P is heated and melted, and fixed as an image on the sheet P. In the case of double-sided printing, the sheet passes through the inversion section 270, where it is switched back and reversed, and is introduced again into the transfer section 240, whereby the back-side image is transferred onto the sheet P. Thereafter, the sheet P passes through the fixing device 260 in the same manner as above, whereby the toner image on the sheet P is heated and fixed, and the sheet is discharged outside the printer, completing the printing process.

[0021] <Paper Library> The paper sheets used for printing in the image forming apparatus 100 are managed by the operator using a database called a paper library. The paper library is stored in the HDD 115 or RAM 113, and is read and written as necessary by each software module. The detailed configuration of the paper library will be described later with reference to FIG. 5.

[0022] 3 is a schematic diagram showing an interface screen for an operator to perform operations such as editing the paper library in the printing system of this embodiment. Reference numeral 300 denotes the entire interface screen displayed on the operation panel 120 by the CPU 114.

[0023] Reference numeral 310 denotes a paper list, in which the papers stored in the paper library are displayed in a list. In the paper list 310, paper attributes such as those shown in columns 311 to 315 for each paper are presented to the operator as accompanying information. Column 311 indicates the paper name of each paper. The paper name is a name designated by the operator or the like so that each paper can be distinguished from the others. Columns 312 and 313 indicate the paper length in the sub-scanning direction and the paper length in the main scanning direction for each paper, respectively. Column 314 indicates the basis weight for each paper. Column 315 indicates the surface quality for each paper. Here, the surface quality is an attribute that indicates the physical characteristics of the paper surface, such as "coated" in which the surface is coated to increase glossiness, or "embossed" in which the surface has an uneven surface. In addition, by touching a portion of the paper list 310 on the operation panel 120 where the paper is displayed, the paper can be selected. The selected paper is highlighted (inverted). In FIG. 3, as an example, "XYZ Paper Color 81" is shown to be selected. Furthermore, if the number of papers recorded in the paper library is greater than the number of papers that can be displayed at one time in the paper list 310, a scroll bar 317 is used. The operator can select any paper by operating the scroll bar 317.

[0024] Reference numeral 320 denotes a new add button, which is a button for adding a new paper to the paper library. Reference numeral 321 denotes an edit button, which is a button for editing the paper attributes of the paper selected in the paper list 310. When the new add button 320 or the edit button 321 is pressed, an interface screen such as that shown in Fig. 4 is displayed. Reference numeral 322 denotes a delete button, which is a button for deleting the paper selected in the paper list 310 from the paper library. Reference numeral 323 denotes a print position adjustment button, which is a button for adjusting the print position for the paper selected in the paper list 310.

[0025] <Paper library editing interface> 4 is a schematic diagram showing an interface screen for an operator to edit paper attributes in the printing system of this embodiment. A screen 400 shows the entire interface screen that the CPU 114 displays on the operation panel 120.

[0026] Text boxes 401 to 404 are text boxes for inputting paper attributes, such as paper name, paper length in the sub-scanning direction, paper length in the main scanning direction, and basis weight. Input into the text boxes is performed using a software keyboard (not shown) or a numeric keypad provided on the operation panel 120. 405 is a combo box for specifying the surface properties of the paper. In the combo box 405, it is possible to specify one from a list of surface properties that are supported by the image forming apparatus 100 and registered in advance.

[0027] A stop editing button 420 is pressed to confirm the paper attributes entered at that time and save them in the paper library. The interface screen 400 is then closed, and the screen returns to the paper library editing screen 300. A cancel button 421 is pressed to cancel the paper attribute editing process, close the interface screen 400, and return to the paper library editing screen 300.

[0028] <Contents of the Paper Library> 5 is a schematic diagram showing a paper library stored in the HDD 115 etc. Although a schematic diagram is used here for the sake of explanation, in reality the paper library is stored as digital information such as XML or CSV.

[0029] Each of the paper information 501 to 508 represents a paper registered in the paper library.

[0030] Columns 511 to 515 show the paper attributes specified by the operator for each paper. Column 511 shows the paper name. Columns 512 to 515 show the paper attributes that indicate the physical characteristics of the paper, and respectively show the paper length in the sub-scanning direction, paper length in the main scanning direction, basis weight, and surface property.

[0031] Columns 520 and 521 indicate the amount of print position deviation for the front and back sides of each paper, respectively. The amount of print position deviation here indicates the amount of position deviation from the ideal print position, and in this embodiment, it is composed of the items of lead position, side position, main scanning magnification, and sub-scanning magnification. During actual printing, the image forming device 100 adjusts based on these amounts of print position deviation so that printing is performed at the ideal print position (i.e., adjusts to cancel the amount of print position deviation) and prints. The lead position and side position indicate the amount of print position deviation in the sub-scanning direction and main scanning direction, respectively, for the paper. The lead position is adjusted by changing the print start position of the image starting from the edge of the paper at the beginning of the paper transport direction, and the side position is adjusted by changing the print start position of the image starting from the edge of the paper on the left side of the paper transport direction. The sub-scanning direction magnification indicates the deviation of the image length in the sub-scanning direction (magnification relative to the ideal length). The main scanning direction magnification indicates the deviation of the image length in the main scanning direction (magnification relative to the ideal length). The skew indicates the degree of parallelism to any one side of a rectangular piece of paper. The trapezoidal degree is the degree of parallelism between the leading and trailing edges of an image printed on the paper, and the squareness indicates the degree of rectangularity of the image printed on the paper.

[0032] These print position deviation amounts are calculated by printing an adjustment chart on which specific marks are arranged and detecting the positions of the marks on the printed adjustment chart. An example of an adjustment chart is explained using Figure 6. Note that the initial values ​​of these print position deviation amounts are 0 for each item, and the initial values ​​are used when new paper information has just been registered in the paper library, or when a paper has been registered but the print position has not been adjusted.

[0033] <Print position adjustment chart and measurement details> 6 is a schematic diagram showing an example of an adjustment chart used for adjusting the print position. The adjustment chart is printed by an instruction from the CPU 114 after the print position adjustment starts.

[0034] Chart manuscript 601 represents the printed adjustment chart. Marks 602 to 605 represent marks printed at specific positions on the adjustment chart. The marks are printed in a total of eight places, at the four corners of both the front and back sides of the adjustment chart, and the images are arranged so that, if the print positions are ideal, the marks are printed at positions that are a fixed distance away from each of the four corners of the paper. The amount of deviation in the print positions can be found by measuring the relative positions from the four corners of the paper on the adjustment chart.

[0035] In this embodiment, the areas indicated by (A) to (H) in Fig. 6 are measured. (A) to (H) are all distances from the marks 602 and 605 to the nearest edge of the paper, and in this embodiment, the specified distances for (A) to (H) are 10 mm.

[0036] <Print position measurement unit operation> The calculation methods of (A) to (H) will be described below with reference to the flowchart in Fig. 8. Please refer to the schematic diagram in Fig. 6 for the positional relationships of the upper left vertical edge, upper left horizontal edge, upper right vertical edge, upper right horizontal edge, lower left vertical edge, lower left horizontal edge, lower right vertical edge, and lower right horizontal edge mentioned in the flowchart below.

[0037] In step 801 (hereinafter, referred to as S801, etc.), the CPU 114 acquires a read image of the adjustment chart via the image reading unit 130. Note that the image reading unit 130 may be an external reading device such as an ADF, or may be a reading device installed inside the printer, such as an in-line sensor.

[0038] In this embodiment, an example is taken of reading by an ADF that transports paper using two rollers, one on the left and one on the right. When transporting paper using the rotation of the transport roller 232 shown in FIG. 2, multiple rollers may be arranged on the left and right depending on the paper size and the torque of the roller. This example is shown in FIG. 10. This is a top view of the image reading unit, and the transport roller 232 in FIG. 2, which is shown in cross section, becomes two transport rollers, roller 1001 and roller 1002, in this figure. Using these rollers, the adjustment chart 1003 is transported in the direction of the arrow 1004. In this case, in order to transport the chart in parallel, these two rollers need to rotate at the same speed. If there is variation in the diameter of the rollers, a speed difference occurs on the circumference of the left and right rollers, making parallel transport difficult. In this figure, for example, if the roller diameter of roller 1001 is larger than that of roller 1002, the speed of the paper passing through roller 1001 increases, and the paper is transported tilted to the left as indicated by arrow 1005, and the image acquired by the reading unit 235 becomes distorted. The state of distortion of the acquired image is shown in image 1006. Although the image is kept horizontal at the beginning of the scan, the speed difference causes the right side to gradually shrink, creating a difference with the left side. As a result, the parallelism of the top and bottom sides is lost. In the subsequent processing, we will assume that the images obtained in this way do not necessarily have parallel top and bottom sides.

[0039] In S802, the CPU 114 acquires the positions of the marks 602 to 605 from the read image. The mark positions are expressed as two-dimensional coordinates in the horizontal (x) and vertical (y) directions of the scanned image, with the upper left coordinate set as the origin (0,0). The mark positions may be expressed as the coordinates of the center of gravity of the mark, or may be calculated with sub-pixel accuracy. Furthermore, the detection process of the mark may be performed by a pattern matching method.

[0040] In S803, the CPU 114 acquires the paper size of the adjustment chart from the paper library editing screen 300. Specifically, the sub-scanning direction paper length (mm) 312 and the main scanning direction paper length (mm) 313 are acquired.

[0041] In S804, the CPU 114 detects the upper left vertical edge and the upper right vertical edge of the paper. The information on the edges is used to calculate the distances (B) and (D). The details of this process will be described later.

[0042] In S805, the CPU 114 detects the lower left vertical edge and the lower right vertical edge. The information on the edges is used to calculate the distance (F) and the distance (H). The details of the process are the same as in S804.

[0043] In S806, the CPU 114 detects the upper left and lower left horizontal edges of the paper. The information on the edges is used to calculate the distances (A) and (E). The details of this process will be described later.

[0044] In S807, the CPU 114 detects the upper right horizontal edge and the lower right horizontal edge of the paper. The information on the edges is used to calculate the distance (C) and the distance (G). The details of the process are the same as those of S806.

[0045] In S808, the CPU 114 calculates the relative position of the mark from the detected eight paper sides. In this process, the CPU 114 calculates the distances (A) to (H) by calculating the normal distance from the marker center coordinates to the linear expression (ρ=xcosθ+ysinθ) of each paper side.

[0046] <Vertical paper edge detection process> The process of detecting the vertical paper edge performed in steps S804 and S805 will be described in detail below with reference to the flowchart of FIG.

[0047] In S1101, the CPU 114 acquires left and right ROI (Region Of Interest) images (partial images) based on the mark position acquired in S802. FIG. 12 is a schematic diagram showing the positional relationship between the left ROI image and the right ROI image when detecting the top left vertical edge and the top right vertical edge. In this embodiment, a region 0.5 mm long horizontally and 1.7 mm long vertically is extracted as the left ROI image, centered on a point 1.0 mm left and 1.0 mm down from the center coordinates of the top left mark. Also, a region 0.5 mm long horizontally and 1.7 mm long vertically is extracted as the right ROI image, centered on a point 1.0 mm right and 1.0 mm down from the center coordinates of the top right mark. The method of acquiring the left and right ROIs when extracting the bottom left vertical edge and the bottom right vertical edge is the same.

[0048] In S1102, the CPU 114 applies an edge detection filter to each of the left and right ROI images acquired in S1101. In this embodiment, the following kernel k is applied as the edge detection filter.

[0049]

number

[0050] The edge detection filter is not limited to the kernel k. A horizontal Laplacian filter or a differential filter may be used. In the following, the image after the edge detection filter is applied is described as Ie(x,y). Note that x and y are coordinate indices.

[0051] In step S1103, a plurality of candidates for both paper edges are detected from the image after the edge detection filter is applied. The details of this process will be described later.

[0052] In step S1104, an optimal combination is determined from the multiple paper edge candidates. The details of this process will be described later.

[0053] <Paper edge candidate extraction process> The paper edge candidate extraction process performed in S1103 will be described in detail below with reference to the flowchart in FIG.

[0054] In S1301, the CPU 114 applies binarization processing to Ie(x,y) calculated in S1102 to obtain a binary image Ib(x,y). Specifically, Ib(x,y) can be calculated by the following formula.

[0055]

number

[0056] Here, th is a threshold value, which may be a fixed value of 128, for example, or may be dynamically determined from the image by Otsu's method.

[0057] In S1302, the CPU 114 performs a Hough transform on the acquired binary image Ib(x, y). The Hough transform is a process for comprehensively detecting straight lines that form a predetermined angle θ from within an image by expressing the straight line with the formula (ρ=xicosθ+yisinθ). The operation of the Hough transform will be described below.

[0058] In the Hough transform, first, obtain all pairs (xi, yi) of x and y coordinates such that Ib(x, y) = 1 (0≦i≦M). Next, calculate the ρij value for each pair (xi, yi).

[0059]

number

[0060] θj (0≦i≦N) is a predetermined value, and in this embodiment, θ is a value from 88° to 92° in 0.5° increments (θ0=88, θ1=88.5..., θ7=92.0). The obtained (θj, ρij) represents a straight line with angle θj that passes through point (xi, yi).

[0061] Next, the Hough table H(ρ, θ) is calculated by counting all combinations of the obtained (θj, ρij).

[0062]

number

[0063]

number

[0064] An example of a calculated Hough table is shown in Figure 14. The value of cell (θ, ρ) in the Hough table represents how many points in the binary image Ib(x, y) lie on the line ρ = xcosθ + ycosθ. For example, the value of cell (θ = 89°, ρ = 2) in Figure 14 is a large value of 219. This allows us to determine that the line defined by (θ = 89°, ρ = 2) exists in the image. On the other hand, the value of cell (θ = 92°, ρ = 2) is 0, which indicates that the line (θ = 92°, ρ = 2) does not exist in the image.

[0065] In S1303, the CPU 114 detects the peak position from the Hough table calculated in S1302. H(θ,ρ)≧H(θ,ρ+1) and H(θ,ρ)≧H(θ+1,ρ) and H(θ,ρ)≧H(θ-1,ρ) and H(θ,ρ)≧H(θ,ρ-1) and H(θ,ρ)>th_hough The point that satisfies the above is regarded as the peak value. Here, th_hough is a threshold value for determining whether the Hough table value is noise or not, and is set to 140 in this embodiment. For example, the peak values ​​detected for FIG. 14 are two candidates, (θ=90°, ρ=5) and (θ=89°, ρ=2).

[0066] As described above, in the paper edge candidate extraction process, multiple paper edge candidates can be extracted from one ROI image. FIG. 15 is a schematic diagram showing an example of detecting multiple paper edge candidates. FIG. 15(a) is a schematic diagram showing a scanned image and a right ROI image. In the scanned image, the paper surface is read at an angle due to the inclination during transportation. FIG. 15(b) is a schematic diagram showing a binary image Ib(x,y) when this process is applied to the upper right ROI image of the scanned image. A black pixel in the binary image Ib(x,y) indicates that no edge exists at the pixel position, and a white pixel indicates that an edge exists at the pixel position. As shown in FIG. 15(b), two paper edge candidates (multiple edges) are calculated for the scanned image. The edge on the left side of FIG. 15(b) is an edge formed by the shadow of the paper edge, and the edge corresponds to the paper edge candidate (θ=89°,ρ=2) in the Hough table of FIG. 14. Due to the tilt during scanning, the edge in question is tilted 1 degree from the vertical direction. Meanwhile, the paper edge candidate on the right side of Fig. 15(b) is a paper edge candidate that was erroneously detected due to a leader streak, and corresponds to the Hough table (θ=90°, ρ=5) in Fig. 14. Leader streaks are always detected as vertical streaks, regardless of the amount of tilt of the document. Multiple paper edge candidates detected by this process are narrowed down to one optimal pair (one set of edge pairs) by the optimal paper edge pair determination process described below.

[0067] <Optimal paper edge pair determination process> The optimal edge pair determination process in S1104 will be described in detail below with reference to the flowchart in FIG.

[0068] In S1601, the CPU 114 judges whether the processes from S1602 to S1605 have been completed for all the paper edge candidates. If not, the process proceeds to S1602, and if the processes have been completed, the process proceeds to S1605. Note that, for example, if three left paper edge candidates and five right paper edge candidates were extracted in S1103, the loop will be repeated a total of 15 times.

[0069] In S1602, the CPU 114 acquires candidates ρL and θL for the first paper edge. The first paper edge candidate is, for example, a candidate for the upper left vertical paper edge.

[0070] In S1603, the CPU 114 acquires the second paper edge candidates ρR and θR. The second paper edge candidate is, for example, the upper right vertical paper edge candidate.

[0071] In S1604, the likelihood of the paper edge pair is calculated taking into account the parallelism of the paper edge pair and the paper size. In this embodiment, the likelihood of the paper edge pair is calculated using the following formula. Note that the smaller the value in the formula, the more likely the paper edge pair is.

[0072]

number

[0073] Here, psize represents the standard value of the paper size. When detecting a vertical edge, the paper length in the sub-scanning direction is set, and when detecting a horizontal edge, the paper length in the main scanning direction is set. Additionally, dpi is the scan resolution. step(x) is a step function that returns 1 when x>0 and 0 when x≦0. Additionally, γ and ε are constants, with γ=1000 and ε=0.5.

[0074] The term (ρR-ρL)cosθL*25.4 / dpi represents the measured value of the paper size determined by the paper edge candidate pair, and the term |psize-(ρR-ρL)cosθL*25.4 / dpi| represents the error between the standard value and the measured value.

[0075] Also, γ step(|θR-θL|-ε) is a penalty term based on the parallelism of the paper. That is, when the difference between θR and θL is 0.5° or less, there is no penalty, and otherwise a penalty of 1000 is given. That is, if the absolute value of the angle difference is less than or equal to the threshold, there is no penalty. In this embodiment, the likelihood is calculated using the above formula, but the present invention is not limited to this. For example, the minimum value of the numerical value obtained by subtracting the length of the paper size in a specified direction from the distance between the paper edge candidates (distance between edges, difference in position information corresponding to the edges) may be set as the most likely paper edge. In this case, the paper edge candidate that is the innermost from the standard value of the paper size is selected as the most likely paper edge.

[0076] In S1605, the CPU 114 determines the paper edge pair with the best (smallest) likelihood as the most likely paper edge pair.

[0077] The specific operation of this process will be explained using the example in Fig. 17. The upper part of Fig. 17 illustrates the results of the paper edge candidate extraction process for the upper right paper edge and the upper left paper edge of the scanned image. For the upper right paper edge, three candidates are extracted: right candidate A (paper stain), right candidate B (paper edge), and right candidate C (leader streak). On the other hand, for the left paper edge, only one candidate is extracted: left candidate (1) (paper edge).

[0078] The table at the bottom of the figure shows the results of the calculation of the likelihood calculated by the optimal paper edge pair determination process and the determined paper edge pairs. The paper standard error in the table represents the value of the term |psize-(ρR-ρL)cosθL*25.4 / dpi| in formula (6), and the parallelism penalty represents the value of the term γ step(|θR-θL|-ε). The likelihood score represents the value defined by formula (6). In terms of the numerical value of the paper standard error, the left candidate (1) × right candidate C has the least error. However, the right candidate C is a leader streak and is not parallel to the left candidate (1), so it is not the optimal paper edge candidate from the perspective of parallelism. The right candidates A and B are parallel to the left candidate (1), and the "left candidate (1) × right candidate B", which is closest to the paper standard value, is determined as the optimal paper edge candidate pair.

[0079] <Horizontal paper edge detection process> Next, detection of a horizontal piece of paper, which is a paper side perpendicular to the vertical paper side, will be described. The process of detecting the horizontal paper side performed in S806 and S807 will be described in detail below with reference to the flowchart in FIG.

[0080] In S1801, the CPU 114 acquires the upper and lower ROI images based on the mark position acquired in S802. FIG. 12 is a schematic diagram showing the positional relationship between the upper ROI image and the lower ROI image when detecting the upper left horizontal edge and the lower left vertical edge. In this embodiment, an area 1.7 mm long horizontally and 0.5 mm long vertically is extracted as the upper ROI image, centered on a point 1.0 mm to the right and 1.0 mm up from the center coordinates of the upper left mark. Also, an area 1.7 mm long horizontally and 0.5 mm long vertically is extracted as the lower ROI image, centered on a point 1.0 mm to the right and 1.0 mm down from the center coordinates of the lower left mark. The method of acquiring the upper and lower ROIs when extracting the upper right horizontal edge and the lower right horizontal edge is the same.

[0081] In S1802, the CPU 114 applies an edge detection filter to each of the upper and lower ROI images acquired in S1801. In this embodiment, upper and lower differential filters are used as the edge detection filters, but the present invention is not limited to this. Other filters, such as a vertical Laplacian filter, may also be used.

[0082] In S1803, a plurality of candidates for both sides of the paper are detected from the image after the edge detection filter is applied. This process is similar to S1103.

[0083] In S1804, an optimal combination is determined from the multiple paper edge candidates.

[0084] The basic flow of this process is similar to that explained in S1104. The process is performed in the flow of S1601 to S1605 in FIG. 16. However, as mentioned above, the parallelism of the top and bottom sides of the image to be processed is not necessarily guaranteed due to the roller speed difference. In this case, the formula (6) assumed in S1604 above on the premise of parallelism does not hold. Therefore, when calculating the likelihood of a piece pair from the first paper edge candidates ρT and θT and the second paper edge candidates ρB and θB, the angle difference Δθ between the upper left vertical edge and the upper right vertical edge calculated in S804, and the lower left vertical edge and the lower right vertical edge calculated in S805 is used as a correction value. For example, when the upper angle θ1 calculated from the upper left and upper right and the lower angle θ2 calculated from the lower left and lower right are used, if the images are acquired ideally parallel, this angle difference will be 0. However, if there is a difference in these angles, for example if θ1 is 87° and θ2 is 90°, the parallelism between the top and bottom of the paper will be lost, and from the angle difference Δθ=3°, it can be estimated that the top and bottom edges are deviated by 3° from the horizontal. Therefore, formula (6) can be transformed as follows:

[0085]

number

[0086] Here, θave is the average angle calculated by (θT-θB) / 2.

[0087] By adding this angle correction value Δθ to the formula, the optimal combination of the upper and lower sides is determined.

[0088] The specific operation of this process will be explained using the example in Fig. 19. The upper part of Fig. 19 illustrates the results of the paper edge candidate extraction process for the upper left paper edge and the lower left paper edge of the scanned image. For the lower left paper edge, two candidates are extracted: lower candidate A (paper stain) and lower candidate B (paper edge). On the other hand, for the upper left paper edge, only one candidate is extracted: upper candidate (1) (paper edge).

[0089] The table at the bottom of the figure shows the results of the calculation of the likelihood calculated by the optimal paper edge pair determination process and the determined paper edge pairs. The paper standard error in the table represents the numerical value of the term |psize-(ρR-ρL)cosθave*25.4 / dpi| in formula (7), and the parallelism penalty represents the value of the term γ step(|θT-θB+Δθ|-ε). The likelihood score represents the value defined by formula (7). When the paper is fed in parallel, the upper candidate (1) × lower candidate A has the least error. However, there is a correction based on the value of Δθ calculated from the left and right sides, and it is found that it is not parallel to the upper candidate (1), so it is not the optimal paper edge candidate from the perspective of parallelism, and the lower candidate B is determined as the optimal paper edge candidate pair.

[0090] <Usage sequence> 7 is a sequence diagram showing a series of processes in this embodiment. In this embodiment, the main communication takes place between the operator and the image forming apparatus 100. Here, the process starts with the paper library editing screen 300 being displayed.

[0091] First, when the operator presses the print position adjustment button 323 in S701, the image forming apparatus 100 determines to start print position adjustment, and in S702, the image forming apparatus 100 displays a print position adjustment screen for specifying a paper feed cassette via the CPU 114.

[0092] Next, in S703, the operator designates a paper feed cassette and issues an instruction to perform print position adjustment processing.

[0093] Next, in S704, the image forming apparatus 100 outputs the adjustment chart illustrated in FIG.

[0094] Next, in S705, the operator sets the adjustment chart output in S704 in the image reading unit 130.

[0095] Next, in S706, the image forming apparatus 100 reads the adjustment chart set in the image reading unit 130 via the CPU 114, and executes the print position adjustment process.

[0096] Next, in S707, the image forming apparatus 100 performs print position adjustment via the CPU 114, and executes the print position adjustment process S802 to S808, thereby storing the print position deviation amount for each paper feed cassette in the paper library.

[0097] In this embodiment, the print position adjustment is performed as described above. Then, using the registered print position deviation amount, image formation is performed as follows.

[0098] Next, in S708, the user instructs the host computer 101 to execute a print job.

[0099] Next, in step S709 , the host computer 101 transmits the print job to the image forming apparatus 100 .

[0100] Next, in S710, the image forming apparatus 100 executes print job 1. At this time, the print position deviation amount registered for the paper feed cassette is read from the paper library and applied to the execution of print job 1.

[0101] Next, in S711, the image forming apparatus 100 provides a result generated by the execution of the print job.

[0102] <Notes> By the above process, the operator can calculate the amount of print position deviation for each selected paper feed cassette simply by scanning the chart, making it possible to adjust the front and back print positions.

[0103] The method for detecting multiple paper edge candidates and calculating the most likely paper edge candidate pair from parallelism and paper size has been described above. According to this embodiment, the relative position of the mark can be robustly detected even in a situation where the brightness of the leader background and the paper white are close, or where there is a lot of noise such as leader streaks. By adjusting the print position so as to cancel the calculated relative position of the mark, accurate front and back position adjustment is possible.

[0104] Furthermore, even if the parallelism of the leading and trailing ends of the paper is lost due to uneven paper transport, the relative position of the mark can be robustly detected by using the angle difference obtained when determining the left and right sides of the paper as a correction value.

[0105] In this embodiment, the left and right ROI images and the top and bottom ROI images shown in Fig. 12 are described as independent images that do not overlap, but this is not necessarily limited to this. It is also possible to cut out an image of a wider range, for example, an image of the four corners of the paper including the marker, and use it for processing as an ROI image common to the vertical and horizontal paper sides.

[0106] Example 2 In the first embodiment, multiple candidates for each of the opposing paper edges are detected by the Hough transform, and the optimal pair is selected based on the paper size and parallelism. However, while the Hough transform can detect multiple candidates, it has the problem that the accuracy of paper edge detection is not high. Therefore, in this embodiment, after detecting the optimal paper edge pair, more detailed paper edge estimation is performed to perform more accurate paper edge detection. The configuration of the printing system in the second embodiment is substantially the same as that in the first embodiment. Therefore, the same components are denoted by the same reference numerals, and detailed descriptions thereof will be omitted. Below, the differences from the first embodiment will be described.

[0107] <Vertical paper edge detection process> The process of detecting the vertical paper edge performed in S804 and S805 will be described in detail below with reference to the flowchart in FIG.

[0108] In S2001, the CPU 114 acquires left and right ROI images based on the mark positions acquired in S802. This process is the same as that in the first embodiment.

[0109] In S2002, the CPU 114 applies an edge detection filter to each of the left and right ROI images acquired in S2001. This process is the same as in the first embodiment.

[0110] In S2003, a plurality of candidates for both paper edges are detected from the image after the edge detection filter is applied. This process is the same as that in the first embodiment.

[0111] In step S2004, an optimal combination is determined from the multiple paper edge candidates. This process is the same as in the first embodiment.

[0112] In S2005, a more detailed paper-side expression is determined based on the first paper-side expression obtained in S2004. The details of this process will be described later.

[0113] In S2006, a more detailed paper-side expression is determined based on the second paper-side expression obtained in S2004. The details of this process will be described later.

[0114] <Detailed paper edge estimation processing> The detailed edge estimation process performed in S2005 and S2006 will be described in detail below with reference to the flowchart of FIG.

[0115] In S2101, the CPU 114 acquires the coefficients (ρ, θ) of the paper side formula calculated in S2004 and the image Ie(x, y) after application of the edge detection filter calculated in S1102. In the following, the coefficients are updated and more detailed coefficients (ρ', θ') are calculated.

[0116] In S2102, the CPU 114 judges whether the following processing has been performed for all y values ​​of the ROI image. If not, the process proceeds to S2103, and if completed, the process proceeds to S2105. For example, if the vertical size of Ie(x, y) is 200 pixels, the processing of S2103 to S2105 is repeated 200 times.

[0117] In S2103, the CPU 114 obtains the edge position x of the provisional edge equation at the height y. Specifically, it is calculated as x=(ρ-y sin θ) / cos θ.

[0118] In S2104, the CPU 114 calculates the position x* at which the edge detection filter response value becomes maximum within ±4 pixels (surrounding pixels) around the position x. Specifically, x* is calculated by the following linear approximation formula.

[0119]

number

[0120] In S2105, the CPU 114 stores the calculated edge position (x*, y) in the RAM 113.

[0121] In S2106, the CPU 114 performs least squares linear approximation on the calculated pairs (x*, y) of edge positions, thereby calculating (ρ', θ') with higher accuracy than (ρ, θ) calculated in S2004.

[0122] <Horizontal paper edge detection process> As in the vertical paper edge detection process, after determining a paper edge candidate pair based on the Hough transform, the paper edge detailed estimation process is executed. Specifically, after executing the processes of S1801 to S1804, S2005 and S2006 are executed.

[0123] <Notes> As described above, in this embodiment, after the optimal paper edge pair is detected, a more detailed paper edge estimation is performed. According to this embodiment, it is possible to perform a highly accurate paper edge estimation while taking into consideration the parallelism of the paper edges and the consistency with the paper size.

[0124] Example 3 In the first and second embodiments, the optimal edge pair is determined from one image based on the standard value of the paper size. However, due to variations in the speed of the transport roller and errors when cutting the paper, the paper size on the scanned image may not match the standard value of the paper size. In the above embodiment, the optimal edge pair is detected based on the paper size, and if there is a discrepancy between the paper size on the scanned image and the standard value, the registration adjustment may not be performed normally. Therefore, in this embodiment, multiple sheets are printed and scanned for registration adjustment, and the most likely paper size is estimated, and then the paper edge is detected. This allows robust position correction to be performed even if there is a difference between the standard value of the paper size and the paper size of the scanned image. Below, only the differences from the first embodiment are described.

[0125] <Print position measurement unit operation> The operation of the print position measurement unit will be described below with reference to the flowchart in FIG. 22 and FIG.

[0126] In S2201, the CPU 114 acquires a read image of the adjustment chart via the image reading unit 130. Note that the image reading unit 130 may be an external reading device such as an ADF, or may be a reading device installed inside the printer, such as an in-line sensor.

[0127] In S2202, the CPU 114 acquires the positions of the marks 602 to 605 from the read image. The mark positions are expressed as two-dimensional coordinates in the horizontal (x) and vertical (y) directions of the scanned image, with the upper left coordinate set as the origin (0,0). The mark positions may be expressed as the coordinates of the center of gravity of the mark, or may be calculated with sub-pixel accuracy. Furthermore, the detection process of the mark may be performed by a pattern matching method.

[0128] In S2203, the CPU 114 acquires an ROI image (partial image) based on the acquired mark position. This process is the same as S1101, and the top, bottom, left and right ROI images of all four corners are acquired.

[0129] In S2204, the CPU 114 applies an edge detection filter to each of the acquired ROI images. This process is similar to that in S1102.

[0130] In S2205, the CPU 114 performs a paper edge extraction process on the result of S2204. This process is the same as S1103. In this step, a Hough table and paper edge candidates are extracted for each of the eight ROI images.

[0131] In S2206, the CPU 114 determines whether the processes of S2201 to S2205 have been completed for all documents. If completed, the process proceeds to S2207, and if not, the process returns to S2201.

[0132] In S2207, the paper size of the document is estimated from the multiple paper edge candidates and the Hough table acquired in S2205. The details of this process will be described later.

[0133] In S2208, the CPU 114 detects the upper left vertical edge and the upper right vertical edge. This process is the same as S804. However, the paper size is not a standard value, but the one estimated in S2207 is used.

[0134] In S2209, the CPU 114 detects the lower left vertical edge and the lower right vertical edge. This process is the same as S805. However, the paper size is not a standard value, but the one estimated in S2207 is used.

[0135] In S2210, the CPU 114 detects the upper left and lower left sides of the paper. This process is the same as S806. However, the paper size is not a standard value, but is estimated in S2207.

[0136] In S2211, the CPU 114 detects the upper right horizontal edge and the lower right horizontal edge of the paper. This process is the same as S807. However, the paper size is not a standard value, but is estimated in S2207.

[0137] In S2212, the CPU 114 determines whether the processes of S2208 to S2211 have been completed for all documents. If completed, the process proceeds to S2213, and if not, the process returns to S2208.

[0138] In S2213, CPU 114 calculates the relative mark position from the eight paper sides detected for the multiple scanned images. In this process, the normal distance from the marker center coordinates to the linear equation (ρ=xcosθ+ysinθ) of each paper side is calculated to calculate distances (A) to (H) in FIG. 6 for each scanned image. Furthermore, the average value of distances (A) to (H) calculated for each scanned image is taken to calculate the average relative mark position, which is used as the registration correction value.

[0139] <Paper size estimation process> The paper size estimation process in S2207 will be described in detail below with reference to the flowchart in Fig. 23. Here, calculation of the horizontal paper size will be described as an example, but estimation of the vertical paper size is performed in the same way. In this case, right and left are replaced with top and bottom, and the process is executed.

[0140] In S2301, CPU 114 initializes the numerical values ​​of the histogram value storage memory to 0. The upper part of Fig. 24 shows the histogram value storage memory initialized by this process. As shown in Fig. 24, in this embodiment, 10 arrays are secured to construct a histogram that divides 209.5 mm to 210.4 mm in increments of 0.1 mm. Each array stores the occurrence frequency of the estimated value of the corresponding paper size, but 0 is stored by this initialization process.

[0141] In S2302, the CPU 114 obtains the right edge candidates ρR and θR obtained in the paper edge candidate extraction process in S2205.

[0142] In S2303, the CPU 114 obtains the left edge candidates ρL and θL obtained in the paper edge candidate extraction process in S2205.

[0143] In S2304, the CPU 114 calculates an estimated value (estimated result) of the paper size using the left and right edge candidates (edge ​​information). Specifically, the calculation is performed using the following formula.

[0144]

number

[0145] In S2305, the CPU 114 updates the corresponding histogram value based on the estimated value. For example, if the estimated value of the paper size calculated in S2304 is 209.8 mm, 1 is added to the value of the fourth histogram bin from the left in FIG. 24. Note that, in this process, weighting may be performed according to the strength of the edge candidate, and the histogram value may be updated. For example, the accuracy of the paper candidate can be reflected in the histogram by adding the value of the Hough table corresponding to the edge candidate to the corresponding histogram bin.

[0146] In S2306, the CPU 114 determines whether the processes of S2302 to S2305 have been executed for all the left edge candidates. If completed, the process proceeds to S2307, and if not, the process returns to S2303.

[0147] In S2307, the CPU 114 determines whether the processes of S2302 to S2306 have been executed for all the right edge candidates. If completed, the process proceeds to S2308, and if not, the process returns to S2302.

[0148] In S2308, the CPU 114 judges whether the processes of S2302 to S2307 have been completed for all manuscripts. That is, it judges whether the estimation results of all manuscripts have been tallied. If completed, the process proceeds to S2309, and if not, the process returns to S2302.

[0149] In S2309, the CPU 114 adopts the paper size corresponding to the most frequent value of the histogram as an estimated value of the paper size representative of the plurality of sheets. Figure 25 is a graph of the updated histogram values ​​of Figure 24. Since the most frequent value of the histogram is 76, the estimated paper size is 209.9 mm.

[0150] <Usage sequence> 26 is a sequence showing a series of processes in this embodiment. In this embodiment, the main communication is between the operator and the image forming apparatus 100. Here, the process starts with the paper library editing screen 300 displayed. In this embodiment, unlike the first embodiment, multiple documents are read and the amount of print position deviation is adjusted.

[0151] First, in S2601, when the operator presses the print position adjustment button 323, the image forming apparatus 100 determines to start print position adjustment. Then, in S2602, the image forming apparatus 100 displays, via the CPU 114, a print position adjustment screen for specifying a paper feed cassette and the number of copies to be printed.

[0152] Next, in S2603, the operator specifies the paper feed cassette and the number of copies to be printed, and issues a command for print position adjustment processing. At this time, the image forming apparatus 100 displays a screen for specifying the number of copies to be printed of the print adjustment chart on the operation panel 120 via the CPU 114. The user specifies the desired number of copies on the operation panel 120, and issues a print command.

[0153] Next, in S2604, the image forming apparatus 100 outputs the adjustment chart exemplified in FIG.

[0154] Next, in S2605, the operator sets the predetermined number of adjustment charts output in S704 in the image reading unit 130.

[0155] Next, in S2606, the image forming apparatus 100 reads a plurality of adjustment charts set in the image reading unit 130 via the CPU 114, and executes the paper edge candidate extraction process S2201 to 2206.

[0156] Next, in step S2607, the image forming apparatus 100, via the CPU 114, estimates the paper size for each paper feed cassette.

[0157] Next, in S2608, the image forming apparatus 100 executes the print position adjustment process S2207 to S2211 via the CPU 114, and the amount of print position deviation for each read image is stored.

[0158] Next, in step S2609, the image forming apparatus 100 stores, via the CPU 114, the amount of print correction calculated by averaging the amount of print position deviation for each read image.

[0159] In this embodiment, the print position adjustment is performed as described above. Then, using the registered print position deviation amount, image formation is performed as follows.

[0160] Next, in S2610, the user instructs the host computer 101 to execute a print job.

[0161] Next, in step S 2611 , the host computer 101 transmits the print job to the image forming apparatus 100 .

[0162] Next, in S2612, the image forming apparatus 100 executes print job 1. At this time, the print position deviation amount registered for the paper feed cassette is read from the paper library and applied to the execution of print job 1.

[0163] Next, in S2613, image forming apparatus 100 provides a result generated by the execution of the print job.

[0164] <Notes> According to this embodiment, even if the standard value of the paper size and the read image size of the document differ due to unevenness in the transport speed during image reading, moisture absorption, cutting errors, etc., it is possible to search for paper edge pairs more accurately.

[0165] Example 4 In the first and second embodiments, a method for selecting an optimal paper edge pair by estimating the angle of the parallelism of the upper and lower sides of an adjustment chart that is not parallel is described. It is known that the longer the paper transport length, the greater the deviation in the angle parallelism. The degree of parallelism also differs depending on the weight (basis weight) of the paper. Therefore, in this embodiment, an example is described in which parameters are dynamically changed based on how the parallelism of the upper and lower sides is not parallelized according to the information in the paper library. The configuration of the printing system in the third embodiment is substantially the same as that in the first embodiment. Therefore, the same components are denoted by the same reference numerals, and detailed description thereof is omitted. Below, a detailed description is given of the means for determining the paper edge pair, which is a difference from the first and second embodiments.

[0166] First, the distortion coefficient used when determining the paper edge pair will be described. As described above, a list in which various information about the paper used for printing is stored is prepared as the paper library. In list 310 in FIG. 3, the papers stored in the paper library are listed here. In paper list 310, the paper attributes as shown in columns 311 to 316 have accompanying information for each paper. Columns 312 and 313 respectively indicate the paper length in the sub-scanning direction and the paper length in the main scanning direction for each paper. Column 314 indicates the basis weight for each paper.

[0167] From this paper information, the amount of skew and distortion of the image caused by transport when reading using the ADF is estimated based on the paper length and basis weight, and this estimated value is stored as a distortion coefficient. The larger the basis weight and the longer the transport length of the paper, the larger this distortion coefficient becomes. This distortion estimate is obtained for each sheet of paper. In this case, since the transport length in the ADF is used, the sub-scanning direction during printing is not necessarily the transport length to the ADF. The distortion coefficient is used to estimate the paper edge pair.

[0168] The process of detecting the vertical paper edge performed in S804 and S805 using FIG. 8 is a process following the flowchart of FIG. 11 used in the first embodiment, and is the same as in the first embodiment, so a description thereof will be omitted.

[0169] Next, the process of detecting the horizontal paper edge performed in S806 and S807 will be described with reference to the flowchart of FIG. 18 used in the description of the first embodiment.

[0170] In S1801, the CPU 114 acquires upper and lower ROI images based on the mark position acquired in S802. The method is the same as that described in the first embodiment. The method of acquiring the upper and lower ROIs when extracting the upper right horizontal edge and the lower right horizontal edge is also the same.

[0171] In S1802, the CPU 114 applies an edge detection filter to each of the upper and lower ROI images acquired in S1801. In this embodiment, upper and lower differential filters are used as edge detection filters, but the present invention is not limited to this. Other filters such as a vertical Laplacian filter may also be used.

[0172] In S1803, a plurality of candidates for both sides of the paper are detected from the image after the edge detection filter is applied. This process is similar to S1103 described in the first embodiment.

[0173] In S1804, an optimal combination is determined from the multiple paper edge candidates. The basic flow of this process is similar to that described in S1104 in the first embodiment, but a correction value is given not only based on the roller speed difference as described in the first embodiment, but also based on the paper distortion coefficient described above.

[0174] For this distortion coefficient, ε in the above-mentioned formulas (6) and (7) is used as a variable according to the distortion coefficient, not as a constant. If the distortion coefficient is d, then it is transformed as follows:

[0175]

number

[0176] As explained in the first embodiment, ε is a parameter for the penalty term based on the parallelism of the paper. In other words, when the difference between θT and θB is equal to or less than ε, there is no penalty, and otherwise a penalty of 1000 is applied. The threshold value for generating this penalty is dynamically switched according to the distortion coefficient. For paper with a large distortion coefficient (where the distortion tends to be read as large), the value of ε is set large, for example 1°, and the allowable range of the angle swing is increased. Conversely, for paper with small distortion, a more strict value of ε is set, for example 0.5°.

[0177] By adding this ε to the equation as a variable based on the distortion coefficient, the optimum combination of the four sides according to the printing paper is determined.

[0178] <Notes> According to this embodiment, it is possible to search for paper edge pairs more accurately in conjunction with the magnitude of the distortion width due to the characteristics of the paper, and it is possible to prevent pairs from going undetected even for paper that is likely to be read with a large distortion.

[0179] In this embodiment, the ADF transport length and basis weight of the paper are used as information for calculating the distortion coefficient, but this is not limited to this. If there is any information that gives some kind of image distortion to the ADF, that information may be added.

[0180] (Other Examples) In the above embodiment, the mark relative position is calculated based on the Hough transform and edge detection processing, but this may be realized by machine learning. For example, as shown in FIG. 9(a), a SVR (Support Vector Regression) may be constructed in which the Hough table and paper size are input and the left and right (ρ, θ) are output. In this case, the learning data is obtained by performing the Hough transform of S1202 on the previously acquired read image, and the correct answer data is the left and right (ρ, θ) value tagged manually. Also, as shown in FIG. 9(b), a model may be constructed in which the image data and paper size are input and the paper edge position is output using a CNN (Convolutional Neural Network). When learning the model, the number of learning data may be increased by image processing by randomly applying minute magnification and minute angle rotation to the scanned image.

[0181] The present invention can also be realized by a process in which a program for implementing one or more of the functions of the above-mentioned embodiments is supplied to a system or device via a network or a storage medium, and one or more processors in a computer of the system or device read and execute the program. It can also be realized by a circuit (e.g., ASIC) that implements one or more of the functions.

[0182] The present invention may be applied to a system composed of a plurality of devices, or to an apparatus composed of a single device. For example, a device having an image reading unit 130 and a device having an image forming unit 150 may be separate devices, and these devices may be linked by communication. Also, a device performing image formation and a device performing image processing may be separate devices, and these devices may be linked by communication.

[0183] The present invention is not limited to the above-mentioned embodiments, and various modifications (including organic combinations of the embodiments) are possible based on the spirit of the present invention, and are not excluded from the scope of the present invention. In other words, the present invention includes all configurations that combine the above-mentioned embodiments and their modifications.

[0184] The definitions of the abbreviations used in each example are as follows:

[0185] ADF stands for Auto Document Feeder. ASIC stands for Application Specific Integrated Circuit. CIS stands for Contact Image Sensor. CNN stands for Convolutional Neural Network. CPU stands for Central Processing Unit. CSV stands for Comma-Separated Values. RAM stands for Random-Access Memory. ROI stands for Region Of Interest. ROM stands for Read Only Memory. HDD stands for Hard Disk Drive. RIP stands for Raster image processor. DPI stands for Dots per inch. SVR stands for Support Vector Regression. XML stands for Extensible Markup Language. [Explanation of symbols]

[0186] 100 Image forming device 130 Image reading unit 150 Image forming section 235 Second image reading unit 236 First image reading unit

Claims

1. A selection means for selecting sheet information; a printing means for printing an image having a predetermined pattern including at least one mark on a sheet corresponding to the sheet information selected by the selection means; a reading means for reading the image from the sheet; a detection means for detecting a plurality of candidates for the edge of the sheet from the image read by the reading means; a determining means for determining a set of edges of the sheet from among the plurality of candidates based on the plurality of candidates detected by the detecting means and sheet size information included in the sheet information selected by the selecting means; means for obtaining a parameter based at least on position information of the at least one mark and position information corresponding to one edge of the set of edges determined by the determining means; an adjustment unit that adjusts a printing position by the printing unit based on the parameter, The image processing device is characterized in that the set of edges is a set of edges among the multiple edges, the absolute value of the angle difference being less than a threshold value, and the distance between the edges and the length of the sheet in a specified direction satisfy a specified relationship.

2. 2. The image processing apparatus according to claim 1, further comprising a calculation means for performing a filter calculation on a partial image of the image.

3. 3. The image processing apparatus according to claim 1, wherein the detection means detects the plurality of candidates using a Hough transform.

4. 4. The image processing device according to claim 1, further comprising an approximation unit that identifies a pixel position in the peripheral pixels of the edge set at which a response of a detection process for detecting the edge is maximum, and performs a first-order approximation on the identified pixel position.

5. 5. The image processing apparatus according to claim 1, wherein the set of edges is a set of edges along a conveying direction among a plurality of edges of the sheet.

6. 6. The image processing apparatus according to claim 1, wherein the reading means comprises a mechanism for conveying a sheet, and a CIS for reading an image from the conveyed sheet.

7. 7. The image processing apparatus according to claim 6, wherein the reading means has a white member disposed opposite the CIS.

8. A control method for an image processing device, comprising: A selection step of selecting sheet information; a printing step of printing an image having a predetermined pattern including at least one mark on a sheet corresponding to the sheet information selected in the selection step; a reading step of reading the image from the sheet; a detection step of detecting a plurality of candidates for the edge of the sheet from the image read in the reading step; a determination step of determining a set of edges of the sheet from among the plurality of candidates based on the plurality of candidates detected in the detection step and sheet size information included in the sheet information selected in the selection step; obtaining a parameter based at least on position information of the at least one mark and position information corresponding to one edge of the set of edges determined in the determining step; and adjusting the print position based on the parameter, A control method for an image processing device, characterized in that the set of edges is a set of edges among the multiple edges, the absolute value of the angle difference being below a threshold value, and the set of edges is a set of edges where the distance between the edges and the length of the sheet in a specified direction satisfy a specified relationship.

9. A program for causing a computer to execute the control method according to claim 8.

Images