JPH05314298A - Data reader - Google Patents

Abstract

PURPOSE:To surely reproduce data in the case of short-time processing to the image data of mesh-shaped patterns by simultaneously deciding the positions of pixel data expressing respective meshes based on geometrical regularity. CONSTITUTION:A line type image sensor 11 is moved at fixed speed in a sub scanning direction on a recording medium where images having the mesh-shaped patterns are recorded. The image data of images obtained by the image sensor 11 do not practically damage the distance accuracy of source images but keeps the geometrical regularity. A CPU 15 detects a data start mark out of the image data impressed from the image sensor 11 and calculates the central positions of the respective meshes in the image data of following mesh-shaped patterns from the detected position while utilizing the regularity of the image data. Then, the CPU 15 identifies the brightness/darkness of the respective meshes by sampling image bits at the calculated central positions of the respective meshes from a series of following main scanning line image data.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data reading device which takes in an encoded image recorded on a recording medium such as paper and reproduces encoded data.

[0002]

Background Bar code technology is known as a typical image coding technology. However, the bar code has a problem that the recording density cannot be increased due to its structure. Therefore, the applicant of the present invention has proposed a new image coding technique in which a mesh pattern expressing information units (bits) by the light and shade formed in a mesh arranged vertically and horizontally is used as a coded image (Japanese Patent Application No. 63-328028).

According to this, since the two-dimensional surface on the recording medium is filled with the mesh of the mesh pattern, the recording density can be improved as much as possible. The problem is that in reproducing data, the captured encoded image must be two-dimensionally searched to identify the lightness and darkness of each mesh.

For this purpose, the applicant of the present invention adds a sampling reference mark which covers the entire scanning length of the mesh pattern to the mesh pattern in a use environment in which the encoded image is distorted and captured, and the data reading device By utilizing the linearity or uniformity of the distance that is approximately established in the local area of the captured image data, the position of each sampling reference mark is tracked and detected, and the local position information is used to detect the local position. We have developed a data reproduction technology that decodes the light and darkness of each mesh of a mesh pattern by finding the position of the mesh within a certain range and sampling the image bits there. Although this data reproducing technique has a data reproducing ability that can be practically used, there is a problem in that virtually all image data needs to be accessed, the amount of data processing is large, and the processing takes time.

[0005]

SUMMARY OF THE INVENTION It is, therefore, an object of the present invention to provide a data reading device capable of surely reproducing data of image data of a mesh pattern in a relatively short time.

[0006]

In order to achieve the above object, the data reading apparatus according to the present invention comprises a recording medium on which an image of a mesh pattern in which data is encoded by the light and dark selectively formed in each mesh is recorded. Image sensor means for reading the image data representing the image without substantially impairing the distance accuracy, and pixel data representing each mesh in the image data based on the overall geometric regularity of the image data. It is characterized by comprising a mesh position determining means for collectively determining the positions of the above, and a mesh light / dark identifying means for identifying the light / dark of each of the above meshes based on the pixel data at each determined position.

According to this structure, the image sensor means reads the image of the mesh pattern on the recording medium as image data without substantially impairing the distance accuracy, so that the image data as a whole has some geometric rules. Will save the sex. Therefore, the position of the pixel data representing each mesh on the image data is collectively determined by utilizing this property. As a result, the processing time can be shortened, the mesh position can be measured with less error, and the bright and dark of the mesh can be accurately identified. The image sensor means can be realized by, for example, a line-type image sensor that moves mechanically at a constant scanning speed and scanning direction.

[0008]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows the overall configuration of a data reading device 10 according to this embodiment. The overall purpose of the apparatus 10 is to capture an image having a mesh pattern 21 as illustrated in FIG. 2 by the image sensor 11, decode the data encoded in the image, and reproduce it.

The image sensor 11 is, for example, a line type image sensor including a photoelectric conversion array such as a CCD device, and a sensor feeding motor 13 controlled by a control circuit 12 causes a coded image as shown in FIG. The recording medium 20 on which 20 is recorded is scanned (moved) at a constant speed in a fixed direction (sub-scanning direction).
The image data of the image 20 that is photoelectrically converted by the image sensor 11 and binarized by such mechanical scanning does not substantially impair the distance accuracy of the original image 20 and maintains its geometric regularity. It becomes a thing. The control circuit 12 controls the circuits related to the image sensor 11 and performs serial / parallel conversion of image data sent in series from the image sensor 11 to perform an interface function for the CPU 15.

The CPU 15 operates according to a program stored in the ROM 16. During scanning of the encoded image 20, the image data provided from the image sensor 11 via the control circuit 12 is used to generate a mesh pattern 21. An image bit in the approximate center of each mesh is selected and used as a discrimination result of the lightness and darkness of each mesh in the RAM
Store at 17. That is, as shown in FIG. 2, there are spaces on the recording medium 14 in the vertical direction of the mesh pattern 22 (48 × 72 mesh matrix configuration) which is the data body of the encoded image 20, and the left side of FIG. The data start mark 22 composed of a black bar is formed at two positions which are scanned prior to the net-like pattern 22 by the line-type image sensor 11 which moves in the lateral direction of the net-like pattern 22 from right to right. , The start position (upper left or lower left corner) of these two data start marks 22 is CP
It is detected by U15 before the arrival of the mesh pattern 22.
The positions of these two data start marks 22 are the recording medium 14
Has a predetermined positional relationship with the mesh pattern 22 above,
Therefore, it has a predetermined positional relationship with the center of each mesh 23.
Since the movement of the image sensor 11 is controlled at a constant speed along the constant scanning direction by the motor 13 or the like as described above, it exists in the image data in which the same regularity as that on the recording medium 14 is captured. By utilizing this regularity, the center position of each mesh 23 in the image data of the coming mesh pattern 21 can be calculated from the detected positions of the two data start marks. So CPU
Reference numeral 15 samples the image bit at the center position of each mesh 23 that has been calculated from the subsequent series of main scanning line image data and stores it in the RAM 17 to determine the bit that is an information unit depending on the brightness of each mesh 23. The encoded data of the expressed mesh pattern 22 is decoded and reproduced.

In this way, in the data reading apparatus 10 of the present embodiment, the image sensor 11 scans the encoded image 20 while identifying the brightness of each mesh on the scanned image data. Therefore, the decoding process can be performed at an extremely high speed, an image memory for storing the scanned image data is not substantially required, and a large storage capacity can be saved.

The operation of the embodiment will be described in more detail below with reference to the flow chart shown in FIG. First, at the initialization of 3-1 check flag, main scanning line counter, RAM1
Internal settings such as 7 addresses and I / O for the control circuit 12
With the O setting, scanning of the encoded image 20 is started. The position of the first mark is detected by the following mark detection loops 3-2 and 3-3. That is, the main scanning line counter is updated at every predetermined period (main scanning line capturing period) to capture the main scanning image data for one line from the image sensor 11, and the fragment of the data start mark 22 (mark It is checked whether there is a black pixel indicating the upper left or lower left of 22). As a result, in FIG. 4, the point P1 on the main scanning line 41 is detected as the position of the first data start mark 22. When the position of the first mark is detected, its position P1 (the line number indicated by the main scanning line count at the time of detection, that is, the X component and the dot number indicated by the image dot count from the end of the main scanning line image, that is, the Y component) is detected. (Specified by and) is stored (3-4), and the work of detecting the position of another data start mark 22 is continued.

In the case of FIG. 4, the second data start mark 22 is detected at the point P2 of the subsequent main scanning line 42. Normally, the main scanning direction and the width direction of the net-like pattern 21 are slightly different (the difference is indicated by an angle θ in FIG. 4).
Two marks are observed on two different main scanning line images, but when they are completely parallel (θ = 0 degree), they are observed at two dot positions with the same spacing on the same main scanning line image. It Position P1 of the two marks 22,
When P2 is detected, both mark detection checks 3-5 are established. From these two detection positions P1 and P2, the angle θ of the inclination 44 of the image data and the main scanning width 45 (see FIG. 4), which will be the reference for the subsequent sampling, can be known, so these parameters are calculated in 3-6. If this main scanning width is divided by the number of meshes in the main scanning direction, here 48, the length U = in the vertical direction (as viewed in FIG. 2) of the mesh seen from the main scanning direction is obtained, and the vector of the main scanning width ( The 48 equal points of the line image) represent the sampling position (Y component) in the main scanning direction. On the other hand, the sampling position (X component) in the sub-scanning direction is obtained from the aspect ratio of the mesh on the image data. That is, the line resolution of the image sensor 11 in the main scanning direction is Ddot / mm, the moving (scanning) speed in the sub scanning direction is 5 mm / sec, and the main scanning cycle time (interval or frequency for capturing main scanning line image).
Is KHz (these parameters are determined by the system), the relative size r in the sub-scanning direction with respect to the unit length 1 in the main scanning direction is given by r = S × D / K. The horizontal length U _x of the mesh is given by (main scanning width / 48) × r (= U _y × r). In other words, the center position P of each mesh on the image data is P = (MU _x + NU _y t _an θ, NU _y −MY _x t _an θ) from the geometrical regularity stored in the image data. (Where M and N are integers, and the center of the mesh (M = 0, N = 0) in the upper left corner of the mesh pattern is indicated by the origin).

Here, the mesh length components U _x and U _y are calculated as real numbers, but the actual sampling from the image data is in a discrete form in which number of main scanning line image and what number of dot. Since it can be done only by the above, the integer coordinate closest to the real number coordinate P plus the actual position on the image data of the mesh of M = 0 and N = 0 (to be exact, the real number position) is the actual sampling position. (S with a number in Fig. 5)
Becomes

Thus, at 3-7, the sampling position of the first (M = 0) bit row is calculated, 3-8 to 3-1.
Preparation for the decoding process of 1 is completed. In the decoding loop, the main scanning line image is extracted (3-8), the sampling position included in the main scanning line image is checked (3-9), and the main scanning line image having no sampling position is detected. If it is skipped (NO in 3-9 and returns to 3-8), and if sampling positions are included, the image bits at each sampling position are sampled as information bits indicating the brightness of the mesh and stored in the RAM 17.
The sampling position of the next mesh is calculated (3-10).
For example, in the case of FIG. 4, the line image up to the main scanning line 43 which meets the sampling position of the first bit row is skipped. By such sequential scanning / decoding, R
Store for AM17 is defined by image format 4
Decoding completion check when 8 × 72 bits are completed 3-
11 is established, the end of scanning is instructed to the image sensor 11 (3-12), and the scanning / decoding process according to the embodiment is completed.

As described above, according to the present embodiment, (a) the data is decoded in parallel with the image scanning and the processing speed is extremely high. (B) For the same reason,
No image memory is required, and the storage capacity of the data reading device 15 can be greatly reduced. (C) If there is a simple data start mark, the regularity of the image data is used to detect all mesh positions from the detection position. This has the advantage that it can be determined and does not require a special sampling reference mark that extends over the entire length of the mesh pattern in the scanning direction.

[0017]

[Modification] Although the description of the embodiment has been completed, various modifications and changes can be made within the scope of the present invention. For example, instead of simultaneously executing the scanning of the encoded image and the decoding of the image data in parallel, the decoding of the image data may be performed after the scanning of the encoded image is completed.

Such a modification (first modification) will be described with reference to FIGS. 6 to 12. FIG. 6 is a flowchart of a process for identifying the lightness and darkness of each mesh from the image data of the mesh pattern 21 loaded in the image RAM 70 as shown in FIG. In the first routine 6-1, one to a plurality of main scanning line images are examined to measure the vertical length of the mesh pattern 21 and the vertical length of the mesh as viewed from the main scanning direction. This process is performed when the coded image 20 has no marks other than the mesh pattern 21 and when the mesh pattern 21 has no mark.
The processing contents are slightly different from the case where a contour line or the like (the thickness is, for example, equal to the width of the mesh) that forms the contour of the mesh pattern is provided around the. If there is a contour line, any main scanning line image except for the margins of the image contains black pixels that are fragments of the contour line, so check the image dot value from both ends of the line image toward the center. When a black image element (which will be referred to as a candidate point) is hit, it may be considered as an edge of the contour line, and the distance between them gives a measure of the width of the mesh pattern in the main scanning direction. On the other hand, in the case of only the mesh pattern, the similar candidate point is the edge of the outermost black mesh in the mesh pattern 21 in the main scanning line image.
If the outermost mesh of is white, that mesh is skipped.

However, in the halftone recording (printing) device of the coded image, the process of suppressing the DC component of the light and dark of the mesh (the length of continuous meshes of the same lightness) can be performed (by digital modulation or random number processing for light and dark). ) Is applied, the line image having a black mesh on the outermost side is hit by examining a certain number or more of main scanning lines. Therefore,
To select the longest candidate point pair (excluded because it is an error that is extremely long compared to the others) among the candidate point pairs obtained on a plurality of main scanning line images as indicated by l, m, and n in FIG. Thus, the vertical length of the mesh pattern viewed from the main scanning direction can be evaluated. In the following description, it is assumed that the mesh pattern 21 has a contour line, but if the processing is slightly complicated as in this example, it can be applied to a coded image of only the mesh pattern 21. In addition, in the case of a contour line, the net-like pattern 2 seen from the main scanning direction is obtained only by examining one main scanning line image.
Although the vertical length of 1 can be measured, in consideration of the possibility that some defects may occur around the mesh pattern 21 due to dirt or the like, a plurality of main scanning line images with a certain interval are examined, and the majority of the measured values is decided. Even if there is a defect depending on the method or the like, it is possible to prevent the influence from occurring.

Further, in the routine 6-1, the size (vertical length) of the mesh seen from the main scanning direction may be measured. This is to check the length (run length) where the same pixel value continues on the main scanning line image, and check its histogram (Fig. 8).
, The peaks of that frequency are observed as harmonic series of run length based on one unit of mesh. Therefore, the vertical length of the mesh and also the vertical length of the mesh pattern can be evaluated. it can. In the case of the histogram shown in FIG. 8, since peaks occur at the run lengths 4, 8, and 12,
The size of the mesh seen from the main scanning direction is about 4 dots. Further, if desired, the vertical size of the mesh can be accurately evaluated up to the digit of the decimal point by performing statistical processing (for example, 3, 4, and 5 in FIG. 8 are samples of the length of one mesh). ), 7, 8 and 9 are considered to be samples of two lengths of the mesh, and these samples are averaged by weighting according to their appearance frequency). However, if the format of the mesh pattern (two-dimensional matrix structure of the mesh) is known, the mesh size can be easily calculated from the width of the mesh pattern.

In the routine 6-2 shown in FIG. 6, the horizontal length of the mesh pattern and the horizontal length of the mesh as viewed in the sub-scanning direction (see FIG. 7) are measured. This is because the inspection direction is changed from the main scanning direction to the sub scanning direction, and the processing content is the routine 6-1.
Is the same as. Up to this point, the vertical and horizontal dimensions of the mesh pattern and the vertical and horizontal dimensions of the mesh, which are observed according to the somewhat regularity of the image, are obtained from the main scanning direction and the sub-scanning direction on the image RAM 70.

In the next routine 6-3, the vertical and horizontal directions of the mesh pattern are determined. The direction of the mesh can be determined by various algorithms. The rotation search utilizes the fact that the cut length of the mesh pattern seen from the correct direction is the shortest. Referring to FIG. 9, first, two lines 2 and 3 intersecting the main scanning line 1 at a considerable angle are selected, and the vertical cutting length of the mesh pattern 21 seen from the image on the lines 2 and 3 is selected. Measure and compare. If the line 2 is shorter, the correct vertical direction is between the line 1 and the line 2 so that the line 4 is divided into two equal parts (the angle may be divided into two equal parts, but each end of the line 1 and the line 2 may be divided into two). , Which is a line connecting the centers between the positions, and the cutting length on the line image is measured. If the cutting lengths of line 1 and line 2 are not equal, the one with the shorter cutting length is selected. A line that further divides the line 4 into two is selected, and thereafter, the line is equally divided into two, and when the cutting length does not decrease, or when two lines having the same cutting length are detected. Abort processing (eg line 1)
And the cut length of the line 2 is equal, the bisecting line 4 has the shortest cut length).

In another approach, as shown by the circles and triangles in the enlarged image of FIG. 10, adjacent image lines (actually, since it is a discrete system, the line L as shown by the circle of FIG. The direction in which the bit patterns of (stepwise) match or best match can be determined as the direction of the mesh pattern. The small cells in FIG. 10 indicate image bit memory cells, and the large cells indicate light and dark meshes. It is advisable to start the bit pattern matching inspection at a stage where the direction of the mesh pattern is narrowed to some extent by comparing the cutting lengths by the above-described rotation search method. For example, when the direction is narrowed down within the angle range of the size of the mesh when viewed at the end of the mesh pattern,
Since the search range of the elongated rectangle is determined thereby, the bit pattern matching check is performed within this search range. Alternatively, the edge of the mesh may be detected within such a search range, and the direction may be determined from the found position of the edge. This is an approach that focuses on the most occurrence of image bit mismatch before and after the boundary between adjacent mesh columns. For example, when a narrow rectangular search area having a width of about a mesh size is obtained, an image inspection of the length of the mesh is performed on each image along the width direction (the sub-scanning direction when determining the vertical direction of the mesh pattern). Perform on bit lines, detect bit inversion positions, connect detection positions that fall within a half width of the mesh among these detection positions, or determine the regression line of these detection positions. Gives the direction of the mesh pattern.

The result of the processing of the routine 6-3 is the vertical and horizontal directions of the mesh pattern or two lines giving the directions. Therefore, information on the positions of the meshes at the upper, lower, left and right ends of the mesh pattern in the two lines is also obtained, which is a routine 6-instead of the routine 6-4 described below.
4b.

In Routine 6-4, a plurality of suitable line images along the vertical and horizontal directions of the mesh pattern 21 obtained in Routine 6-3 are examined to find the left side of the mesh pattern 21,
Detect the contour lines (if equipped with contour lines) or the center lines of the mesh columns or rows at the right, top, and bottom edges, and
Determine the position of the mesh in the corner. FIG. 11 shows an example of an algorithm for detecting the leftmost mesh row by parallel search. The image RAM is arranged along the vertical direction of the mesh pattern.
The line 1 at the end on 70 and the line 2 which is considered to pass through the inside of the mesh pattern are examined, and if the line 2 includes a plurality of black pixels indicating the existence of the mesh pattern, the line 1 and the line 2 between If line 3 is selected and if line 3 also passes through the mesh pattern, line 4 that bisects line 1 and line 3 is selected, and the 2 min search is continued in the same manner, and the interval between the two lines is sufficient. As it becomes closer (for example, about the size of the mesh), one line consists of only or almost white pixels that indicate the outside of the mesh pattern, and the other line is inside the mesh pattern (assuming there is no outline here). That includes a black pixel group (for example, a pixel group whose run length 2 has an interval approximately equal to the vertical length of the net-like pattern and whose evaluation value of the mesh length is approximately equal to an integral multiple thereof). Obtained Et al., Truncation processing, the latter line search results, i.e. the line passing through the left edge of the mesh of the mesh pattern (center line).

If a more accurate center line of the leftmost mesh is desired, a line pair l ₀ and l ₁ and a line pair l ₂ and l ₃ (this line pair are arranged at 3-bit intervals as shown in the enlarged image of FIG. 12). Is separated from the former line pair by about the size of the mesh) every time one inspection is completed by shifting it by 1 bit to the right, and the white pixel in the sub-scanning direction between the line pair l ₀ and l _{1 is} black. The number of the line l ₀ when the number of pixel changes to the bit is maximum, and the line l ₂ when the number of bit changes between the line pairs l ₂ and l ₃ is the maximum.
, The left boundary E0 between the mesh pattern and the outside is between l ₀ and l ₁ , and the leftmost mesh row and the second mesh row are between l ₂ and l _3. Is the boundary E1. If a line M (hatched in the drawing) that divides the boundary E0 and the boundary E1 into two is selected, this is the accurate center line of the mesh at the left end (note that the boundary detection due to the maximum discrepancy in image bit patterns). The method is also effective when there is distortion in the captured image data, and the boundaries of each mesh of the mesh pattern may be examined and created, and the internal points may be sampled.

The mesh positions at the four corners of the mesh pattern 21 are given by the intersections of the center lines, for example, the upper left corner is given by the intersection of the left end line and the upper end line.

Routine 6-4b is an alternative process to routine 6-4. It finds the intersection between the line in the vertical direction and the line in the horizontal direction of mesh pattern 21 detected in routine 6-3b, and the intersection point is obtained. Is the center of the mesh of the row and column of the mesh pattern 21. Specifically, the positions and image formats showing the fragments of the meshes at both ends in the line image in the vertical direction of the mesh pattern 21 detected by the routine 6-3b, that is, the number of meshes in the vertical direction of the mesh pattern 21 (in FIG. 2, 48), the line number of the mesh whose center is the intersection is calculated, and similarly, the routine 6-3
A row of meshes centered on the intersection point based on the positions of the mesh fragments at both ends in the line image in the horizontal direction of the mesh pattern 21 detected in b and the number of meshes in the horizontal direction of the mesh pattern 21 (72 in FIG. 2). The number is calculated.

After the routine 6-4 or 6-4b, in the routine 6-5, the center of each mesh of the mesh pattern 21 on the image RAM 70 is set to the standard mesh (4
Routine 6 is calculated from the center position of the mesh of the corner or the mesh in which the row and column numbers are specified), the evaluation values of the size of the mesh in the vertical and horizontal directions, and the vertical and horizontal directions of the mesh pattern 21.
At -6, the lightness and darkness of the mesh is identified by sampling the image bit at the center of each mesh.

It should be noted that not only a single image bit at the center of the mesh is sampled, but also image bits around it (for example, up, down, left and right) are sampled with the center of the mesh as a reference, and the brightness of the mesh is determined from the result. Further, the reliability may be evaluated.

Next, a second modification will be described with reference to FIGS. 13 and 14. For convenience, assume a mesh pattern with contour lines. The matrix structure of the mesh of the mesh pattern is known. In the first routine 14-1 of the decoding process (FIG. 14), a plurality of different line images passing through the memory center of the image RAM 70 (FIG. 13) are examined, and the black pixel first appearing from both ends of each line image toward the center is checked. Focusing on a fragment of the contour line, the detection position is sampled. To select a plurality of line images, it is convenient to make the first group of line images intersect the upper and lower contour lines of the mesh pattern and the second group of line images to intersect the left and right contour lines of the mesh pattern. Well, this is easily determined because the range in which the mesh pattern is present falls within a certain fluctuation range. As a result, the sampled detection position file is predicted to be the first group that is expected to be on the upper contour line, the second group that is expected to be on the lower contour line, and the position on the left contour line. Third
It is divided into a group and a fourth group which is expected to be located on the right contour line. For each group, if straight line fitting by a regression line or the like is performed except for the sampling positions that are out of alignment, that is the locus of the contour line, and the intersection points between the four straight lines obtained by this are the mesh patterns. The positions of the four corners are determined (14-2). In this method, the mesh pattern 1 is formed due to stains on the recording medium 14.
This method is also effective when the peripheral line portion 1 is partially destroyed, and the positions of the four corners of the mesh pattern 21 can be accurately measured. There is also an advantage that the processing time is short.

After the positions of the four corners are determined, the center of each mesh is calculated in the same manner as in the first modification, and the image bits located there are sampled to complete the decoding process (14-3, 14-4).

Next, a third modification will be described with reference to FIG. In the decoding process of this modification, the image RAM
The image RAM 70 is divided into four by the main scanning line and the sub-scanning line passing through the memory center on 70, and the position of the corner of the mesh pattern in each quadrant is determined. Explaining the first quadrant shown in FIG. 15, the main scanning line 1 and the sub-scanning line 2 are examined first, and the black pixel first found from the end is regarded as a fragment of the contour line and its position is stored. next,
The diagonal line 3 connecting the upper right of the image RAM 70 and the center is examined, and the position of the black pixel first found from the end is stored.
Next, image R between main scan line 1 and diagonal line 2
The line connecting the midpoint at the upper end of the AM 70 and the center of the image RAM 70 is examined, a similar black pixel is detected, and the position is stored. Here, the positional relationship of the black pixels regarding the lines 1, 4, and 3 and the positional relationship of the black pixels regarding the lines 2, 3, and 4 are examined. As a result, in the case of FIG. 15, it is found that none of the three black pixels is on a straight line. In other words, there are corners of the net-like pattern on the line passing between line 3 and line 4. Therefore, a similar black pixel is detected by drawing a line passing through the center of the memory and the middle point of the lines 3 and 4 at the upper end of the memory.
The straight line check between the black pixels of No. 5 and the straight line check between the black pixels of lines 2, 3, and 5 are performed. According to FIG. 15, it is determined that the lines 1, 4, and 5 are on a straight line. Therefore,
The center of the upper right corner of the mesh pattern is between line 3 and line 5. Therefore, a line 6 that divides the line 3 and the line 5 into two is drawn, and the search range is narrowed down in the same manner. If the space at the memory end between the two lines for drawing the bisector becomes sufficiently small (for example, a distance of 1 dot), the position of the black pixel on that line is located at the upper right corner of the mesh pattern (to be exact, the contour line The upper right point).

Next, a fourth modification will be described with reference to FIGS. 16 and 17. In this example, 4 are placed outside the mesh pattern 21.
One alignment mark R (here, a black circle) is provided, and from the image RAM 70 in the decoding process,
The center of the alignment mark R is measured, the center position of each mesh of the mesh pattern 21 is calculated from the result in accordance with some regularity, and the image bit at that position is sampled to identify the brightness of each mesh. To do. In order to detect the center of each alignment mark R, line images in the direction of, for example, 45 degrees are examined one by one from the edge of the image RAM 70 as shown in the enlarged image of FIG. When a line image including a run length of black pixels regarded as a fragment of is hit, the search range is narrowed down, the line image is examined finely (for example, at 1 dot intervals) within the range, and the projection A is taken. Furthermore, the line direction is, for example, 9
It is rotated by 0 degree, the line image in that direction is examined at fine intervals, and its projection B is taken. The edges E1 and E2 and the edges E3 and E4 of the projections A and B are found, a line passing through their centers is obtained, and their intersections are obtained.
This intersection is the center of the alignment mark R. By performing such processing at the four corners of the image RAM 70, the center positions of the four alignment marks R are determined. In the case of this method, even if a maximum of two marks among the four alignment marks R are destroyed, the destruction can be detected (because the shape of the projection and the distance between edges are greatly different from the assumed reference, detection is possible). If the positions of the remaining two sound alignment marks R are measured, the center position of each mesh of the mesh pattern can be calculated accurately from the regularity of the image data.
There is an advantage that the decoding durability against stains on the top is strong.

As another modification, for example, the image sensor may be of a magnetic / electric conversion type that reacts with magnetic ink or the like. Instead of a mechanically scanned image sensor, a fixed area image sensor may be used, although the cost is high.

[0036]

As is apparent from the above description, in the present invention, the image sensor means for taking in the image of the mesh pattern on the recording medium in a format in which the distance accuracy thereof is substantially preserved is provided, and the image data of the taken-in image is provided. The position of the pixel data representing each mesh in the image data is collectively determined based on the geometrical regularity that the whole has, and the brightness of each mesh is identified based on the pixel data at each determined position. Therefore, the processing amount for data reproduction can be reduced, the processing time can be shortened, and decoding with few errors can be performed, which is suitable for real-time use.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of a data reading device according to an embodiment of the present invention.

FIG. 2 is a diagram exemplifying a coded image of a mesh pattern which is a reading target of the embodiment.

FIG. 3 is a flowchart of scanning / decoding processing according to the embodiment.

FIG. 4 is a diagram showing an example of scanning of an image sensor with respect to an encoded image.

FIG. 5 is an enlarged view showing a scanning example.

FIG. 6 is a flowchart of a decoding process according to a first modified example.

FIG. 7 is a diagram showing an outline of image data on an image RAM, which is used for explaining measurement of the length of a mesh pattern seen from the main scanning direction.

FIG. 8 is a diagram showing a run-length histogram relating to a main scanning line image.

FIG. 9 is a view used for explaining measurement of a vertical length of a mesh pattern by rotation search.

FIG. 10 is an enlarged view of an image and is used to explain the determination of the direction of a mesh pattern by detecting the coincidence of image bit patterns.

FIG. 11 is a diagram used for explaining detection of a boundary of a mesh pattern by parallel search.

FIG. 12 is a diagram used to explain the boundary determination between mesh columns by detecting the maximum mismatch of image bit patterns.

FIG. 13 is a diagram according to the second modification and used for explaining the detection of the contour line of the mesh pattern on the image RAM.

FIG. 14 is a flowchart of a decoding process of the second modified example.

FIG. 15 is a diagram showing a principle of detecting a corner position of a mesh pattern on an image RAM by a binary search according to a third modification.

FIG. 16 is a diagram showing an image RAM in which a coded image in which a registration mark is added to a mesh pattern is taken in according to a fourth modification.

FIG. 17 is a diagram showing the principle of measuring the position of an alignment mark.

[Explanation of symbols]

 10 data reader 11 image sensor 14 recording medium 15 CPU 16 ROM 20 coded image 21 mesh pattern

Claims (1)

[Claims] 1. Image data representing the image is read from a recording medium on which an image of a mesh pattern in which data is encoded by light and dark selectively formed on each mesh is recorded, without substantially impairing the distance accuracy. Image sensor means, mesh position determining means for collectively determining the position of pixel data representing each mesh in the image data based on the overall geometric regularity of the image data, A data reading device, comprising: a mesh light / dark discrimination means for discriminating the light / dark of each mesh based on the pixel data at the position.