WO2001089199A1

WO2001089199A1 - Method for combining scanned partial-image data

Info

Publication number: WO2001089199A1
Application number: PCT/DE2001/001755
Authority: WO
Inventors: Jörg Olaf VON WECHGELN; Michael Hansen
Original assignee: Heidelberger Druckmaschinen Aktiengesellschaft
Priority date: 2000-05-16
Filing date: 2001-05-09
Publication date: 2001-11-22
Also published as: DE10024028A1

Abstract

The invention relates to a method for combining first partial-image data (6) and second partial-image data (7) to form complete-image data (9). According to said method: first reference points (14) are determined in the first partial-image data (6) and corresponding second reference points (15) are determined in the second partial-image data (7); new pixel grids are defined (18) section by section in the partial-image data (7), based on the second reference points (15); new pixels are interpolated in the new pixel grids (18); and the first partial-image data (6) and the new pixels are combined to form the complete-image data (9). The positions of the corresponding second reference points (15) are calculated by determining the maximum correlation coefficient between a search pattern (20; 23) and a search area (21; 24).

Description

Method for merging scanned field data

The invention relates to the field of electronic reproduction technology and relates to a method for merging partial image data which are obtained by several optoelectronic scans of an image original. The overall image data of the original image is created by merging.

In a scanner for optoelectronic scanning of templates, such as images, graphics and texts, the templates to be scanned are converted into electrical signals, which are then further converted into digital data for processing in electronic reproduction technology. If it is a flat bed scanner, the originals are arranged on a flat original carrier, and an optoelectronic scanning element scans the originals pixel by line and line, whereby the original carrier and the

Move the scanning element relative to each other. The scanning element essentially has a light source for line-by-line illumination of the original, an optoelectronic converter, for example a CCD line, for converting the scanning light coming from the original into the image signals and a scanning lens for sharply imaging the original on the optoelectronic converter and for adjusting the Reproduction scale for originals of different sizes or for setting different scanning resolutions. Such a scanner is known for example from DE 195 34 334.

There are flatbed scanners with a fixed original carrier and one

Scanning organ which is in the sub-scanning direction, i.e. moved perpendicular to the direction of the scanning line (the main scanning direction) along the original carrier. This design is typical for so-called desktop scanners. However, there are also embodiments in which the original carrier moves along a fixed scanning element.

The scanning resolution in the direction of the scanning line depends on the number of sensor elements in the CCD line and the respective original width Direction of the scan line. In the case of inexpensive scanners, the scanning element is usually designed such that the entire width of the original carrier is imaged on the CCD line. This results in a fixed resolution of the scanner for all templates. If templates are scanned that do not fill the entire width of the template carrier, the template is only imaged on a part of the sensor elements in the CCD line. The scanning resolution perpendicular to the direction of the scanning line is determined by the relative speed between the original and the scanning element.

In the case of more complex scanners, the scanning resolution in the direction of the scanning line can be varied by means of the imaging scale, in that the scanning lens reproduces the respective original width completely and with optimum image sharpness on the CCD line. This means that smaller original widths can be scanned with higher resolution than larger original widths. The maximum adjustable scan resolution results from the number of sensor elements in the CCD line divided by the width of the original. To change the scanning resolution in the line direction, the image width, i.e. the distance between the CCD line and the scanning lens, and the object distance, i.e. the distance between the scanning lens and the original can be set by moving the scanning lens with a fixed focal length and / or the CCD line or by scanning lenses with different focal lengths into the beam path. However, it is disadvantageous that large originals cannot be scanned with a high resolution. A high scanning resolution in connection with a large-format original is required, for example, if already screened color separation films are to be scanned from finished printed pages in order to process them in the form of digital image data in an electronic reproduction system.

To solve this problem, it is provided in even more complex scanners not only to move the scanning element perpendicular to the direction of the scanning lines but also in the direction of the scanning lines. This allows a large format document to be scanned in columns in several scans. After the first column has been scanned, the scanning element moves in the direction of the scanning line shifted by the column width, and the second column is scanned, etc. For the individual column scans, the imaging scale can be set so that the column width is mapped to the entire CCD line. In this way, a large format original can also be scanned with high resolution. Such a scanner is known for example from EP 0 833493.

The partial image data obtained by scanning the individual columns must be electronically combined to form an entire image. According to the prior art, this is done by copying the partial image data line by line into a file of the overall image data. 1 shows the scanning of a section of a template (2), which is mounted on a template carrier (1), in a first column (3) and a second column (4). The scanning element for scanning the second column (4) is expediently displaced by less than the column width, so that an overlap area (5) is formed on the right edge of the first column (3) and on the left edge of the second column (4). 2 shows the partial image data obtained during the scanning of the columns (6) for the first column and (7) for the second column. A seam line (8) is defined in the overlap area at which the partial image data are to be combined to form the overall image data (9). A line (10) of the total image data (9) is then formed by combining a line segment (11) from the first partial image data (6) and a line segment (12) from the second partial image data (7). The line segment (11) begins at the left edge of the first partial image data (6) and ends at the seam line (8), and the line segment (12) begins at the seam line (8) and ends at the right edge of the second partial image data (12). The length and position of the row pieces (11) and (12) to be removed can be taken from the column widths, the definition of the seam line (8) in the overlap area (5) and the displacement of the second column (4) relative to the first column (3 ) be calculated.

This simple combination of partial image data according to the prior art has the disadvantage that mechanical and optical installation tolerances of parts of the scanning member and tolerances in the movement of the scanning member during the scanning are not taken into account. Such tolerances mean that the drawing data on the seam line (8) do not match exactly. These errors are particularly noticeable at high resolution.

It is therefore an object of the present invention to compensate for the registration errors which occur due to mechanical and optical tolerances when the partial image data are combined, and to generate overall image data which contain no visible errors.

The invention is explained in more detail below with reference to FIGS. 1 to 10.

Show it:

1 shows the scanning of a template in two columns,

2 shows the joining of partial image data into total image data according to the prior art,

3 shows the occurrence of a registration error due to a misalignment of the CCD line,

4 corresponding reference points in the overlap area of the partial image data,

5 shows the principle of the compensation of registration errors according to the invention,

6 shows an arrangement of horizontal search patterns and search areas,

7 is an enlarged view of a horizontal search pattern and search area,

8 shows the interpolation of correlation coefficients with a polynomial, 9 shows an arrangement of vertical search patterns and search areas, and

10 shows the definition of a pixel grid in a partial section of the second partial image data.

Fig. 3 shows an example of the occurrence of a registration error when joining the data if, as a result of installation tolerances, the CCD line is somewhat crooked compared to the direction of movement (arrow direction) of the scanning element when scanning the first column (3) and the second column (4) is. For illustration, the misalignment has been drawn in a greatly exaggerated manner in FIG. 3. It is clear that in the overlap area (5) the end of line i in the first column (3) and the beginning of line i in the second column (4) do not contain the same image information. The continuation of row i in the first column (3) is in the second column (4) in a row j in front of it. For the partial image data, this means that a certain line segment (13) can be found on the one hand at the end of line i in the first partial image data (6) and on the other hand at the beginning of line j in the second partial image data (7).

3, the corresponding image data in the overlap area (5) are not vertically shifted by entire lines in a real scanning system, but also by fractions of a line height. In addition, as a result of further mechanical and optical tolerances, there are also horizontal displacements of the corresponding image data in the overlap area (5), likewise by fractions of a pixel width. Furthermore, in general, these horizontal and vertical register errors are not constant during the movement of the scanning element when scanning a column, but change in the overlap area (5) along the seam line (8).

Fig. 4 shows this general situation of the registration errors in the overlap area (5) of the partial image data. The first partial image data (6) have been adopted here as a reference system. On the seam line (8) of the first partial image data (6) first reference points (14) have also been adopted. As previously explained, the second reference points (15) corresponding to the first reference points (14) in the second partial image data (7) are shifted horizontally and vertically by the registration errors, so that they are there on a new curved seam line (16).

5 shows the principle of the method according to the invention for compensating the registration errors. The illustration shows a horizontal strip of the first and second partial image data (6) and (7), the height of which corresponds approximately to the vertical distance between two first reference points (14), the grid being the

Pixels are shown very coarsely. First, the positions of the second reference points (15) corresponding to the first reference points (14) are determined by correlating the first partial image data (6) and second partial image data (7) present in the overlap area (5). Then the curved seam line (16) is approximated by straight lines between two adjacent second reference points (15). The line segment (17) between two adjacent second reference points (15) forms a boundary of a new pixel grid (18) in the second partial image data (7), which is geometrically distorted compared to the original pixel grid (19), i.e. is rotated and stretched or compressed. The new pixel grid (18) is selected such that the same number of lines is contained in the second partial image data (7) along the straight line segment (17) as in the first partial image data (6) along the connecting line between the first reference points (14). ,

At the positions of the pixels in the new pixel grid (18), new pixel data are interpolated from the neighboring pixels of the original pixel grid (19) (marked with small squares in FIG. 5). Any interpolation method is used for this, for example bilinear or bicubic interpolation. The interpolated pixels from the second partial image data (7) are then combined with the original pixels from the first partial image data (6) (marked with small circles in FIG. 5) to form the total image data (9). The transition point between the pixels from the first partial image data (6) and the interpolated pixels from the second partial image data (7) does not have to take place exactly at the seam line (8), as drawn in FIG. 5. The transition can also be any number of pixels to the left or right of the seam line (8), as long as the transition point is still in the overlap area (5). The transition point can, for example, also be varied from line to line in the overall image data (9) by a random value. The seam lines (8) and (16) therefore do not have to indicate the location of the transition between the partial image data. They only serve to determine the geometric parameters of the registration errors and to compensate for the registration errors in the combined overall image data (9) by piece-wise adaptation of the geometry. The steps of the compensation method according to the invention for the registration errors are explained in detail below.

First, the positions of corresponding first reference points (14) and second reference points (15) are searched for in the overlap area (5) of the first and second partial image data (6) and (7). This is done by calculating and maximizing the correlation between the right edge of the first field data (6) and the left edge of the second field data (7). In a preferred embodiment of the invention, the horizontal and the vertical positions of the first reference points (14) and the second reference points (15) are determined separately.

6 shows an arrangement of horizontal search patterns (20) on the right edge of the first partial image data (6) and associated search areas (21) on the left edge of the second partial image data (7) for determining the horizontal positions of the first reference points (14) and the second reference points (15). Sections from a line in the overlap area (5) of the first partial image data (6) are selected as search patterns (20) at arbitrary vertical distances, for example sections of 50 pixels each. The starting point of the search pattern (20) is set, for example, as the first reference point (14). As assigned search areas (21), sections from the second partial image data (7) are selected which comprise several lines and also more picture elements per line and which include the expected position of the respectively associated second reference point (15), for example sections of 5 lines x 70 each pixels. 7 shows a detail of a search pattern (20) in the first partial image data (6) and the assigned search area (21) in the second partial image data (7). In the search area (21), a window (22) the size of the search pattern (20) is moved pixel by pixel in each line of the search area (21). In each position of the window (22), the correlation coefficient r (d) is determined from the image data in the search pattern (20) and the image data g _{i + d} in the window (22) of the search area (21). The parameter d specifies the displacement of the window (22) in the line of the search area (21).

In equation (1), f _{m is} the mean value of the pixels in the search pattern (20) and g _{m is} the mean value of the pixels in the window (22). The maximum rmax and the associated window position dmax are determined in each line of the search area (21).

Since the window position dmax only indicates the possible horizontal position of the second reference point (15) with one pixel, according to a preferred embodiment of the invention the correlation coefficients at the window positions dmax-1, dmax and dmax + 1 are interpolated by a second degree polynomial, and the exact position d_relation is determined, at which the maximum extreme value of the polynomial lies. This is shown in Figure 8.

After the exact position d_relationship has been determined in each line of the search area (21), the median value of these positions is determined in accordance with a preferred embodiment of the invention and stored as the horizontal position of the second reference point (15). In order to make the calculation of the exact positions d_related more reliable, according to a preferred embodiment of the invention, before the correlation coefficient r (d) is calculated, it is checked whether there is a sufficiently high signal structure in the search pattern (20) and in the window (22). The mean square deviation of the Pixel gray values are calculated both in the search pattern (20) and in the window (22). If they are below a threshold T1 of, for example, 1 gray level / pixel, the window position is not taken into account for the further calculation. If the mean square deviation in the search pattern (20) is below the threshold, a new position of the search pattern is selected. The following condition must therefore be met for a sufficient signal structure.

^ ∑ (f _t ~ f _m f ≥ l AND ^ ∑ (g _{i + d} - g _m f ≥ Tl (2)

To stabilize the results, according to a further preferred embodiment of the invention, only correlation coefficients rmax that exceed a threshold T2, e.g. rmax> 0.95.

After the exact horizontal position d_relationship has been determined for every second reference point (15) in the manner described, the exact vertical position of the second reference points (15) is calculated using the same method. 9 shows an arrangement of vertical search patterns (23) on the right edge of the first partial image data (6) and associated search areas (24) on the left edge of the second partial image data (7). Vertical stripes, which are one pixel wide, in the overlap area (5) of the first partial image data (6) are selected as the search pattern (23), for example stripes each 50 lines high. As the assigned search areas (24), vertical sections from the second partial image data (7) are selected which are several pixels wide and also higher than the search patterns (23) and which include the expected position of the associated second reference point (15), for example sections of 5 pixels x 70 lines each. The exact horizontal and vertical position of the second reference points (15) is then used to approximate the curved seam line (16) by straight lines (17) between two adjacent second reference points (15). 10 shows again in detail the formation of the pixel grid (18) of the pixels to be interpolated in the second partial image data (7) according to a preferred embodiment of the invention. A section with three adjacent second reference points (15) is shown. The upper two reference points (15) are connected by the straight piece (25) and the lower two reference points (15) by the straight piece (26). In the uppermost reference point (15), the line direction for the upper section of the new pixel grid (18) is defined by a perpendicular (27) to the straight line piece (25). In the middle reference point (15), the line direction for the lower one is correspondingly defined by a perpendicular (28) on the straight piece (26)

Section of the new pixel grid (18) set. The perpendiculars (27) and (28) meet at points (29) and (30) on the right edge of the second partial image data (7). According to the required number of lines in the upper section, the required number of pixels is evenly distributed on the straight line (25) and on the connecting line between points (29) and (30). Likewise, the required number of pixels per line is evenly distributed on the vertical (27) and (28). After the distribution of the pixels on the outer boundaries of the upper section, the locations of all pixels in the new pixel grid (18) result from linear interpolation of the pixel coordinates on the boundaries. For each line segment between two adjacent reference points (15), a separate section with a respectively adapted new pixel grid (18) is defined in this way.

In the foregoing description of the present invention, for simplification and better illustration, it was assumed that the first sub-picture data (6) were scanned without tolerances, so that a straight seam line (8) results, and that the second sub-picture data (7) are subject to registration errors , resulting in a curved seam line (16) (Fig. 4). The second partial image data (7) are therefore adapted to the first partial image data (6) by interpolation of new pixels. The method can also be used in reverse, ie new pixels are interpolated in the first field data (6) in order to adapt them to the second field data (7). In practice, however, are both the first The partial image data (6) and the second partial image data (7) are subject to errors due to mechanical and optical tolerances. In order to take this fact into account and since it is only important to compensate for the relative registration errors of the partial image data to one another, according to a further preferred embodiment of the invention the determined registration errors, ie the angle and length deviations of corresponding straight line segments between adjacent reference points (14) or (15) distributed over the first partial image data (6) and the second partial image data (7) and new pixel grids (18) resulting therefrom are defined in both partial image data and new pixels are interpolated. This gives an even better and more uniform image quality of the overall image data (9).

Claims

claims

1. A method for combining first partial image data (6) and second partial image data (7) to form overall image data (9), characterized in that - in the first partial image data (6) first reference points (14) and in the second

Sub-image data (7) corresponding second reference points (15) are determined,

- on the basis of the second reference points (15) new pixel grids (18) are defined in sections in the second partial image data (7), - new pixels are interpolated in the new pixel grids (18), and

- The first partial image data (6) and the new pixels are combined to form the total image data (9).

2. The method according to claim 1, characterized in that the positions of the corresponding second reference points (15) are determined by determining the maximum correlation coefficient between a search pattern (20; 23) and a search area (21; 24).

3. The method according to claim 2, characterized in that the maximum correlation coefficient for the horizontal and the vertical position of the corresponding second reference points (15) is determined separately.

4. The method according to any one of claims 2 or 3, characterized in that the positions of the second reference points (15) are determined by polynomial approximation of correlation coefficients.

5. The method according to any one of claims 2 to 4, characterized in that the correlation coefficients are only determined if a sufficiently high signal structure is present in the search pattern (20; 23) and in the search area (21; 24).

6. The method according to any one of claims 2 to 5, characterized in that only for determining the positions of the second reference points (15) Correlation coefficients are used that are greater than a threshold value (T2).

7. The method according to any one of claims 1 to 6, characterized. that new sections in sections based on the second reference points (15)

Pixel grids (18) are defined in the first partial image data (6) and the second partial image data (7),

- New pixels are interpolated in the new pixel grids (18), and

- The new pixels interpolated in the first partial image data (6) and the second partial image data (7) are combined to form the total image data (9).

8. The method according to any one of claims 1 to 7, characterized in that in the overall image data (9) the transition point between the first sub-image data (6) and the second sub-image data (7) is varied from line to line.