WO1992015169A1

WO1992015169A1 - Process and device for generating signals corresponding to the information content of raster-scanned images

Info

Publication number: WO1992015169A1
Application number: PCT/DE1992/000085
Authority: WO
Inventors: Axel Bauer; Roland Czolbe; Herbert Dethardt
Original assignee: Linotype-Hell Ag
Priority date: 1991-02-14
Filing date: 1992-02-08
Publication date: 1992-09-03
Also published as: DE4104467A1; CN1065171A

Abstract

Process and arrangement for transmitting raster-scanned images. Digital video signals for pixels which are appreciably smaller than the scanning width are derived by line-by-line and column-by-column scanning. The video signals undergo effective low pass filtering in the line and column directions by a filter window which includes at least the pixel that have passed through a raster mesh and which is shifted by an offset, line-by-line and column-by-column.

Description

Method and arrangement for generating signals which correspond to the information content of rastered images

The invention is based on a method according to the type of the main claim.

Often, artwork that has already been scanned is scanned optoelectronically in order to generate electrical signals suitable for further processing. When further processing to a differently rastered template, Moirέ can arise due to similar raster frequencies.

For a Moirέ-free further processing, electrical signals are therefore necessary which, although they reproduce the image content as precisely as possible, are largely free of spectral components caused by the grid. For this purpose, it is known to use a widened beam cross section when scanning screened originals. With this type of "descreening", however, significant quality losses, in particular sharpness losses, have to be accepted.

The object of the present invention is to provide a method and an arrangement for generating signals which correspond to the information content of rastered images, the information content of the images being converted into electrical signals as uninfluenced as possible, without the occurrence of disturbing spectral components caused by the rasterization. The method according to the invention with the features of the main claim has the advantage that electrical signals are available for the purpose of transmission and / or further processing, which largely correspond to the information content of the image, without spectral components caused by the screening leading to Moir Mo interference.

The arrangement according to the invention with the features of claim 12 has the advantage that the large amount of data resulting from the scanning can be processed with a relatively small amount of circuitry.

The measures listed in the subclaims enable further advantageous developments and improvements of the method and the arrangement according to the invention.

Exemplary embodiments of the invention are shown in the drawing using several figures and are explained in more detail in the following description. Show it:

1 shows a schematic illustration of a filter window used according to the method according to the invention,

2 shows a section of a grid of a print template in a schematic representation,

3 shows a representation of the scanning of the original and the offset of the filter window, 4 shows a block diagram of an arrangement for carrying out the method according to the invention,

5 shows a more detailed block diagram of the arrangement according to FIG. 4,

6 shows a quantization characteristic curve which is used in the analog / digital conversion of the video signals obtained during the scanning, and

7 is a diagram for explaining an Kaiser-Bessel window, which is preferably used for low-pass filtering.

Identical parts are provided with the same reference symbols in the figures.

1 schematically shows a filter window F in which 24 x 24 samples are taken into account. The video signal represents one pixel in each case. To distinguish it from the halftone dots of the original raster and from the pixels of the signals generated according to the invention, the pixels are hereinafter called fine pixels (FBP). In addition, the image to be transferred is used as a template and the grid in a corresponding manner as

Called template grid. In the example shown in FIG. 1, the dimension b of a fine pixel is 1/8 of the raster width Rv of the original raster. The halftone dots of the template are shown schematically by circles in FIG. 1.

The filter window F shown in FIG. 1 is applied to the entire template, shifted by an offset V in each case. The offset V occurs row by row and column by column. The size of the offset V, which must be an integer, depends on the oversampling factor Ue, the raster width Rv of the original raster, the raster width Ra of the output raster and the scale M to be selected according to the equation V = (RvUe) / (2-Ra-M ), in which

Rv is the grid size of the template grid or the image,

Ue the oversampling factor,

Ra the raster width of the output raster or the size of the pixels, which are given by the filtered video signals, and

M is the scale (100% = 1).

FIG. 2 shows an example of a template grid with grid points staggered line by line, so that diagonal grid lines are created, the grid width being Rv. The raster width Ra of the output raster then results in Ra = Rv / Ue.

3 shows schematically the displacement of the filter window F when scanning the template VL, which is interrupted in both directions in order to be able to display the filter window F on a favorable scale. In addition, the filter windows F shown with different line types are slightly shifted in the figure for the sake of clarity, but this does not correspond to reality. The total area scanned is larger than the template VL to be scanned in order to ensure perfect filtering of the template right up to the edge.

First, the filter window F is shifted n times in the scanning direction by the offset V (filter window F11, F12 ... F1n) until the entire original has been scanned in the scanning direction. Lines Z0 to Z23 are recorded. Then there is an offset V in the feed direction by eight lines, so that in the following Lines Z8 to Z31 are detected in the scanning direction (filter window F21, F22 ... F2n). This is repeated up to the filter window Fmn.

In order to obtain a lead, storage of the fine pixels of 32 lines each is provided, which is indicated in FIG. 3 by a dotted line.

4 shows a device for carrying out the method according to the invention as a block diagram. The analog video signals generated by a scanning device 1 (not described in more detail) are fed to an analog / digital converter 2. Details that are important in the analog / digital conversion in the context of the method according to the invention will be explained in more detail later in connection with FIG. 6. The analog / digital converter 2 generates a digital signal with n binary digits for each fine pixel, which are output in parallel to an image memory 3.

The image memory 3 essentially has the task of

Buffer sample values of 32 lines so that the sample values required for the filtering can be input simultaneously into the following digital filter processor 4. The output signals of the filter processor 4 are fed to an interface 5, with which the gradation adjustment is possible. The filtered digital video signals can be taken from the output 6.

FIG. 5 shows the arrangement according to FIG. 1 in more detail. The analog / digital converter 2 supplies digital video signals with four binary digits. In the signal processing steps following the analog / digital conversion, each binary position of the digital video signals is processed separately. In addition to the advantage of relatively simple processing circuits the advantage that an adaptation of the arrangement to video signals with different bit widths, for example those with three binary digits, is possible in a simple manner without the individual circuits having to be changed. The following description of the individual processing steps and the corresponding circuits therefore relates to a binary position of the digital video signals.

In a series / parallel converter 7, the samples of 24 fine pixels, which are present in series at the output of the analog / digital converter 2, are converted into parallel signals. As a result, the number of samples that are required from a scan line for later filtering is forwarded in parallel.

With a maximum sampling frequency of 35 MHz at the output of the analog / digital converter 2, a new 24-bit data word is then present at the output of the series / parallel converter 7 every 685 ns. These data words are written into a read / write memory 8 which has 32 lines per binary position with 49152 fine pixels each - that is, a capacity of 24 x 64 K per binary position. An address control unit 9, which contains an address counter for writing and reading, is assigned to the read / write memory 8. The address control unit 9 takes over the control of the offset V of the filter window F in the feed direction. Furthermore, the end of the scanning process is determined by the address control unit 9 by counting the lines. Via a bus system 12, the address control unit 9 is given the offset V in the feed direction by a CPU (central processor unit) 13.

For the later low-pass filtering, 24 lines of a scanned image section lying in parallel are required. Due the line-by-line scanning of the original and the series / parallel conversion, however, the image data are stored line-by-line in the read / write memory 8. With each write operation, sample values of pixels from 24 columns each are written in parallel into the read / write memory 8. Accordingly, when reading from the read / write memory 8 for each word read, the samples of 24 pixels lying next to one another are present in parallel at the input of the following column / row converter 10.

By appropriate addressing, 24 rows are sequentially written into a memory of the column / row converter 10, so that an image section of 24 rows by 24 columns is created. When reading out from the memory of the column / row converter 10, this image section is rotated by 90 °, so that 24 rows appear in parallel at the output of the column row converter 10. The columns are read out sequentially, creating a 24-row sequential data stream from column 0 to column n.

A 24-line by 32-column image section is taken from the data stream into the window memory 11. The offset V in the scanning direction is determined by appropriate control when reading out the data from the window memory 11. Corresponding control signals are supplied by the CPU 13 via the bus system 12.

The function of the digital filter processor can be described as follows: 24 24

i = 1 j = 1 p is the amplitude value of the respective pixel of the filtered video signal, h, j the filter coefficients, pj the amplitude values of the fine pixels, i the row index within the filter window F and j the column index within the filter window F.

By dividing the signal processing into different bit levels:

P = ∑ p _k .2 = ∑∑∑ hij. pi _jk . 2k kkij

k represents the index of the respective binary position.

Pijk is a binary position of a fine pixel and can therefore only have the values 0 or 1.

The coefficients hij preferably correspond to an Kaiser-Bessel window multiplied by a Bessel function, which enables high blocking attenuation with a filter flank that is not too steep. A short description with references is available in the book Schönfelder, Helmut (ed.): "Digitale Filter in der Videotechnik", Berlin 1988, pp. 98 and 99. Examples of the coefficients for the fine pixels of a column or a row are shown in FIG. 7, which shows the course of the coefficients in one direction of the filter window F for an oversampling factor Ue. After the oversampling factor Ue has been entered, the CPU 13 determines suitable coefficients. A manual

However, these "proposed" coefficients can be changed to adapt to the prevailing circumstances. The multiplication of the sampled values of the fine pixels within the filter window F by the coefficients is carried out for a binary position with the aid of a read / write memory 14. For this purpose, the coefficients are previously written into the read / write memory 14 by the CPU 13 via the bus system 12. The sample values are fed from the filter memory 11 to the address inputs of the read / write memory 14. The partial sums from the samples and the coefficients are available at data outputs of the read / write memory 14.

In circuit 15 there follows an addition over one line in each case within filter window F and in circuit 16 an addition of the line totals over the columns of filter window F. The filtered signal is thus available at the output of circuit 1. As already mentioned, in particular the circuits 7 to 16 can be designed relatively simply by separating the individual binary positions. Short processing times are also possible thanks to parallel processing.

The binary digits are summarized in a computing unit 17 weighted according to their significance. In the processing described so far, different quantizations and coefficients provide different word widths. In the arithmetic unit 17, the present results are therefore shifted to the right until the most significant binary digit of the range of values resulting from the scanning of the original lies on the most significant binary digit of 12 binary digits. The results are simultaneously cut to 12 binary digits.

The most significant binary position of the range of values resulting from the scanning of the original is current gradation setting determined and saved. The control of the computing unit 17 also takes place via the bus system 12 from the CPU 13.

The computing unit 17 is followed by a circuit 18 for setting the gradation, which essentially consists of a read / write memory which is loaded via the bus system 12 with a look-up table (LUT). The data stream comprising eight binary positions can then be extracted at output 6.

Using the LUT provides an excellent opportunity to adjust the entire arrangement. For this purpose, a template is scanned with a given gray wedge of, for example, 27 steps and a desired amplitude value is entered into the LUT for each gray value. Not only can any gradation be set, it can also compensate for undesirable influences, such as tolerances in the scanning distance (gray fog suppression), different illuminance levels during scanning and changes in the overall amplitude when switching the filter characteristics.

When choosing the amplitude resolution or the number of binary parts of the digital video signals, the following circumstances are important in addition to the circuitry or computing effort:

- Fine pixels, which are only partially covered by halftone dots of the template, are to be quantized sufficiently fine,

- It must be taken into account at the highest quantization level (maximum light) that even raster dots whose area is small compared to the area of a fine pixel are still be recorded, and at the lowest quantization level it must be taken into account that halftone dots are still recorded with almost 100%. - However, interference should be eliminated, which is caused in particular by the gray veil or fingerprints.

Noise caused by roughened glass plates should also be suppressed.

In order to meet these partially contradicting requirements, however, an advantageous quantization characteristic could be found which allows the above-mentioned interference to be suppressed without the reproduction of small raster points being impaired. An exemplary embodiment of such a characteristic is shown in FIG. 6, which shows the relationship between the analog video signals Sa and the digital video signals Sd in the case of 3-bit quantization. The values of the digital video signals are given in decimal, while the transmittance or the transparent portion of a fine pixel is plotted on the axis of the analog video signals Sa.

In the depths (small signal Sa), the analog / digital converter is adjusted with the offset adjustment in such a way that the quantization threshold between 0 and 1 is as high as possible, but at most in such a way that a 100% grid point is still certain can be distinguished from a 98% grid point and the 98% grid point is recognized with a digital value of> 0. The above-mentioned disturbance variables, which lie below this quantization threshold, are therefore set to 0 (digital). This is achieved with an additional offset of half a quantization level compared to the linear adjustment. There is thus an additional suppression of disturbance variables in the black of 5%. In the light (large signal Sa), the analog / digital converter is adjusted with the gain adjustment so that the quantization threshold between 6 and 7 is as low as possible, but minimally so that a 0% grid point of a 2% - Screen point can be distinguished and the 2% screen point is recognized with a digital value <7. In light, the above-mentioned disturbance variables, which lie between the maximum signal Sa and the quantization threshold 7/6, are set to 7 (digital). This is achieved by a misalignment of the gain by one quantization level (1 LSB) compared to the linear adjustment in the light. With 3-bit quantization, the interference rejection gain is 11.

As described, the thresholds in light and in the depths cut off the disturbances mentioned above. The digital signal Sd therefore practically only contains raster information. The resulting non-linear transmission characteristic is compensated for by appropriate programming of the LUT 18.

Claims

1. A method for generating signals which correspond to the information content of rasterized images, characterized in that by rows and columns

Sampling digital video signals are derived for pixels that are much smaller than the raster width, that the video signals are subjected to a low-pass filtering effective in the row and column direction with a filter window that at least those that fall on a grid mesh

Includes pixels and is shifted by an offset in rows and columns.

2. The method according to claim 1, characterized in that the filter window comprises approximately 24 x 24 pixels.

3. The method according to claim 1, characterized in that the offset V is determined as follows:

V = (RvUe) / (2-Ra-M), where

Rv is the grid size of the template grid or the image,

Ue an oversampling factor,

M is the scale (100% = 1).

4. The method according to claim 1, characterized in that the low-pass filtering is carried out with an Kaiser-Bessel window.

5. The method according to claim 1, characterized in that in the analog / digital conversion, an amplitude quantization corresponding to three to four binary digits takes place.

6. The method according to claim 5, characterized in that the corner points of the quantization characteristic are shifted to suppress signal interference.

7. The method according to claim 6, characterized in that the one corner point of the quantization curve in black

5% and the other corner point in white is shifted by 11%.

8. The method according to claim 1, characterized in that a separate low-pass filtering is carried out for the individual binary positions of the digital video signals.

9. The method according to claim 1, characterized in that for the purpose of low pass filtering video signals of pixels of a number of lines are stored, the number of lines corresponding at least to the size of the filter window.

10. The method according to claim 1, characterized in that the amplitude values of the filtered video signals are evaluated with the aid of a table stored in a read / write memory.

11. The method according to claim 10, characterized in that a template with predetermined gray levels, preferably with a gray wedge, is scanned for adjustment purposes and that in a table is written into the read / write memory, which contains selected amplitude values for the individual gray levels.

12. Arrangement for performing the method according to claim 1, characterized in that an image memory (3) is connected to a scanning device (1) via an analog / digital converter (2), which stores a part of the image required for low pass filtering , and that a digital filter processor (4) is connected to an output of the image memory (3).

13. The arrangement according to claim 12, characterized in that the image memory (3) and / or the digital filter processor (4) each in a plurality of similar circuits for one

Binary position of the output signals of the analog / digital converter (2) are divided.

14. Arrangement according to claim 12, characterized in that in the digital filter processor (4) a read / write memory is provided for multiplication and addition of a binary position with filter coefficients.

15. The arrangement according to claim 12, characterized in that the digital filter processor (4) is followed by a gradation table containing read / write memory (5).

16. Arrangement according to one of claims 12 to 15, characterized in that a central processing unit (13) for calculating and / or storing the offset of the

Filter window and / or the coefficient and / or the gradation table provided and with control inputs of the Image memory (3) and / or the digital filter processor (4) and / or the read / write memory (5) containing the gradation table is connected.