CN110502192B

CN110502192B - Print preview with adjusted resolution

Info

Publication number: CN110502192B
Application number: CN201910411277.6A
Authority: CN
Inventors: F·克洛普纳
Original assignee: Heidelberger Druckmaschinen AG
Current assignee: Heidelberger Druckmaschinen AG
Priority date: 2018-05-17
Filing date: 2019-05-17
Publication date: 2021-02-26
Anticipated expiration: 2039-05-17
Also published as: CN110502192A; DE102018207725B3

Abstract

The invention relates to a method for identifying, by a computer, a deviation between image outputs of different rendering processes in a workflow of a printing process, a first rendering process generating a first image from a pre-press stage PDF file and a second rendering process generating a second image from the pre-press stage PDF file, comprising: the first and second images are converted into gray images, and the gray images with large resolution are reduced to the size of the images with small resolution; quantifying the effect of the resolution difference according to an image comparison metric; calculating a difference image of the first and second images and binarizing according to a first threshold; applying an average filter to the difference image to obtain strongly biased pixel shares around each pixel; setting pixels with the peripheral strongly deviated pixel share less than 0.5 in the difference image as 0; partitioning the remaining pixels in the difference image; calculating sections larger than a second threshold value and outputting the sections as content deviation; the printing process is corrected taking into account the content deviations of the first and second images.

Description

Print preview with adjusted resolution

Technical Field

The invention relates to a method for identifying deviations between image outputs of different rendering processes.

The technical field of the invention belongs to prepress stage (Druckvorstufe).

Background

When processing print jobs, the data of the customer's print job are correspondingly preprocessed by the workflow system of the printing plant, usually in the so-called prepress phase. This means that: the text and/or graphics forming the print job are preprocessed in the pre-press stage in the workflow system in such a way that the print job can be passed directly to one or more printing machines for completing the print job. Here, it is not important: the prepress processing is carried out directly at the printing plant or, for example, in the Web-to-Print domain on a central job receiver server.

In this case, the user (i.e. the processor at the end of the prepress phase or the printing personnel of the printing plant) checks before the print job is transferred to the printing press (or to the plate exposer in the case of offset printing): whether the data of the print job has been properly preprocessed by the workflow system. This is typically done by previewing the corresponding print job data. Here, a preview image (in most cases in the form of a PDF file) is generated by the workflow system, which provides the user with an overview of the print job data that has been pre-processed to completion. Based on the preview image, the user can recognize the appearance of the data that has been preprocessed and is to be sent to the printing press (or plate exposure machine). Thus, the user can identify whether the data was correctly preprocessed or whether further correction is possible. However, the previews thus generated correspond only to the original print data, but do not completely coincide with them. Otherwise one does not need a preview but can look directly at the original print data to be sent to the printing press (or plate exposer). These original print data are in most cases generated in uncompressed format (e.g. bitmap) since compressed image formats (e.g. JPEG) are often accompanied by a corresponding quality loss. In contrast, very inefficient: such bitmaps are exposed in the preview with a correspondingly high memory requirement. Thus, a compressed format with reduced resolution is used on the display for previewing. The Rendering Engine (Rendering-Engine) used in previewing in the Heidelberg Print software is the same as that used to generate the Print data (Adobe PDF Print Engine). Only the parameters are different, in particular the resolution. In addition, when making (or processing) a PDF file using Adobe InDesign or Adobe Acrobat, in order to view the PDF file, an additional rendering engine (such as Adobe PDF Library) is mostly used. Due to these differences, these preview images are often inconsistent with the original print data in the form of bitmaps. In order to avoid the user from erroneously making an insignificant correction by considering the discrepancy as an error in the preprocessed print data at this time, such deviations must be recognized in time. The reverse situation is even more important. If the preview image is shown to be correct with respect to the customer's intent, but the print data contains deviations, this must be recognized in advance anyway.

The US patent application US 2004/0086156 a1 relates to the known prior art on this subject. This patent application discloses a method for plate monitoring at the prepress stage, wherein a raster image processor makes a first print image at a first resolution and then generates a second, non-rasterized print image, stores the two rasterized print images, and then compares the two rasterized print images to one another in order to thereby be able to evaluate the quality of the raster processor.

German patent application DE 10218068 a1 is another known prior art. This patent application discloses a method for communicating information about hue values and color values between prepress, print and quality control stages. The application is based on the following tasks: the coordination among the prepress stage, the printing stage and the quality control stage is improved. This is achieved by comparing the CMYK image structure data with the control image of the quality control stage of the current color measurement instrument. In a first step, therefore, the two pieces of information are brought to a uniform resolution by summing (zusmammenfassung) the image points of the higher-resolution image. If the resolution difference is not an integer relationship, the summarization is achieved by interpolation. Next, this printed image is analyzed and an image area suitable for the desired analysis is searched for.

Algorithms for comparing images for various application domains are also disclosed in the prior art. These algorithms are applied to the production of preview images. For example, are: a metric (Metriken) for determining the image quality from a reference image (SSIM, PSNR); methods in the field of image registration (mutual information, normalized cross-correlation); or feature-based methods (SIFT, SURF), and are furthermore mainly used in the field of image matching (rediscovery of certain features in different images).

However, the described application scenario is accompanied by difficulties in that: the images to be compared typically have a large difference in resolution (e.g., 72dpi for preview and 1200dpi for print). The deviations between these images, which are caused by the resolution, cannot be reliably distinguished from the deviations of the content to be recognized by the methods mentioned without modification. Content deviations are undesirable effects such as strongly distorted or missing fonts or objects that are produced when rendering (render). Furthermore, with all these methods too many false positive errors can be identified (Falsch-Positiv-Fehler). That is, deviations due to resolution (e.g., small fonts at low resolution) are no longer correctly represented and are therefore identified as content deviations. Furthermore, the greater the difference in resolution, the greater the false positive bias found with these mentioned methods.

Therefore, methods are also used in the prior art to find such false positive deviations between the bitmap and the re-digitized printed output. However, the methods employed to date have had the disadvantage that: in these methods more parameters must be set manually correctly to find deviations of a certain size. This is very costly and not always successful.

Disclosure of Invention

The object of the present invention is therefore to provide a method for printing a preview of a printing process, which allows the print data to be produced to be previewed as faithfully as possible.

This object is achieved by a method for identifying, by a computer, deviations between the image outputs of different rendering processes in a workflow of a printing process, wherein a first image is generated in a first rendering process on the basis of a PDF file of a prepress stage and a second image is generated in a second rendering process on the basis of said PDF file, comprising the steps of: converting, by the computer, the first image and the second image into grayscale images and downscaling (hermerskalierien) the grayscale image having the larger resolution to the size of the image having the smaller resolution, respectively; quantifying the effect of such resolution differences by said computer in accordance with an image comparison metric; calculating, by the computer, a difference image between the first gray image and the second gray image, and performing binarization (Binarisierung) according to a first threshold; applying, by the computer, an average filter to the binarized difference image to obtain a proportion fraction (Anteil) of pixels with strong deviations around each pixel; setting, by the computer, 0 to those pixels in the binarized difference image for which the proportional share of pixels strongly deviating around themselves is less than 0.5; partitioning (Segmentierung), by the computer, the remaining pixels in the binarized difference image on which the average filtering has been performed; finding out sections greater than a second predetermined threshold value by the computer and outputting the sections as content deviations; the printing process is corrected taking into account the deviation of the content found between the first image and the second image. The method according to the invention is characterized in that: manipulating (manipulating) the difference image such that as far as possible only content deviations remain. The implementation mode is as follows: the second image, which corresponds to the printed image to be generated in the first instance at a higher resolution, is downscaled in a manner as realistic as possible (wirklichkeitsgenauen) to the resolution of the first image, which corresponds to the printed preview image at a lower resolution. Based on this, a number of method steps are then used to determine: where there is a difference between the first printed image and the second printed image. These differences are then presented to the user accordingly, so that the user knows: the differences found between the first image with the lower resolution and the second image with the higher resolution or errors in the ready-to-print data must be corrected. Conversely, it is also possible that the preview has errors, but the print data is correct, and is avoided according to the method of the invention. Thus, the user can perform a correction procedure for the created print data to be sent to the printing press (or the plate exposure machine) without worrying about ignoring the error or correcting a false error that does not exist in the originally created print image data. That is, the method according to the present invention corrects a difference between the preview image and the actual print data, which is caused by the resolution and may cause an erroneous setting of the printing process.

Advantageous and therefore preferred developments of the method result from the preferred embodiments and the description with the figures.

In this case, a preferred refinement of the method according to the invention provides that, after the determination of the segments, the segments are marked in the first image by the computer on the basis of a bounding box (Begrenzungsboxen). One way to effectively show the user in front of the display the found deviation between the first image and the second image is to: these deviations are marked by means of a bounding box. It is also possible here that: when the bounding boxes showing the respective deviations are clicked, the user is shown a part based on the respective high-resolution bitmap (i.e. the second image) at this location for visual comparison. Thus, the user can accurately recognize the difference between the first image having the lower resolution and the second image having the higher resolution.

In this case, a further preferred development of the method according to the invention is: after determining the sections, the computer outputs the determined number of content deviations. In addition or as an alternative to displaying the ascertained deviation on the display by means of a bounding box, the user can also be presented with the amount of content deviation between the first image and the second image. For example, if only a small deviation or ideally no deviation at all is detected, the user does not have to check the differences caused by the resolution and can immediately proceed with the original evaluation of the first image.

In this case, a further preferred development of the method according to the invention provides that the first image relates to a preview image for the printing process and the second image relates to a bitmap for executing the printing process, wherein the preview image has a resolution which is significantly lower than the bitmap. The core of the whole method is as follows: there is a large difference in image resolution between the first image and the second image, and thus these content deviations may also occur during the different rendering processes used to make the two images. In order to evaluate the produced printed image by means of the preview image, the existing deviations with respect to the original printed image in the form of a bitmap must of course be correspondingly small. At least the user must know clearly the existing differences, that is, he must know: the specific errors in the preview image are not real errors but are only caused by the difference in resolution and the different rendering processes.

In this case, a further preferred refinement of the method according to the invention provides that the PDF files are used in a workflow for the printing process, and that the first rendering process and the second rendering process use different rendering engines based on the PDF standard. Because the corresponding PDF file or JDF in the workflow process is in a PDF-specific format developed specifically for the printing industry, a PDF standard-based rendering engine is also used for making previews that are also made as PDF files. A PDF rendering engine is also used for making the second image in bitmap form in a raster process. Typically an Adobe PDF Print Engine. However, it is also feasible to use different engines based on the PDF standard. If this happens, this may also be the reason for the discrepancy between the first printed image and the second printed image (i.e. the preview image and the print bitmap).

In this case, a further preferred development of the method according to the invention is to correct only those errors in the first image which are also present in the second image in order to correct the printing process. The significance of the method according to the invention is that: errors in the bitmap (i.e. the original print data) are detected as often as possible even if they do not occur in the preview image due to differences in the image resolution and/or rendering process. False positive errors in the preview image that are not present in the bitmap should also be found. Of course, such false positive errors should not be corrected, but only those errors in the preview image that are also present in the bitmap data, and thus reflect real errors. Of course, these real errors in the bitmap data must be corrected all the time, even if they do not appear in the preview image. Thus, when comparing two images (i.e., the bitmap and the preview image), all of the differences are always present.

In this case, a further preferred refinement of the method according to the invention provides that the remaining pixels in the binarized difference image for which mean value filtering has been carried out are partitioned by means of the row line reconstruction method (zeilenkizidenzverfahren). The partitioning in the penultimate method step, by means of which the retrieved content deviations are located, can be optimally carried out by means of a row-line reconstruction method. However, in principle, other partitioning methods are also possible.

In this case, a further preferred refinement of the method according to the invention provides that the predetermined first threshold value coincides with the predetermined second threshold value. Experience has shown that: using the same threshold for the first and second thresholds will give satisfactory results. However, it is also possible to select a different value for the second threshold (for example taking into account the resolution of the two images).

In this case, a further preferred refinement of the method according to the invention provides that the first threshold value is calculated by the computer on the basis of the quantized resolution difference. The advantage of this method is that no rigid thresholds or parameters need to be manually set, since these thresholds are determined in a dynamic manner according to image quality metrics.

In this case, a further preferred refinement of the method according to the invention provides that the method is applied several times in order to identify iteratively all existing deviations. Since these thresholds are chosen in dependence of the overall difference of the image, the variance of the roughness also has an effect on it. Therefore, in images with large deviations, small deviations are usually not recognized. This is a problem. Nevertheless, after the coarse deviations have been eliminated in the first iteration (Durchgang) of the method according to the invention, the method is used several times in subsequent iteration steps in order then to also reliably find out the smaller deviations, which has proven to be a good measure. In this case, coarse deviations are eliminated after the first pass, which is achieved by operating the produced PDF and by correspondingly continuing to apply the method according to the invention to the subsequently operated produced PDF.

Drawings

Such an invention and structurally and/or functionally advantageous refinements of the invention are described further below on the basis of at least one preferred embodiment with reference to the drawings. In the drawings, mutually corresponding elements are denoted by the same reference numerals, respectively.

The figures show:

FIG. 1: examples of inkjet printer system configurations;

FIG. 2: an example of a rendering-based bias;

FIG. 3: examples of printed images (grayscale images) in the form of bitmaps;

FIG. 4: examples of printed images in the form of preview images;

FIG. 5: examples of difference images;

FIG. 6: examples of partitioned biases;

FIG. 7: an example of a printed image in the form of a preview image with marked deviations;

FIG. 8: schematic flow of the process according to the invention.

Detailed Description

The method according to the invention is used for an inkjet printer 3 in a specific workflow system. Such a workflow system is exemplarily shown in fig. 1. The workflow system runs on one or more computers 1 that process respective print jobs 5. In this case, the print job 5 to be printed on the inkjet printer 3 is rasterized by the raster image processor 2, and the rasterized print images 4 are transferred from there to the inkjet printer 3 for the corresponding final printing.

A preferred embodiment variant of the method according to the invention is again schematically illustrated in fig. 8 with reference to a single method step.

The method comprises the following steps:

1) the higher resolution grayscale image I (x, y) is downscaled to the resolution of the preview image J (x, y). A grey scale image in the form of a bitmap 7 is shown in figure 3 and a preview image 8 is shown in figure 4.

2) Calculating the structural similarity index SSIM (I, J)

3) Calculating a difference image 10D (x, y) ═ I (x, y) J (x, y) · luminance

Binarization: d (x, y) ═ {0, if D (x, y) <255 × 1SSIM (I, J)), otherwise } (see fig. 5)

4) An average filter size of 7x7 was used: s (x, y) ═ mean _7x7(D (x, y))

5) Binarization: s (x, y) {0, if S (x, y) <0.51, otherwise }

6) Partitioning an image by row line reconstruction method (see FIG. 6)

7) The area (number of pixels) of these segments is determined, and

zeroing out segments with area less than 255 × 1-SSIM (I, J)

8) The remaining section 11 in the preview image 12 is marked with a bounding box. This is outlined in fig. 5, where the detected and marked deviations 13 can be seen well. Furthermore, in a particularly preferred development variant, the parts of the bitmap 7 exhibiting high resolution are used as a comparison when the bounding box is clicked.

Thereafter, the user can perform a change in the PDF or renderer setup based on the presented deviation 9 and eliminate the deviation. This then ensures that: errors are excluded before the printing process, rather than after the printing process, thereby saving significant time and cost.

The important explanation for this is that: the spike visible in fig. 3, deviation 9 as bitmap 7, is an example of rendered deviation 6 and is produced by a diagonal junction (Miter-Join) where the two meeting lines connect. Again, this is shown in fig. 2. If the angle between the lines is very sharp, the tip of such a bevel connection will be very long. To avoid this, a so-called bevel Limit (Miter-Limit) is provided which limits this length. That is, the spikes in the image of FIG. 3 can be eliminated by selecting a smaller amount of skew angle, or by making the angles between the line segments in vector form not too sharp.

The method according to the invention recognizes coarse content deviations 9, such as strongly deformed or missing objects, better than the methods disclosed in the prior art. Furthermore, the number of image areas erroneously identified as correlation deviations is small. By means of image quality metrics

The threshold is determined automatically, which also enables the method to be employed more flexibly than, for example, methods used in other prior art.

However, since the thresholds are chosen in dependence on the overall difference of the

images

7,8, a coarse deviation 9 also has an effect on this. Thus, in

images

7,8 with large deviations 9, often no small deviations 9 can be identified. To avoid this, in a further, supplementary embodiment variant, the method according to the invention is applied several times after the coarse deviations 9 have been eliminated by operating the PDF in order to iteratively recognize all deviations 9. Furthermore, in a further embodiment variant, the parameters (in the form of the threshold values) of the method can be adapted manually. This may be done, for example, at a user interface. Thus, the user may select the recognition sensitivity.

In addition to this, image comparison can be theoretically bypassed by analyzing parameters of the PDF file and the rendering engine. However, because the new version renderer is used regularly, the solution must be adapted regularly. The parameters were selected automatically by using the Structural Similarity Index (SSIM), which gives satisfactory results in experiments, whereby the parameters need to be set manually only in special cases.

List of reference numerals

1 computer

2 Raster Image Processor (RIP)

3 printing machine

4 rasterized printed image

5 print Job

6 rendering-based biasing

7 bit map

8 Preview image

Deviation between 9 bitmap and preview image

Difference image between 10 bit map and preview image

11 detected and partitioned deviations

12 preview image with marked, detected deviations

13 detected and marked deviation

Claims

1. A method for identifying, by a computer (1), deviations between image outputs of different rendering processes in a workflow of a printing process,

wherein in a first rendering process a first image is generated based on a PDF file of a prepress stage and in a second rendering process a second image is generated based on the PDF file of the prepress stage,

the method comprises the following steps:

-scaling, by the computer (1), the first image and the second image into a grey scale image and downscaling the grey scale image with the larger resolution to the size of the image with the corresponding smaller resolution;

-quantifying, by the computer (1), the effect of the resolution difference in accordance with an image comparison metric;

-calculating, by the computer (1), a difference image (10) between the first and second grayscale images and binarizing according to a first threshold;

-applying, by the computer (1), a mean filter to the binarized difference image (10) to obtain a proportional share of strongly biased pixels around each pixel;

-setting, by the computer (1), 0 pixels in the binarized difference image (10) for which the proportion of pixels with strong deviations in their surroundings is less than 0.5;

-partitioning, by the computer (1), the remaining pixels of the binarized difference image (10) on which the mean filtering has been performed;

-finding, by means of the computer (1), a section (11) that is greater than a second predetermined threshold value and outputting the section (11) as a content deviation (13);

-correcting the printing process taking into account the derived content deviation between the first image and the second image.

2. The method of claim 1, wherein the first and second light sources are selected from the group consisting of,

it is characterized in that the preparation method is characterized in that,

after retrieving the section (11), the section (11) is marked in the first image by the computer (1) according to a bounding box.

3. The method of claim 1 or claim 2,

it is characterized in that the preparation method is characterized in that,

after the section (11) is determined, the computer (1) outputs the determined number of content deviations (13).

4. The method of claim 1 or claim 2,

it is characterized in that the preparation method is characterized in that,

the first image relates to a bitmap for the printing process for performing the printing process, and the second image relates to a preview image, wherein the preview image has a resolution significantly lower than the bitmap.

5. The method of claim 1 or claim 2,

it is characterized in that the preparation method is characterized in that,

using the PDF file in a workflow for a printing process, and

the first rendering process and the second rendering process use different rendering engines based on the PDF standard.

6. The method of claim 5, wherein the first and second light sources are selected from the group consisting of,

it is characterized in that the preparation method is characterized in that,

only such errors in the second image are revealed for correction of the printing process: the error is also present in the first image.

7. The method of claim 1 or claim 2,

it is characterized in that the preparation method is characterized in that,

the partitioning of the remaining pixels in the binarized difference image (10) that has been mean filtered is performed by a row line reconstruction method.

8. The method of claim 1 or claim 2,

it is characterized in that the preparation method is characterized in that,

the first threshold value and the second predetermined threshold value coincide.

9. The method of claim 1 or claim 2,

it is characterized in that the preparation method is characterized in that,

the first threshold is calculated by the computer (1) from the quantized resolution difference.

10. The method of claim 1 or claim 2,

it is characterized in that the preparation method is characterized in that,

the method is applied several times in order to identify iteratively all existing deviations.