DE102006052515A1 - Two dimensional monochrome print images print quality objective determination method, involves determining diffuse measure from standard deviation of characteristics e.g. percentage area coverage, within printing screen frame - Google Patents

Abstract

The method involves partially digitalizing a print image using an image detection system. A bitmap image is formed, and a diffuse measure is determined from a standard deviation of characteristics e.g. percentage area coverage within printing screen frame of a predetermined screen. The product of the diffuse measure and balance point measure is formed in order to form a dimensionless structure value.

Description

The The invention relates to a method for evaluation, in particular monochrome Printed images, especially at medium tonal values.

The For example, the invention finds application in the field of gravure printing. This is usually a semi-autotypic printing process, i.e. the printing forms can be variable in depth and area. Due to the process also the maximum tonal values are rasterized. Because of the engagement low-viscosity paints becomes a "flood" of the individual Pressure points separating webs of the printing form on the substrate reached. This flooding of the webs begins in the midrange. These In principle, conditions also apply to non-semi-autotypical printing forms in gravure.

Therefore There are tonal values that appear on the substrate from discrete pressure points are formed, and tonal values in which the areas between the pressure points are completely flooded (idealized) as well as the transition between these areas.

Ideally there would be a smooth transition by a steadily increasing, cross-linking flooding, i.e., flooding (cross-linking) takes place from the Pressure points evenly and preferably in all directions instead. In reality there is this uniformity Not.

By an uneven flood from first Visually imperceptible or hardly noticeable pressure points can lead to the formation come from printing point agglomerates, which improves the quality of the printed image affect as structures become visible that are not present in the template are.

This phenomenon is in the Tiefdruck u.a. known as Waving and results from the Interactions during the wetting process of the ink on the paper; in well matched color-paper combinations is included appropriate press parameters to minimize the problem possible.

Unfortunately There is no single approach to assessing this phenomenon. Partially, a ranking order is created by a few observers and provided with key figures. This approach is not very objective, hardly leaves a comparability over longer periods too, shows no reliable Differentiation and is time consuming. Therefore, are reliable Characteristics neither customer for the evaluation of a printing result, still printing companies for the evaluation of papers regarding the phenomenon still paper manufacturers for the improvement of your products.

task The invention is to provide a method with which a automatic evaluation of printed images can be done, in particular with a waving index (visually effective overall impression on the Print patterns) and at least one structure parameter (marking the acting structures) with good reproducibility and high agreement in terms of the visual impression of a group of technical observers can be determined.

The measuring method should meet the following requirements:
The structures extracted from the image by the method should correspond to the visually apparent noise pattern.

The Pictures can preferably can be normalized in terms of their brightness. It can Structural parameters are calculated that have a high correlation with a visual ranking of the structure images, determined by Professionals.

The Structure characteristics can combined with material characteristics so that a high correlation with a visual ranking order of the printed patterns, determined by Professionals.

The respective correlation for 3rd and 4th should be greater than 0.9 be.

The Task is solved by that a printed image at least partially by means of an image acquisition system digitized and to the obtained digitized image Bitmap image is formed and determined to this a scattering measure is derived from the standard deviation of a parameter, in particular the percentage area coverage, within the grid window of a given grid.

In a preferred embodiment, this parameter is the percentage area coverings. It can be made according to the invention by determining visually disturbing, not technically avoidable structures, by means of image digitization and image analysis evaluation an automatic evaluation.

at The evaluation of the print patterns has shown that the paper parameter "brightness" or "whiteness" is an important one Plays a role, as lighter papers offer a better contrast (which basically positive for many aspects of print quality and thus desired is), which unfortunately also emphasizes undesirable structures, makes it visually more recognizable, because the resolution of the human eye a function of structure size and contrast is.

consequently a distinction must be made between the paper brightness that the Waving more or less highlights (contrast), and structural parameters, the really annoying ones Represent structures.

According to the method it can thus be provided, first or later Calibration to measure the brightness of the paper, e.g. with a Spectrophotometer. This step may be omitted if in terms of above-mentioned Brightness calibratable / calibrated imaging systems be used.

• a) Integration der Rasterstruktur (z.B. Medianfilter, 5·5, 4-fach) und extrahieren der dann entstehenden dunklen Strukturen (z.B. Schwellwerttrennung über mittleren Grauwert, korrigert um die mittlere Standardabweichung aller Bilder), Eliminierung visuell nicht relevanter Strukturen (z.B. Close-Operation, 2×2 square, 4-fach).
• b) Eliminierung der gleichmäßigen Strukturen (hier Raster) durch Frequenzanalyse (z.B. Fourier), Eliminierung visuell nicht relevanter Strukturen (z.B. Bandbreiten- oder Ringfilterung) und Erstellen eines Grauwertbildes mit den störenden Strukturen (Rücktransformation), Schwellwerttrennung (wie vorgenannt).

With an image acquisition system - if necessary, after a previous brightness calibration - the monochrome (monochrome) printed images digitized. In this case, for example, scanner, microscope + digital camera, or similar. and, for example, an 8-bit grayscale resolution is used. After image acquisition, a conversion of the acquired image into a black-and-white image, in particular a bitmap, can preferably take place. This can be done, for example, using digital filters, such as blur, thresholds and average tonal values. The soft-focus filters can be used to create structures that the human eye perceives. Disturbing structures can be extracted, for example, by means of a threshold value separation (in particular a separation of the image contents based on selected threshold values). For the process or the human eye not relevant structures can be eliminated in a further step, ie in particular be deleted from a preprocessed image. The digital images can be preprocessed so that the interfering structures can be extracted, ie in particular the preprocessed bitmap image only shows the disturbing structures, in particular in black and white. There are several possibilities for this, two of which are briefly described here:

• a) Integration of the raster structure (eg median filter, 5 × 5, 4 ×) and extract the resulting dark structures (eg threshold value separation over average gray value, corrected by the mean standard deviation of all images), elimination of visually irrelevant structures (eg close operation , 2 × 2 square, 4-fold).
• b) Elimination of the uniform structures (in this case raster) by frequency analysis (eg Fourier), elimination of visually irrelevant structures (eg bandwidth or ring filtering) and creation of a gray value image with the interfering structures (back transformation), threshold separation (as mentioned above).

For the rating the process resulting Bitmap images are used according to the invention statistical variables, e.g.

1. Scattering Measure [St]

To determine the scattering measure, the bitmap is evaluated with a matrix or a raster which, for example, is 1.8 × 1.8 mm ² in a preferred embodiment. According to the invention, the standard deviation of the percentage of area coverage within these windows of the matrix or of the grid, which results in the scattering measure, is determined. This is a measure of the homogeneity of the surface and indicates how evenly the agglomerations are distributed over the surface. The higher the scatter, the more restless or inhomogeneous the measured surface. Thus, this scattering measure forms a statistical quantity for determining the homogeneity of a printed image according to the invention.

2. centroid [Sc]

to Determination of the centroid a picture detail which in one embodiment is e.g. about 1/64 of the picture may overlap as a matrix for evaluating the percentage area coverage in predetermined steps, e.g. in a version in about 20 pixel steps over a Bitmap image moves.

When Focus is defines the difference between the two image sections, which are the highest and lowest area coverage exhibit.

It can Point clouds (clustering of individual objects) and agglomerates determined be over that Measuring field are distributed, with a matrix in a defined Size and in defined steps the structure pattern of the bitmap image scans. This makes it possible to identify and evaluate the focal points of the structure image.

3. Waving Index [WI]

The product of the scattering measure determined according to the invention and the preferred centroid divided by 1000 forms a dimensionless structure value, which is defined as the Waving Index. WI = (St * Sc) / 1000

applies one the procedure on the respective paper classes at (different Brightness and color acceptance e.g. in painted and unpainted Papers), one achieves a correlation significantly larger than 0.9 based on the visual rating. With the same waving index (+/- 0.05) the lighter paper is the visually worse cutting paper.

Of the ideal waving index is equal to 0 (absolutely evenly gray Area). Depending on the paper, good values are 0.5-0.7, cutting poor papers depending on the variety with values greater than 1.25 or 1.6 from.

A Characteristic curve WI over Tone shows the dependency from the printed tonal value and helps the actual critical area, Identify with changing substrates, colors and printing conditions.

4. Structural features

to further detailing can the coherent ones Structures (print agglomerates) within a bitmap as individual objects evaluated and in the form of a report file for printing stock development be used. This report provides information about e.g. Sizes and Number of objects.

With it is over these values the pure quality determination out possible, the printing process and especially the papermaking regarding of the waving phenomenon to improve.

When specific structural parameters show area / largest single area, box size (length / width ratio of Single objects), scope of individual objects and other good correlations for visual evaluation. these can e.g. to be evaluated multidimensionally depending on the under 1.-3. defined sizes.

The 0 shows for illustrative purposes an output image (left) and the S / W bitmap image determined therefrom according to the invention, which only shows the interfering structures and can be statistically evaluated according to the invention.

Further versions and introductory principles are described below. there is re all following and previous versions to determine that in connection with an execution mentioned Features can be used not only in the specific execution, but also with the other versions. All disclosed features this entire description of the invention are essential to the invention be classified and combined with each other or used in isolation.

1 Introduction

1.1 topic of this work

The Theme of this work is "Development and evaluation of a method for determining the waving phenomenon in the Gravure ".

1.2 problem definition

The gravure printing process is a semi-autotypic printing process. Semi-autotypical means that the printing forms are both surface and depth variable. So that the squeegee also in the maximum Tonwer If you have a guide through the bars, these tonal values must also be rasterized. As printing ink, low-viscosity inks are used. This leads to the so-called flooding, ie, that a flooding of the separating webs of the printing form takes place on the substrate. The flood begins in the midrange.

• • Tonwerte, die auf dem Bedruckstoff aus einzelnen Druckpunkten gebildet werden, und
• • Tonwerte, in denen die Flächen zwischen den Druckpunkten idealerweise komplett geflutet sind sowie
• • den Übergang zwischen diesen Extremen.

That's why there is

• • tonal values that are formed on the substrate from individual pressure points, and
• • Tones in which the areas between the pressure points are ideally completely flooded and
• • the transition between these extremes.

Ideally it should be a smooth transition through evenly rising, cross-linking flooding (i.e. flooding takes place from the pressure points evenly in all Directions instead). In reality is there this uniformity Not. By an uneven flood from first Visually imperceptible or hardly noticeable pressure points can lead to the formation of pressure point agglomerates come that the quality affected by the printed image, as structures become visible that are not present in the template are.

This phenomenon is in the Tiefdruck u.a. known as Waving.

Unfortunately There is no single approach to assessing this phenomenon. Often is a rank order created by a few observers and with Key figures provided. This approach is not very objective, hardly leaves a comparability over longer periods too and is time consuming.

Therefore are reliable Characteristics neither the customer for the evaluation of a printing result, still printing companies for the evaluation of papers regarding the phenomenon, still paper manufacturers for the improvement of your products.

1.3 Approaches to solving the problem

• • Bewertung durch Heranziehen der Histogrammfunktion und bedruckten Fläche
• • Automatische Bewertung durch Fast Fourier-Transformation und Filterfunktionen
• • Automatische Bewertung von Verteilung, Form und Wirkung in Verbindung mit einer Matrix- oder Mosaikfunktion

To solve the problem, various approaches based on different methods have been discussed:

• • Evaluation by using the histogram function and printed area
• • Automatic evaluation by Fast Fourier transformation and filter functions
• • Automatic evaluation of distribution, shape and effect in combination with a matrix or mosaic function

For this Investigation was the evaluation of distribution, shape and effect used in conjunction with a matrix or mosaic function. This method calculates a Waving Index (visually more effective Overall impression on the print samples) and structural parameters (marking the acting structures) with good reproducibility and high agreement regarding the visual impression of a group of technical observers.

It Print patterns are digitized with a suitable method and so preprocessed that it extracts the visually disturbing structures can be. The resulting bitmaps become specific structural parameters evaluated. For improved evaluation, further material characteristics can be used be used.

2 The gravure-specific phenomenon "Waving"

2.1 Gravure-specific problems in the printing process

In gravure printing, there are several phenomena that can occur due to the interaction of machine, printing form, printing ink and substrate. Some of these phenomena should be briefly highlighted in a table, including the Waving. For the phenomena Missing Dots, Mottling and Durchdruck, investigations on the papers to be examined later have already been carried out in previous projects.

2.1.1 Definition of the phenomenon "Waving"

• • Tonwerte, die auf dem Bedruckstoff als einzelne Druckpunkte vorliegen,
• • Tonwerte, in denen die Flächen zwischen den Druckpunkten idealerweise komplett geflutet sind

und den Übergang zwischen diesen Extremen.Since gravure printing is a semi-autotypic printing process (the printing forms are variable in depth and area abel) with low-viscosity paints, is in the printed image

• • tonal values that are present on the substrate as individual pressure points,
• • Tones in which the areas between the pressure points are ideally completely flooded

and the transition between these extremes.

The Printing form exists (contrary to the printed image) even to full tone out of cups, because only through the bars a guide for the Squeegee assured is. In illustration gravure, the flood begins already in the midrange. Ideally it's a smooth transition through evenly rising, cross-linking Flooding (i.e., flooding is uniform across the pressure points Directions), but in reality there is this uniformity Not. This creates a visible restlessness, which is also called waving referred to as.

waving means that partial pressure points begin to flood irregularly and form visible agglomerates. The viewer often takes these as disturbing Cloud or waveform true, which is why the terms cloudyness or cloudiness.

Under Thus, there is an ideal state of the midrange in the aspect of waving (Flooding is uniformly cross-linked from the pressure points in all directions instead), but depending on different parameters inevitably leads to disturbing effects.

The term "waving" thus always refers to the visually perceivable unevenness of the printing surface in the transition between raster and flooding printing, which is based on irregular agglomeration. The individual agglomerates have very different shapes and dimensions (see also 3 in the next section, "Waving Printing Samples"), which are affected by substrate, ink, and printing conditions. These visually perceptible irregularities must be grasped mathematically and metrologically in such a way that the result is reflected in what one perceives visually.

2.1.2 Pressure Samples with Waving

You can observe the waving particularly well on printed surfaces. This is where visual irregularities are most noticeable. Ideally, different tonal fields are printed next to each other, so you have the opportunity to compare the strength of the Wavings in different tonal values. In 1 such a strip is shown. On this one can clearly see the waving appearance.

Under the microscope you can, as in 2 shown clearly recognize the agglomerates and individual pressure points. The recording was made in 55% field Cup 0 on paper 6 with the magnification 0.8.

With the program Image-Pro Plus it is possible to sort out the shapes of agglomerates after appropriate processing. In 3 some agglomerates are shown.

2.1.3 Causes of Waving

• • Bedruckstoff
• • Gravurart, Rasterwinkelung und Form der Näpfchen
• • Temperatur und Viskosität der Farbe

The reasons for waving are not the subject of this work and are therefore treated only briefly. An exact investigation has not yet taken place on this topic or has not yet been published. But fundamentally, in the context of this work, I found out that several factors influence waving decisively:

• • substrate
• • Engraving type, screen angle and shape of the cups
• • Temperature and viscosity of the paint

It There is still a great deal of research into how and to what extent these and other factors influence waving.

It was noticeable that in the investigation of the elongated wells the preferred direction the agglomerates were detected in the compression direction.

Likewise one could with enlarged photographs (see 4 ) clearly see that at SC-Papie (Supercalandered supercalendered) agglomerates partially formed along paper fibers along. The fibers in this case may also affect the shape and size of some agglomerates.

Even the unprinted paper may possibly already provide information about the expected Waving, because here are the individual fibers depending on the paper as in 5 clearly visible.

3 The Waving Index (Wi)

Unfortunately There is no single approach to assessing this phenomenon. Often is a rank order created by a few observers and with Key figures provided. This approach is not very objective, hardly leaves a comparability over longer periods too and is time consuming.

Therefore are reliable Characteristics neither customer for the evaluation of a printing result, still printing companies for the evaluation of papers regarding the phenomenon or paper manufacturers for the improvement of your products.

Around the waving phenomenon in its quality to be able to determine it requires a metrological method, the one on the visual Impression-related measure for a Output measured object. This measure becomes referred to as the Waving Index and will be abbreviated to Wi below.

3.1 Claims to the Waving Index

Of the Waving index should output a number that says how strong the Waving is pronounced in a measured pressure test or a series of measurements. The focus is on the visual perception of the viewer.

These Number should make it possible Objectively compare different measured samples and give them one certain quality assigned. Ideally, the Wi also has different engravings, raster and Tonwertfelder a global significance, so that comprehensive Comparisons possible are. Also diagrams of the waving quality in the individual tone values a paper type would be representable - one Art Waving Characteristic. With the Wi, it should also be possible to make statements about the Substrate or other printing parameters.

3.2 Claims to the method of determination of the Wi

to Determination of the Wi should be preceded by an algorithm and be applicable reproducibly. This can e.g. in the form of a program, Plug-ins, macros or similar done so that the operator the exact operation of the algorithm not know, but can use it intuitively.

Ideally the Wi is over determines a one-shot technology, that is, the operator performs the Measurement on a pressure test through and receives immediately after operation of the Algorithm the Wi. Furthermore it should be possible be, more and more detailed information about to get the measured sample. This can be in the form of an issued Be reports that provide important details in a short

One Program, plug-in or macro the advantage that the measurement if necessary also on the spot of companies carried out if they have a suitable measuring device. In contrast to many expensive special measuring instruments could for the Waving e.g. a scanner are used. The corresponding program could then get the data from a TWAIN interface, or it will as a plug-in or Macro for common Programs like Photoshop or Excel installed.

Also a calibration of the meter would be ideal and could be done in Implementation be realized, so that measured values of different Systems would be comparable.

3.3 Description of the print samples

For the investigation, monochrome prints were used which were printed on a special gravure printing machine under test conditions in the gravure laboratory of the Bergische Universität Wuppertal. In addition to the investigation of waving studies were already held on the areas Mottling, Missing Dots, density, gloss and pressure. From each paper sample several copies were available for the investigations.

Of the Printing cylinder was externally engraved with a special template, so for various investigations specially used for this purpose engraved fields could become. With the printing form were 10 different paper patterns printed. These papers are manufactured in three groups divided (see Table 2). Papers 1 to 5, 6 to 7 and 8 to 10 were wound on a roll in succession to preferably print all three groups in one pass. Consequently should a repeated setup of the machine be bypassed and the printing conditions such as Room temperature, web speed and viscosity kept constant. With constant parameters is a better one Ensure comparability of different paper patterns. To minimize fluctuations, were before the actual papers Several hundred meters of waste paper printed. The used black Color is metrologically and visually best detected.

3.3.1 The papers

The paper patterns have been numbered from 1 to 10 and remain the same throughout the following investigation:
Papers 1-5, 6-7 and 8-10 differ within their groups in composition or different preparation. Exact details were from

paper Manufacturer not published, but the effects are evident in the various studies become visible.

3.3.2 The printing form

A smaller section of the printed image is in 6 shown. You can see three areas running in the printing direction, which differ in the type of engraving. The left print area was created with the Engraving Cup 0 (compressed cups, grid 70 L / cm). The middle area was created with the engraving type Cup 2 (elongated wells, grid 70 L / cm) and the right area with the engraving type Cup 4 (square wells, grid 100 L / cm).

• • ein 4 cm langes Volltonfeld, einen 25 cm langen 10%igen Tonwertfeld mit einem integrierten, 0,5 cm breiten Tonwertverlauf von 0% über 100% bis 0%
• • ein reproduziertes Halbtonbild mit 5,3 cm Länge, einem weiteren 27 cm langen 10%igen Tonwertfeld
• • einen Bereich mit 14 Tonwertfeldern á 2 cm Länge für die Werte 5%, 15%, 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 70%, 80%, 90% und 100%

Each cup area is about 4 inches wide and contains the same elements:

• • a 4 cm long solid field, a 25 cm long 10% tone field with an integrated, 0.5 cm wide Tonwertverlauf from 0% over 100% to 0%
• • a reproduced halftone image of 5.3 cm in length, another 27 cm long 10% tonal field
• • an area with 14 tonal fields á 2 cm in length for the values 5%, 15%, 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 70%, 80%, 90 % and 100%

Also located between Cup 2 and Cup 4 a strip that alternately fields of 1 cm in length with 100% and 0% area coverage having.

3.4 The judgments to be assessed and measured fields

3.4.1 Selection of the field

Because the waving as a phenomenon occurs in the midrange, the selection of usable ones is limited Tone value fields in the range 40-70%. As a basis for measurements In the context of this study, a single field is intended to be used rated and measured on all papers. An investigation, the all Fields and cups included, has not been possible in the context of this work, so that a meaningful determination on a field was necessary. For measurements of other cups and Tone values will be needed for further project work. There will be then show to what extent results of this investigation on other cups and Tonal transferable are.

When The object under investigation was the 55% field in Cup 2 (elongated wells, 70s grid). In this field there is no complete flood and single Pressure points are still recognizable. This also causes the agglomerates as disturbing Clearly perceiving objects.

3.4.2 Preparing the samples

Out For the various proofs, one sample per paper was selected. Important was that the samples submitted for visual evaluation at the same time as the basis for metrological evaluation were used. Thus, it was ensured that visually evaluated Aspects can also be found in the metrological results had.

The The samples to be evaluated should have as large an area as possible in order to be measured also visually representative to be evaluated. Furthermore should this possible few other errors exhibit. If necessary, use additional fields for consideration to be able to the two neighboring tonal fields (50% and 60%) were also applied to the sample strip.

The samples to be evaluated were cut out and mounted on a 135 mm long and 25 mm wide white cardboard strip. 7 shows a finished strip. The attachment used was a double-sided, thin adhesive layer that did not penetrate the samples. As a result, the printed image was preserved and was perceived unchanged. By trimming from 40 to 25 mm wide, the labels for the tonal values were omitted and it was annoying print strips positioned at the edge, which were not directly in the center of the image. This reduced irritation, making it easier to focus on waving. Cracks or creases were also avoided when mounting.

Around fading, yellowing and mechanical damage were prevented the 10 strips protected from light and safely kept.

3.5 Visual description of the samples

at a consideration of all samples falls in a direct comparison on that they look different bright. If you choose a viewing distance of 20-25 cm, you can with good lighting recognize the waving very clearly. The Waving leaves paper 7 and 9 very restless appear, because here are clearly large, disturbing structures visible. These are still noticeable even if you have the viewing distance 40-50 cm increased. paper 3, 4 and 8, on the other hand, are very quiet because they are not big agglomerates recognizable. Only on closer inspection will agglomerates become visible, but moving in a smaller size frame than paper 7 and 9. If you increase the viewing distance, these areas appear very quiet and almost trouble-free. Put the papers 3, 4, 7-9 right next to each other and looks at them comparatively from a viewing distance from 40-50 cm, clear differences in the area are visible. The other papers are located qualitatively between these papers: agglomerates are recognizable, but do not immediately come across as extremely annoying.

Of Furthermore, contrasts occur within some tonal fields, because one sees surfaces there with groups of, sometimes small, agglomerates, next to which are lighter surfaces are located. These local contrasts are also still out a big one See vision distance clearly.

Besides, it seems it so that the eye from a certain viewing distance several neighbors tight agglomerates no longer perceives as individual structures, but to one this summarizes. This creates structures partially extending several millimeters over the printing surface.

If not only the 55% field, but also the subsequent 50% and 60% fields, then you can see a significant increase in the Wavings with increasing tone value. It varies depending on the paper strong. Papers that have better quality in the 55% field can be poorer in the 60% field.

4 Visual ranking of the samples

Around one possible objective qualitative rank order for the expression of Wavings were the samples are given to several professional observers from the printing industry submitted for assessment. The aim of the survey was to give orientation for later metrological To get examinations, because an algorithm should spend a WI and reflect in it the visual impression in a number.

4.1 Determination with the pairwise comparison (PVi) and direct comparison (DV)

Around To form this order of rank, became the so-called method of pairwise comparison (in the literature also as pair comparison index designated) applied. Furthermore was added performed a direct comparison of all.

Of the Pairwise comparison will follow with PVi, the direct comparison abbreviated to DV.

at The rank order ranges from 1 to 10: Rank 1 represents the best and rank 10 the worst quality of the samples.

4.1.1 Method of pairwise comparison (PVI)

One Observer determines based on a defined criterion two simultaneous pressure samples from a series of Samples that are better or both are equally good. This quality criterion was in the following attempt exclusively waving. On it was expressly pointed out to other disorders, such as. To exclude stripes from the assessment.

Around Achieving objective results requires the highest possible number of observers be questioned. A lack of objectification by a low Number of observers may be due to the professional qualifications of the Observers are counteracted. The investigation was carried out with five Observers.

The observers' assessment was made under the same conditions: all the individuals were sampled in a viewer's console (see 8th ), which was illuminated with D65 light. In order to be able to judge as well as possible, the samples to be compared should be as close as possible to each other. So the test strips to be compared were always kept right next to each other. They could also be moved to avoid disturbing reflections from the lighting.

Each sample is compared to each other. It must be decided which sample is better or whether both samples are equally good. The results are entered into a quadratic dominance matrix D [g _ij ] of the nth order (n = number of samples) with the elements g _ij using a 2-0-1 point system. The scoring system looks like Table 3 ([3], pp. 293-302; [4], p. 45, p. 46).

The total number N of pairs to be compared is calculated according to this formula: N = (N 2 - n) / 2

at The present 10 samples therefore had 90 pairs per observer be compared. Because the experiment takes a lot of time and the observer overwhelmed could be should use a pairwise comparison of a maximum of 10-12 samples be compared ([3], pp. 293-302; [4], p. 45, p. 46).

Since each sample is compared to each other, there are corresponding elements: The upper diagonal half contains the judgments j versus i, so that the fields lying symmetrically to the main diagonal contain the inverse judgment. Corresponding elements total gij + gji = 2. If this was not the case, the same sample pair was rated twice differently (see 9 ). This procedure is called intransitive judgment ([3], pp. 293-302).

For N observers, N dominance matrices can be added to a combined dominance matrix R [rij] (see 10 ). The row sum (Σz) shows the absolute frequency of preference of a sample. This results in the classification into the rank order, which can be represented either isomorphically by an average calculation (Mw) or absolute (RPVi) (see 10 ) ([3], pp. 293-302).

The intransitive judgments can also be calculated and displayed in a combined matrix, so that it is easy to deduce which pairs of samples were judged most differently. For better illustration, couples with intransitive judgments were highlighted in color, depending on their thickness (see 11 ).

The ascertained ranking order (by mean value) was plotted into a bubble diagram for better illustration and the standard deviation was used as the size for the bubble diameter. So you can clearly see which papers were subject to high uncertainties in the valuation ( 12 ).

4.1.2 Direct comparison (DV)

After this the observers have performed the pairwise comparison, they should then one Perform a direct comparison, i.e. all 10 samples should be compared directly and be ranked. The attempt was also performed under D65 light.

sense The direct comparison was to check if one also this method for can use the formation of a rank order, because it is to perform a multiple faster than the pairwise comparison.

The rank orders of all observers were combined into a combined ranking order and then plotted into a bubble diagram ( 13 ). Again, the standard deviation is used as the size of the bubble diameter, so that uncertainties in the rating are clearly visible.

4.2 Results of PVi and DV

If you look at the results of the two methods, you see that is judged similarly on the papers. Within the rank order some places are reversed, but in both methods three quality groups can be distinguished. In these three groups the same papers are found in both methods:
If the two methods are correlated with each other, the three groups are also clearly visible in the diagram. Since the coefficient of determination R ² = 0.939, it is also possible to use the direct comparison for determining a ranking order for time pressure, since the results of both methods are very similar ( 14 ).

The two rank orders can also be considered comparatively in an ascending order of rank ( 15 ).

Out The diagrams can be determined by the standard deviation for each method the papers with the biggest uncertainties read out. These are PVi papers 10, 5 and 2 (all qualitative medium-value papers) and DV paper 8, 9, 7, 2 and 1. Even with big ones Fluctuations remain the papers but all within their quality group.

These Uncertainties show that there are several aspects in the Assessment included. These aspects seem to have been individually evaluated by observers to be. This is explained in the section "Digitizing the samples" on page 28, because the different aspects are metrologically must be captured and incorporated into the algorithm.

Very it was striking that among the well-performing samples all the dark ones Papi-re were found and among the poorly performing papers very bright papers. To find a connection, the Results in section »Excursion: See «in more detail.

5 Excursus: See

In a graph (Figure 16) with a correlation of the rank PVi and the brightness L *, one can clearly see that the well-rated papers are darker than the mediocre and poorly rated papers. Some of the poorly valued papers are light to very light. This is also clearly visible in a direct comparison by placing all the strips with the printed samples side by side.

A similar picture is obtained by correlating the average gray value of the samples with the rank PVi. However, the coefficient of determination is only R ² = 0.5283.

These results show that an observer in a visual assessment of the Wavings not only only assessed the waving, but obviously also the brightness of the substrate plays a dominant role in the rating. This assumption can be confirmed by literature on the human eye.

5.1 Optical conditions

The Perceptibility of small structures taking into account the brightness is subject to several conditions, here briefly below consideration the meaning for the waving will be performed.

5.1.1 Visual acuity and resolution of the human eye

• • Räumliche Sehschärfe
• • Kontrastsehen

The term visual acuity can be subdivided into two sub-concepts that play a role in visual acuity:

• • Spatial visual acuity
• • contrast vision

The spatial visual acuity allows it to recognize shape details and the contrast acuity relates on the ability Perceiving brightness differences ([5], p. 123). These two Factors affect the resolution of the Au tot.

The structure of the retina allows multiple information about an object to be conveyed in parallel. The total information grows, the smaller the still separately perceived details ("resolved") The absolute size of the still resolved objects depends on the viewing distance (Sehweite) Therefore, the minimum visual angle for two still separately recognizable structures (edges, lines o. is Ä.) as the angle w _min indicated. the smaller the angle w _min, the greater the resolution of (see FIG. 19 ). However, as already mentioned above, this is influenced by the object shape, the contrast and the illuminance ([6], p. 90).

5.1.2 Gray scale viewing and color vision

In particular, because the evaluated samples are monochromatic gray areas, it should be noted that in greyscale viewing, the spatial resolution due to the excitation of all cones is significantly higher than in color vision (see 20 ). For example, edges and structures with fine details can be easily recognized, but slowly changing gray-scale gradients tend to be poorly recognized ([7], p. 197).

5.1.3 Contrast vision at different brightnesses

As already mentioned in the section "Visual acuity and resolving power of the human eye", the spatial resolution of the human eye depends on the luminance. In 21 Illustrative curves of spatial contrast sensitivity are shown. They show what contrast (modulation) a sinusoidal light-dark grating must have at a certain spatial frequency in order to be recognized (in the figure called brightness modulation degree). At high luminance levels, sensitivity is highest at frequencies around 10 lines per degree. At lower and higher frequencies the sensitivity decreases ([2], 5.139, p. 139).

5.1.4 The special problem of surface pressure (under the aspect of the Wavings)

On printed surfaces (in this case with the uniform tone value fields) Waving is particularly visible, because falling on a surface disorders and irregularities by e.g. Agglomerates especially on. Presumably, agglomerates fall less in the case of prints that are already restless due to the motif on. That's why surfaces are also ideal for the visual and metrological evaluation.

5.2 Influence of brightness on perceptibility of the Wavings

The size of a pressure point is in the 55% raster, depending on the paper in Cup 2 between 97 and 148 microns. By engraving the pressure points are different in their extent in length and width. The width is between 97 and 122 microns, the length between 125 and 148 microns. The distance between the individual pressure points is well below 100 μm.

The guideline of the resolution of the human eye is careful observation at 2 'or 1/30 (see 19 ). At a viewing distance of about 25 cm under ideal conditions, the resolution of the eye is therefore about 145 μm. As a result, it is no longer possible to perceive the individual pressure points in the mid-tone range - if they are not already flooded - as individual points, because the distance between them is too small for the human eye under normal conditions. Only with a Nonienbetrachtung under optimal lighting conditions still individual points could be perceived (see 19 ).

considered If you place the measuring fields on the individual samples, you can use the individual pressure points actually not very difficult or under bright lighting dissolve. To the one, because the contrast between the printing dot and the printing substrate a on the other hand, because the pressure points in This grid of raster fields are too close together (too high spatial frequencies). But as soon as a certain number of pressure points agglomerate, may vary depending on the number of agglomerating pressure points and visual conditions the limit to visibility is clearly exceeded.

• • Mit steigendem Kontrast ist es möglich, höhere Ortsfrequenzen wahrzunehmen. So können im Gegensatz zu kontrastarmen Proben (z.B. Papier 3 und 8) auch kleinere Agglomerate bzw. Störungen wahrgenommen werden.
• • Auf hellem Papier werden Störungen generell sehr viel deutlicher wahrgenommen, als gleichgroße Störungen auf dunklerem Papier. Innerhalb einer gleichmäßigen Fläche wirkt ein Agglomerat auf einem hellen Papier also sehr viel unruhiger als auf einem dunklen Papier.

If, as with paper 7, the contrast between the printing dots or the agglomerates and the printing substrate is high, this has a decisive influence on the visual evaluation. This contrast influences the assessment of two aspects that are central to the further course of the study:

• • With increasing contrast, it is possible to perceive higher spatial frequencies. Thus, in contrast to low-contrast samples (eg paper 3 and 8), smaller agglomerates or disturbances can be perceived.
• • On bright paper, noise is generally perceived much more clearly than the same amount of interference on darker paper. Within a uniform area, an agglomeration on a light paper looks much more restless than on a dark paper.

For paper types So this means that a poorly printed bright paper in terms of waving is rated significantly worse than one comparable printing dark paper. Light paper reinforces the effect of printing problems, while a dark paper them softens. Especially for Paper manufacturer is likely this will be an interesting finding.

Also when translated into an algorithm based on visual ratings based, must these findings are integrated. This is described in the section »Integration paper brightness in the algorithm «.

5.3 Integration of paper brightness into the algorithm

There a visual rating very much of the brightness of the paper is influenced, it makes sense, first the Waving Index into one Brightness value and a structure value to divide.

In a measuring method, after normalizing the measurements to a brightness, in the algorithm extracted from the measured data, the pure structural data of the agglomerates be, so that brightness for a comparison of the results with the pure structural value irrelevant is. As a result, the papers are mutually exclusive with respect to the Wavings comparable. To evaluate their visual effectiveness, would have to be in another step, the brightness be taken into account. The eye can, as in the section »contrast shown at different brightnesses «, in certain (paper-based) contrast ratios not all agglomerations and irregularities detect that would be detected by digitizing the samples. A clean Structural analysis would so do not correlate with the rank order.

Around independently evaluate the structures of the Wavings from paper brightness to be able to a new order of rank is necessary, which exclusively refers to the structures. This requires a normalization to a brightness occur. The formation of a new structural ranking will be in the section »Education a structural ranking « Page 40 in the context of the algorithm description.

6 Digitization of the samples

• • Digitalisierung mit Kamera und Mikroskop
• • Digitalisierung mit einem Scanner

In order to be able to submit the print samples to an automated examination, they must or Parts of it are digitized. These data can then be further processed with appropriate programs. For digitization, two different options were used:

• • Digitization with camera and microscope
• • Digitization with a scanner

each Method has several advantages and disadvantages, which are explained below.

6.1 Digitization with camera and microscope

A possibility to digitize the samples is the recording via a microscope to which a Camera is connected. This camera has a TWAIN interface with Connected to a computer, making the recordings directly in one Program (such as Adobe Photoshop or Image-Pro Plus) and created can be processed.

For the microscope image can on the device the parameters magnification, Aperture, sharpness and lighting are set. The lighting is done via a separate lighting unit, which has several different lighting levels features.

The camera outputs a picture of 768 × 562 pixels. When importing images, various settings can be selected (see 23 ).

One greater The advantage of this method is an inserted live image, which is the to microscopic object shows. This makes it possible before to search for a section of the recording and to position it exactly. Also the different possible ones enlargements and adjustments by aperture or lighting are for investigations an advantage, because you can choose the optimal one depending on the object recording quality can achieve.

But The system also has significant disadvantages. For example, e.g. the recording or measurement section (caused by the lens on the microscope) very much limited, and it is further reduced with increasing magnification. Therefore, depending on the magnification always Only a small portion of the pressure sample will be captured and will take multiple shots over that entire field needed. The time required for Multiple shots is not insignificant. Besides, it is possible that Agglomerates greater than the measurement section are and therefore not as a complete object be recorded. When visually viewing the shots is partly the waving no longer recognizable due to the enlargement. Also solve multiple measurements the problem is not because no seamless capture is possible. The only way to get around this problem would be a larger recording area.

Also, the different measurement parameters may not be reproducible or calibratable. For example, the aperture is infinitely adjustable via a rotary knob. Even a noise of the camera has problems. To investigate this, a picture was taken several times and a histogram was output. These histograms were compared, and the standard deviation clearly indicates camera noise (see 24 ). The noise shown can additionally be influenced by fluctuations in the lighting or the ambient light. On the video track the signal deviation is about 3%, but can reach up to 100% in the high and low tone values.

Under the consideration, that the algorithm ideally works without any special equipment the use of camera and microscope is a disadvantage, because this equipment belongs not to the usual equipment in printing houses and paper mills, so this may be expensive to buy.

• • Kamera: Licht 1/4, Blende 22, Vergrößerung 0.8
• • Import: Alle Einstellungen auf Standard, Y8-Graustufen, PAL

From each paper 6 pictures were taken under the microscope. The images have been taken at various points on the 55% field in order to capture the largest possible area and to minimize overlaps. The faulty area with the strip was left out. The same conditions applied to all shots:

• • Camera: light 1/4, aperture 22, magnification 0.8
• • Import: All settings to default, Y8 grayscale, PAL

Because all recordings had the same conditions, they differ significantly in their quality. While with bright papers individual pressure points are clearly recognizable, with dark papers no dif referencing Digitization of the printed image more recognizable.

6.2 Digitization with a scanner

A different possibility to digitize the samples is the use of a commercial scanner. A scanner has opposite Microscope and camera some big ones Advantages. For example, it is possible to have a very large reception area in one Scanning - ever according to size of the original even several samples at the same time. The different magnifications of the microscope correspond to the flexible scan resolution, the depending on the device may have a different maximum, which is different from the optical and mechanical quality depending on the system is. At a correspondingly high resolution, so are individual Raster points well digitizable.

One scanning several samples in parallel takes little time and it is no need for multiple shots because the entire sample area as a whole can be detected. Cuttings and edits can then done by software on the computer. A commercial Scanner belongs in print shops to the basic equipment, so that no expensive equipment additionally purchased would have to be. In addition, then possibly even cross-system (Windows, Apple Macintosh and Linux / Unix). There is also the possibility, to perform a calibration, so that also results from different Scanners and systems can be compared.

Consequently a scanner seems an ideal device for a one-shot technology to be, because he is almost everywhere available, cheap and calibratable with appropriate software, and with "one Push of a button "is a big area measurable.

To digitize the print samples for this investigation with a scanner, the Agfa DuoScan Agfa machine was used. All samples were placed next to each other and recorded in a scan (see 26 ).

The scan resolution was 4000 dpi. With this scan resolution, a fineness of 6.36 μm per pixel can be achieved. This ensures that enough pixels are available for rendering the halftone dots. Subsequently, a 28972897 pixel area was taken from each test strip so that a sufficiently large file per paper was available, which could be further processed by metrology. Care was taken to ensure that the disturbing pressure strip, which clearly penetrates the printed image on almost all samples, did not appear in this area. Although this did not cover the entire area, it was much larger at 18.4 × 18.4 mm than when taken with a camera and a microscope. In a visual comparison, it was found that the waving phenomenon is sufficiently well visible. Even donuts are clearly visible on the pictures (see 27 ). Even with dark paper, this method provides a similarly good and differentiated digitization of the printed image as with a microscope and a camera.

7 alternatives to the proposed procedure

This Section discusses alternative approaches to developing an algorithm. The final procedure later gets detailed in the section »The Algorithm and its Evaluation «on page 40. For time reasons only one procedure was completed implemented, but the different approaches are here below briefly described.

7.1 Considerations on the process options

Around to find a way of proceeding, were various considerations hired. These went in different directions. So, that was a consideration starting from an ideal grid, the disturbances in the grid by agglomerations out. With a Fourier transformation, which is also part of image processing programs, one can do such calculations relatively easy to perform and apply filter functions. Another consideration was about the Proportions of the printed area and the histogram function to make statements about waving. A third consideration, which then as more detailed Was followed up by a procedure for distribution, Form and effect in combination with a matrix or mosaic function.

• • Fast Fourier-Transformation und Filterfunktionen
• • Histogrammfunktion und bedruckte Fläche
• • Verteilung, Form und Wirkung in Verbindung mit einer Matrix- oder Mosaikfunktion (s. Kapitel 8)

Here are the approaches at a glance:

• • Fast Fourier transform and filter functions
• • Histogram function and printed area
• • distribution, shape and effect in combination with a matrix or mosaic function (see chapter 8)

at all procedural approaches visual perception should be a focus. Both first two approaches, the more detailed below explained However, this focus came into the background (the influence brightness and contrast were not taken into account).

7.2 Histogram function and printed area

For an ideal grid of isolated pressure points, two bars are visible in a histogram: one bar represents the frequency for the tonal values of the pressure points, the other represents the frequency of the tonal values of the substrate (see 28 ).

However, the real pressure point behavior is different, because the autotypic printing process makes the transition from the printing dot to the printing material not sharp, but rather fluid. Therefore, a histogram of an ideal real raster would have a smooth distribution of two-peak tone values. For some papers, this can be clearly seen, while other papers have a steadily falling off the peak curve (see 29 ). The histograms were generated from the average of the 6 images taken with microscope and camera.

To obtain better differentiation and comparability in the histograms of the digitized print patterns, an 8x8 blur (Gaussian blur) was first performed on each paper swatch and then the histogram was taken from the average of the 6 scans (taken with camera and microscope). An overview of all histograms is in 30 shown.

in the Ideally would you take a single bar at a certain gray value can, but as the real histograms show, is a narrow and high distribution to be found in the gray area. Starting from these curves was the Question discusses whether one can do such a histogram Statement about the waving can do.

• • Höhe des Maximums
• • Standardabweichung vom Mittelwert
• • Mittelwert
• • Gesamtbreite am Full

The histogram curves can be characterized by various features:

• • height of the maximum
• • Standard deviation from the mean
• • Average
• • Total width at Full

However, a closer look at the maxima does not show a correlation with the ranking order PVi (see 31 ). The other features also show no correlation to rank order.

Also a division of the papers into the classes is good medium bad not possible by the characteristics of the curves in this way. There none of the results from this approach was meaningful until that point, was due to time constraints no further investigation over the previous results.

Perhaps could This approach was taken up again in further investigations when certain aspects are changed and taken into account become. Therefore the histograms were used only small sections, would be one larger contiguous area (e.g. the digitized with the scanner print sample) more meaningful. This would be also the problem of a different recording quality can be corrected. Furthermore would the different paper brightnesses in the investigations with be included. Furthermore, the method of blurring to discuss - possibly There is a more skillful way to get meaningful curves over the To achieve histogram formation.

7.3 Fast Fourier Transformation (FFT) and filter functions

One Another approach to the quality of the Wavings is to perform a Fast Fourier Transform (subsequently abbreviated to FFT) in conjunction with filter functions and an inverse transformation. The explanation of FFT will be as short as possible here held to provide a basic understanding. A detailed Description can be read in the specified literature.

7.3.1 The Discrete Fourier Transform (DFT)

A discrete Fourier transform (DFT) counts under the term image processing the global operations, i.e., all pixels of the input image needed to the result for to calculate an output pixel. Like the digitized print samples digital images are so-called 2-dimensional spatially discrete signals. The procedure for a 2-dimensional Fourier transformation is formally based on the one 1-dimensional.

By summing suitable sinusoidal signals, one can generate any non-sinusoidal (periodic) signal (see 32 ). By a Fourier analysis, any non-sinusoidal signal can be decomposed into sinusoidal components and typically represented in the frequency domain. When applied to images, one speaks of the spatial frequency range, the original image is located in the local area. The frequency range (also spectrum) shows the sinusoidal components of a non-sinusoidal signal ([9], pp. 119-121). In this frequency space many operations, eg filter functions, can be carried out more easily.

at digital images, the Fourier analysis is performed by the discrete Fourier transform and calculates a Cartesian representation of the spectrum. This can be well processed by the computer. To make this presentation clearer To shape the representation can be converted into the polar representation become ([9], p.121).

7.3.2 The Fast Fourier Transformation (FFT)

A discrete Fourier transform is very computationally intensive and the fastest algorithm for realizing the DFT is the so-called Fast Fourier transformation ([9], p.127).

7.3.3 Apply FFT to Raster

Since the FFT refers to periodic variables that can be decomposed into elementary sin and cosine vibrations, it can be applied to a raster to eliminate it. A raster consists of periodic signals so that it can easily be converted into a spectrum, amplitude or phase representation via an FFT using appropriate programs such as Image-Pro Plus, Ad Oculus or a plug-in in Adobe Photoshop (see 33 ) to perform further operations.

These further operations can e.g. Be filtering. So it is possible with high or low pass filters filter out high or low frequency components. Also individual Filtering can be done via special Input masks or directly by drag-and-drop in the spectrum. With these filter functions could For example, you filter out a grid so that only the consist of irregular agglomerations stay.

By an inverse transformation the filtered spectrum is converted back into an image representation and can then follow be evaluated or further processed.

But the FFT has some drawbacks that have led to the discontinuation of this procedure for waving processing. In the case of the FFT, due to the algorithm, for example, edge effects (visible in 36 ). Other issues are discussed in the next section.

7.3.4 Application of the Fast Fourier Transform on the print patterns

The Program Image-Pro Plus offers a ready-made function for Fast Fourier transform so that's the possibility consists of further processing digitized images with this program function. There are many other possibilities filter functions (e.g., high or low frequency filters), or also the possibility perform operations directly in the spectra shown.

The images taken with the microscope were only partially suitable for one Further processing with the FFT, because of the different Brightness and paper good, it is difficult to get the individual Grid points to find sharply demarcated. The individual grid points also have no uniform shape.

A good further processing with the FFT seems only after a good preprocessing of the images to be possible, so that the printed areas differ as well as possible from the unprinted areas. This can be achieved by applying different filters to the images. Really, it is very difficult to highlight the printed areas. A macro that is equally applicable to all papers, can not exist, because they differ too clearly. So there must be a paper-specific or possibly even an individual pre-processing, in order to continue working with a paper-specific algorithm. This would be a hindrance for a proposed one-shot technology rather.

Out time constraints was therefore only an example of an ideal grid with a prefabricated Macro edited. This macro succeeded on an ideal grid (see the following section »Using AutoFFT) worked but not with the digitized print samples, so this one Approach was not followed up.

7.3.5 Use of AutoFFT

There there are many pre-made on the manufacturer side of Image-Pro Plus Macros with various functions for downloading have I am there for a solution which automatically recognizes regular structures in pictures and can eliminate them. The tool is called "Automatic Filtering Using a Fast Fourier Transform "(AutoFFT).

It allows it, regular structures while removing the underlying image details to obtain. A scaling factor to be entered defines the effect the FFT on the regular structures.

The procedure is shown here on an ideal grid with artificially produced agglomerates. The ideal grid was created in Adobe Photoshop and provided there with appropriate artificial agglomerates. This image was then read into Image-Pro Plus (see 35 ).

The read-in image is now further processed with the AutoFFT tool and, following an automatic back transformation, a further-processable result image is output, depending on the settings (see 36 ).

For example, a threshold separation can now be made on this image, which only highlights the flooded areas between the grid points (see 37 ).

The Result clearly shows the elaborated flooded connections between the individual grid points. It would be very easy to process and as a basis for additional measurements can be taken.

Unfortunately So far, this method has only been used on an artificially created grid Agglomerates successfully applicable. To use this method at real Perform data to be able to would the pictures are correspondingly expensive to get prepared. It is not excluded that there is a possibility There is time to evaluate waving with an FFT Within this work, it was only possible to show this approach and on the basis of an artificial example created the possibilities display. Also over The visual aspect of the FFT is in need of discussion, because obtained results should correspond to the vision of an observer.

8 The algorithm and its evaluation

As in the section »Integration paper brightness in the algorithm «on page 28 paper brightness affects the visual rating. Therefore, it makes sense first the waving index into a brightness value and a structure value to divide. The brightness value can be measured easily from the L * or Y values. To determine a structure value, must first a structure ranking order are formed. This one will be back created by technical observers. The samples to be evaluated must be brightness normalized and extracts the structures to be evaluated from the print patterns become.

These Structure evaluation serves as a reference for the metrologically to be recorded Results, which are also due to these brightness normalized images must be made.

8.1 Formation of a Structural Ranking

Around To obtain a rank order of the structure images had to be in a first step, the structures from the digitized print patterns extracted and put into an assessable form. These structures were assessed by observers in a further step.

8.1.1 The generation of structural patterns

For the generation Structural images became the print patterns digitized with the scanner used and all steps executed in the program Image-Pro Plus. The Examples in the illustrations are based on paper 6.

The central question that posed in an extraction of structures must be, is: What does the viewer see? This approach of a attempted simulation of sight served as the basis for the generation of Structure images.

First of all The digitized print samples of the original size were 2897 × 2897 pixels on 1000 × 1000 Scaled pixels. Although this information has been lost, but this size is for further processing and as a basis for Sufficient measurements. The scaling affects a reduction the computing time for Filtering operations off.

at a viewing distance of about 20-25 cm takes a viewer the individual halftone dots on the print pattern are not true, but instead they become an area dissolved. Within this area the agglomerates are disturbing Structures perceived. Therefore, these structures had to be extracted so that they could be evaluated by the viewers alone.

For this reason, a filter was first applied, which dissolves the grid while maintaining the agglomerates as surfaces. This filter is called median (PROCESS → FILTERS → ENHANCEMENT → MEDIAN) and was applied to the print samples in the size 5 × 5 in four applications (pass: 4) (see 39 ). The strength and number of passes were checked visually: It was important that the impression of agglomerates was maintained. The filtered images were viewed on the monitor both from near and from a great distance and compared with a view of the real pressure sample. The median filter does two things: it replaces a central pixel with the mean of its neighboring pixels, and at the same time creates a sort of blur.

If one compares the original image with the filtered image, one sees clearly the disturbing agglomerates as dark surfaces (see 40 ).

In a further step, a threshold value separation was carried out (PROCESS → TRESHOLD). Since the measured standard deviation from the average gray value for all print patterns was approximately 5, this value was used as a reference and the separation threshold for all samples was set to the threshold value = (average gray value) -5. This was an attempt to make a brightness normalization. Again, this was compared visually up close and away with the real print patterns. In order to get a correct view, Black on White must be selected during the preview (see 41 ). After performing the threshold separation, an inversion of the image had to be performed (ENHANCE → INVERT IMAGE) - this made the structures of the agglomerates black.

Around to filter out the smallest structures that are also not visually detectable were after the threshold separation of the filters Close (PROCESS → FILTERS → MORPHOLOGICAL → CLOSE) applied. For this purpose, the options 22 Square and 4 Passes were chosen.

The now left remaining games are also in a direct comparison in the Pictures of the previous processing steps visible.

Around Edge effects that have arisen during image editing for the rating and exclude measurements the pictures were cut from the format 1000 × 1000 to 800 × 800, that is, a trimming of 100 pixels was performed on each side.

8.1.2 Visual description of the structure patterns

In 43 All ten structural patterns are arranged according to their paper. Here the structures extracted from the print patterns are clearly visible.

Because for the latter mathematical and statistical coverage of structural features important is what exactly is captured due to visual effectiveness should, every structural pattern with its visual abnormalities briefly described briefly in Table 5 below.

On On closer inspection, the structural patterns seem to have a preferred direction to flood. The flooding forms in this field in cup 2 tends to be in the pressure direction (this would also have to be measured be detected by a narrow in the printing direction displacement matrix on the Picture is running. Furthermore It is important to find out if it is a preferred direction for the others Gaits and in which direction this lies).

8.1.3 Formation of a Structural Ranking with the direct comparison

For education Ranking was due to time constraints the method of direct comparison used. As in the section »Results of PVC and DV « Page 21, this method can also be representative Deliver results.

The 10 structural samples were submitted to six technical observers for evaluation. You should rank in a direct comparison of 1 (good) to 10 (bad) with regard to a criterion. This criterion was as homogeneity, uniformity or rest of the pattern defined. It was also pointed out that the amount of blackening for the Rating is not crucial, but only the visual impression a homogeneity. The laid order should also be from a greater distance (about 1 m) be considered to better assess difficult to classify samples to be able to.

8.1.4 Results of the Structural Ranking

The results of the direct comparison were recorded in a table and then transferred to diagrams. The result of the graphs shows very clearly that the extremely good and poorly rated papers (paper 4 and 7) were rated very clearly (with paper 7 the standard deviation for the representation was artificially increased) (see 44 ). Other papers had large uncertainties (Paper 2, 6 and 8), but the variations were only within their quality group. These three groups (good - medium - bad) are also visible when sorting in ascending order (see 45 ).

Out The results show that even-looking images are very good be evaluated, with uniformity both with much blackening (Paper 4) as well as with little blackening (paper 8) can. It seems to be important that the existing blackening evenly the area is distributed. But when there are many white spaces, it seems this for the observers are difficult to evaluate (paper 8, paper 6). This confirmed also in talks with the observers.

There in a first pass accidentally the uncut 1000x1000 pixels huge Pictures submitted for evaluation had a second pass with the 800 × 800 Pixel big Samples performed to ensure that the metrologically and visually evaluated Samples are identical. This error showed that samples were rated worse with a strong edge effect in the first round as in the second round, where this edge effect has been cropped was.

8.2 Detection of structural features

To The previous visual assessment had to find a procedure that captures the structural features mathematically and statistically. The result of such a measurement must be in accordance with the structure ranking order correlate well. In order to grasp these structures, different approaches were chosen and compared with the result of the structure evaluation. Hereinafter Section will be first the different possible ones measurable structural features listed.

8.2.1 Perform calibration and measurements

With the program Image-Pro Plus it is possible to use the function Count / Size (MEASURE → COUNT / SIZE) It is possible to count, measure and sort objects in a picture. For a better understanding and for later investigations, which are based on this work to be carried out faster, the procedure is to be illustrated by means of some screenshots from the program Image-Pro Plus. This also ensures that identical settings are used.

First of all a calibration has to be done (MEASURE → CALIBRATION → SPATIAL). The necessary values can be calculated from the originally scanned image. For the 800 × 800 pixel images, an area of 18.4 mm has been recorded, ie 18.4 μm are represented per pixel (see 46 ).

This calibration can be saved and opened and used as needed. To check the values, simply use the Measurements function (MEASURE → MEASUREMENTS) to drag a measurement line across the image width. The associated measured values are output in a table. These values can be used to check the calibration settings (see 47 ).

For each individual measured object about 50 different information can be output, including the location, the area and the scope. Before taking a measurement, various parameters must be set to obtain meaningful measurement results. In a first step, the image to be examined is opened and then the function Count / Size (MEASURE → COUNT / SIZE) is selected (see 48 ).

Select Measurements (MEASURE → COUNT / SIZE → MEASURE → SELECT MEASUREMENTS) to select the desired information to be included later in the output file (see 49 ). For each option a short explanation is shown with an example. For the sake of simplicity, everyone was chosen here to have the largest possible database, from which one can later create selective evaluations.

Once all desired information has been selected, the options for the measurement must be set. This is done both in the Count / Size dialog box and in the Options dialog box (MEASURE → COUNT / SIZE → OPTIONS). In the main dialog box, the options should be appropriate 48 be selected in the Count / Size Options dialog box accordingly 50 , Since exactly what has to be measured, which has also been evaluated visually, both the function Clean Borders must be disabled, as well as With Holes selected in the Outline Style menu.

After the presetting of all options can now be measured. Simply click on the button area Count in the main dialog box Count / Size. The captured objects are displayed accordingly in the picture (see 51 ).

The values for all objects can be exported to a file (MEASURE → COUNT / SIZE → FILE → DATA TO FILE), so that they can be further processed with other programs, eg Microsoft Excel (see 52 ).

In addition, the Count / Size function (MEASURE → COUNT / SIZE → IMAGE MAKE SORTED OBJECTS IMAGE) also offers the option of sorting all objects according to a selectable criterion (see 53 ).

This possibility is very useful for the study of structures, because it gives you the opportunity to see all structures sorted by their size (see 54 ). So one can isolate individual objects of similar size or compare them.

8.2.2 Possible measurable structural features

In the data obtained with Count / Size has become different aspects examined. These are briefly listed below and with relevant results also represented with a corresponding correlation to the structural rank.

8.2.2.1 Number of objects

The number of objects varies between 284 (paper 8) and 399 (paper 4). However, this number says nothing about the quality of the Wavings, because it does not matter how many objects on a Flä but how they are distributed there. This can also be clearly understood visually.

8.2.2.2 Total Area of Objects (Area polyon)

How out 55 It can be clearly seen that the total area of the object (the accumulated area of all objects) is not a measure with which the quality can be determined. Among the good performing papers are both papers with a small total object area as well as papers with a very high total object area. Among the poorly performing papers are only papers with a high total object area. For papers with a high total object area and a good rank, the high area ratio must be very evenly distributed over the entire image area. This can also be checked visually. A closer examination of this very important aspect is described in the section "Examinations of surface homogeneity" on page 55. Another possibility is to look at only the largest objects. If you look again at the structural patterns from this point of view, you can clearly see that the size of the objects has a dominant effect. This is also reflected in 56 again.

But if the biggest object some papers a similar one Magnitude has, must other factors play a role, which are the different ones Assessment effect. For example, e.g. visually detectable that the largest object just as annoying is perceived if it is not positioned under similar sized surrounding objects is. This topic is described in the section "Investigations on the homogeneity of the surface" on page 55 again treated in more detail.

8.2.2.3 Box size of the objects

• • Die XY-Ausdehnung im Koordinatensystem ist Grundlage des Boxmaßes
• • Die tatsächliche Länge und Breite werden miteinander multipliziert und man erhält eine Box, die das Objekt direkt umgibt

The extents of an object determine the box size surrounding it. It is possible with Image-Pro Plus to determine two different box dimensions:

• • The XY extent in the coordinate system is the basis of the box measure
• • The actual length and width are multiplied together and you get a box that surrounds the object directly

For this Investigation became the first option used. The difference to the pure object surfaces is, that the box u.U. more expressiveness for the visual effectiveness has. So a big, very jagged Object little area shares have, it would but visually perceived as dominant because it is about a huge Image portion extends.

As 57 shows, the box size has a similarly dominant influence on the rating as the area of an object.

8.2.2.4 Scope of objects

The size of an object also has a visual effect. A large, fissured object has a larger circumference than, for example, a circular object of the same area. The relationship between the largest scale and the structural rank is in 58 visible, noticeable.

8.2.2.5 Characteristics of the 5 and 10 largest objects

• • 5 größte Flächen: R² = 0,7367
• • 5 größte Boxen: R² = 0,8169
• • 5 größte Umfänge: R² = 0,6392

Since the measurement of an object does not say anything about other large objects, the 5 and 10 largest objects (by area) were used for further investigation and the respective feature values were summed up. This resulted in the following correlations for the 5 largest objects:

• • 5 largest areas: R ² = 0.7367
• • 5 largest boxes: R ² = 0.8169
• • 5 largest scopes: R ² = 0.6392

• • 10 größte Flächen: R² = 0,6954
• • 10 größte Boxen: R² = 0,6432
• • 10 größte Umfänge: R² = 0,5518

For the 10 largest objects, these correlations resulted:

• • 10 largest surfaces: R ² = 0.6954
• • 10 largest boxes: R ² = 0.6432
• • 10 largest scopes: R ² = 0.5518

8.2.2.6 Further Examinations: Scope per box / area Etc

Further investigations were carried out, but the above mentioned features show the highest correlations. With a summation of the largest object surfaces one can create a kind characteristic curve (see 59 ). This indicates how much proportion of the object total area has the k summed largest object surfaces. For example, it is visible in paper 1 that the 20 largest objects already account for half of the total object area. A rather smooth and gentle rise in the curve indicates a calmer surface.

8.2.2.7 Evaluation

Based The obtained different values for individual objects can be one very accurate statement about Sizes, number and shapes are made. However, this is not enough to reproduce the visual perception. They are unique Dominant factors determinable, but not alone the effect determine. It is possible, but to determine individual quality groups with these characteristics a fine differentiation is not possible in this way.

To have to other features of the structural patterns are measured. The characteristics of shape, size and size For example, do not make any statement about the distribution over the area and neighborhood relationships of structures. To make a statement about the Area homogeneity meet to be able to It takes a completely new approach become. This is described in the section "Examinations of the homogeneity of the surface" on page 55 described.

8.2.3 Creation of a structure report

Although it is not possible to make any statements about the distribution of the objects on the basis of the data obtained, a report for a measured print pattern can be output using an Excel mask. This report might include important things like sizes and number of objects. This gives you very detailed information about the structures. To do this, the files exported from Image-Pro Plus are simply imported into Excel. Due to the prefabricated mask, the data is automatically sorted and output. This can be done in a form like in 60 happen.

This Report would be a useful addition to the later Waving index. The report makes it possible, for example to see what a change e.g. the pressure parameter causes - this can e.g. resizing be the agglomerates. With such values, a paper can be quite be described exactly.

8.2.4 Studies on the homogeneity of the area

With the above features it is not possible to describe if a area is homogeneous. To do this new approaches chosen be, because the homogeneity is an important visual factor.

8.2.4.1 The scattering measure

To the structure pattern was subdivided into a matrix the size of 8x8 fields Has. Each field represents So an area of 1.84 × 1.84 mm. This size captured single agglomerates and is therefore not too fine or too coarse. In addition, can the eye in this area very well structures and contrast differences perceive.

There MS Excel just the way of 256 columns in the table layout offers, the images of 800 × 800 on 80 × 80 be scaled. This was initially a black and white bitmap and then scaled to appropriate size. A visual check showed that from corresponding distance both structural patterns are comparable are.

This 80x80 pixel image was read into Image-Pro Plus to get a table of pixel values (MEASURE → BITMAP ANALYSIS → FILE → DATA TO FILE). These data were loaded into a prepared mask in MS Excel. This mask calculates the amount of blackening for each field of the matrix. From all 64 fields, the mean value and the corresponding standard deviation are calculated and displayed (see 61 ).

The Standard deviation is a measure of the homogeneity of the area. she indicates how evenly the Agglomerations over the area distributed or the dispersion around the mean. Therefore this will be Measure called litter measure. The higher the scattering measure, the more restless or inhomogeneous is the measured area. This Measure must now be checked with the values of the structure ranking order.

For this purpose, the scattering measure for all papers and the correlation to the structural rank was calculated (see 62 ).

With a coefficient of determination of R ² = 0.8595, a quality group can already be determined very well. While there are variations in observers (represented by bubble size), these variations are only within the quality group. The quality groups are in 62 clearly. But even this measure alone is not enough to simulate the perception of structures.

By certain layers of the agglomerates can cause problems with this measure come. This is e.g. the case when a large agglomerate is the same Share about several fields of the matrix is distributed. Then it would be with this method only would be detected poorly but visually clearly visible. The same is true for punk clouds or accumulations of agglomerates. To capture these characteristics, must have chosen a second method who finds these priorities.

8.2.4.2 The centroid

Around the problem of point clouds (accretions of single objects) and Agglomerates that have several Fields of the matrix are distributed, to solve, a matrix has been developed in a defined size and in defined steps the structure pattern scans. That's it possible, to identify and evaluate the focal points of the structure image. Even if big Agglomerates under similar huge surrounding agglomerates an area distributed, this is not disturbing. This would have to be also reflect the emphasis.

developed were two different ways to realize such a sampling matrix. With the program Mathcad The company Mathsoft has developed a small program, this one Solve task with flexible parameters. Parallel to this was created a mask in MS Excel that performs a scan function on the 80 × 80 Pixel big Structural patterns performs.

Mathcad: The implementation of the sampling matrix in Mathcad has some key advantages. In Mathcad, there is no limit to data processing as in MS Excel. This allows you to use the 800 × 800 pixel structure patterns. These can also be evaluated directly by the program without the need for another bitmap analysis step. The step size and matrix size can be modified as desired. For this investigation, the window size was set to win = 100 (ie 1/8 of the image area) and the step size step = 20. Both values are figures in pixels. All specified values are gray values.

Excel: In Excel, the window size of the sample matrix with 30 pixels (ie 1/3 of the image area) and a step size set by 1 pixel. The values obtained are blackening parts. Also in Excel was the difference between the minimum and maximum Value used.

Of the value obtained represents the difference between the minimum and the maximum value. The higher this value, the more disturbing a point cloud or agglomerate is visually perceived.

As visible from the results of Mathcad is the maximum Value for all papers of a similar size. This suggests, that due to the chosen Window size fields The white space is taken contain. This can be bypassed by changing the window size become. Another possibility that would be 5 maximum values to output and to form an average. Likewise one can proceed with the minimum value.

Depending on the window size and increment, different values for the centroid are obtained. 64 shows these differences. In another investigation, it may be possible to determine an ideal window and step size.

That the centroid as a single value does not give an accurate statement about the quality of the structure can, shows the correlation of centroid and structural rank ( 65 ). The data from the Mathcad calculation was used.

8.3 Valuation-Relevant Structural Characteristics

Around To get a structure value, both the litter measure and the the centroid in the calculation be included. As well as larger structures, as in the section "Possible measurable structural features" on page 50 has been considered, this was considered in a calculation of a structural value. But because the sample matrix for the center of gravity value such structures also recorded and even considered point clouds in addition, is the size of objects already integrated in this value.

Since both an increasing scattering measure and an increasing centroid indicate disturbing visually perceptible structures, these two values for the structure value were multiplied together. This value was then set in relation to the determined structure lengths. Since the Mathcad centroid had the highest correlation, it was used. With the structural value calculated in this way, the visual evaluation can be simulated very clearly (see 66 ). With the exception of one paper, all values are close to one line. The coefficient of determination is R ² = 0.9154. This confirms that the measure of gravity can correct the weaknesses of the scattering measure. The correlation continues to increase (R ² = 0.9913) when paper 4 on the best rank is taken out of the correlation calculation.

8.4 Integration of the brightness

With the structure value, an evaluation form is now available that can measure structures according to their visual effectiveness. If one correlates this value with the rank PVi, one clearly sees that a further step is necessary to combine the structure value with a factor such that the desired waving index is formed (see 67 ).

If one tries to multiply the brightness as originally planned, however, there are no good correlations. For the Wi (here: structure value multiplied by Y / 100) the following correlation is obtained (see 68 ).

you Although recognizes clearly the respective groups, but in the result one receives none satisfactory correlation. Also other attempts the brightness to integrate over all papers bad correlations. To find the reason have to various aspects will be discussed again.

If you 67 Look closely again, the following occurs:

67 shows clearly two levels: On and above the trend line are all LWC papers, below the trend line all three SC papers are positioned. If one separates these two or three paper classes and only correlates the structure value with the rank PVi, then one obtains a correlation of R ² = 0.8975 (linear) and R ² = 0.9369 (potentially) for the LWC 50 papers. (please refer 69 ).

Correlation across the LWC 50 and LWC 70 papers gives R ² = 0.8943 (linear) and R ² = 0.9332 (potential) (see 70 ).

The three SC papers correlate with R ² = 0.9998 (linear) and R ² = 0.9986 (potentially) (see 71 ).

• • Genaue Arbeitsweise des Filters
• • Helligkeitsnormierung durch Schwellwerttrennung
• • Ideales Messfeld in verschiedenen Cups
• • Gewichtung der Helligkeit bei der Errechnung des Wi

These diagrams show that there are clear relationships between the respective paper class and the structure value. Although the rank PVi can be determined with the structure value and the paper class, this method is not a global solution. The following aspects are to be discussed in further investigations:

• • Accurate operation of the filter
• • Brightness normalization by threshold value separation
• • Ideal measuring field in different cups
• • Weighting of brightness in the calculation of Wi

Further Explanations These points are given in the section "Outlook".

In order to obtain a brightness value, one can use both the Y and L * values. There you are If the digitization of the measuring fields also receives a medium gray scale, the question arises as to whether this could also be used as a brightness value. Especially with regard to a one-shot technology, this would be an ideal method. 72 shows that these two values do not correlate ideally with each other. Possibly, an average of multiple measurements would give higher correlations.

Out time constraints had another investigation on the integration of brightness be broken off at this point, but also in the Section »Outlook« on page 65 important starting points called.

9 Presentation of the final procedure

In This section will briefly review all previous steps of the chosen Procedure listed schematically. For further investigations are Multiple measurements for the respective print patterns needed.

9.1 steps

To each step will be brief explanations given to the methods used.

9.1.1 Digitization of the measuring field

The Print samples are captured digitally with a scanner. Then done a trimming so that the area to be examined is saved as an image file can be. If necessary, the digitized measuring field is scaled.

9.1.2 Extract the structures

• 1. Filter Median
• 2. Schwellwerttrennung
• 3. Filter Close
• 4. Beschnitt

For extracting the structures, several filters are applied one after the other to the digitized measurement field:

• 1. Filter median
• 2. threshold separation
• 3. Filter Close
• 4. Trim

9.1.3 Assessment of structures

• 1. Streumaß: Das Streumaß ist ein Maß für die Homogenität der Fläche und zeigt an, wie gleichmäßig die Agglomerationen über die Fläche verteilt ist. Je höher das Streumaß, desto unruhiger bzw. inhomogener ist die gemessene Fläche.
• 2. Schwerpunktmaß: Das Schwerpunktmaß ermittelt Punktwolken (Anballungen von Einzelobjekten) und Agglomerate, die über das Messfeld verteilt liegen mit einer Matrix, die in einer definierten Größe und in definierten Schritten das Strukturmuster abtastet. Dadurch ist es möglich, die Schwerpunkte des Strukturbildes zu ermitteln und zu bewerten.
• 3. Strukturmerkmale: Unter Verwendung einer Excel-Maske und der ermittelten Objektdaten aus Image-Pro Plus wird ein Report für ein gemessenes Druckmuster ausgegeben. Dieser Report liefert Informationen über z.B. Größen und Anzahl der Objekte und erhält so sehr ausführliche Informationen über die Strukturen und einzelne Objekte.

For the evaluation of the structures, three pieces of information are used. The structure value is formed from the scattering measure and the centroid, as well as additional information on individual objects (structural features).

• 1. Scattering Measurements: The Scattering Measure is a measure of the homogeneity of the area and indicates how evenly the agglomerations are distributed over the area. The higher the scatter, the more restless or inhomogeneous the measured surface.
• 2. Center of gravity: The center of gravity measures point clouds (clustering of individual objects) and agglomerates that are distributed over the measuring field with a matrix that scans the structure pattern in a defined size and in defined steps. This makes it possible to identify and evaluate the focal points of the structure image.
• 3. Structural features: Using an Excel mask and the determined object data from Image-Pro Plus, a report for a measured print pattern is output. This report provides information on, for example, the size and number of objects and thus provides very detailed information about the structures and individual objects.

9.1.4 Integration of brightness or material properties

Around need to form the Wi to the already determined structure value brightness or material properties to get integrated. This may e.g. in the form of the Y value or by a value for a certain class of paper happened.

9.1.5 Output Wi / Report

The output of the Wi can be done both as a single value and in the form of a report. The Re port contains detailed information about Wi, Scattering, Structure value, Brightness, and Structural features. Thereby, a character of the paper can be determined under the waving aspect.

10 Conclusion

In This conclusion will be different problems and Expectations raised in dealing with the topic Waving have done.

The Finding a suitable rating method was very much overall Time intensive. A formation of rank orders with the pairwise Comparison and subsequent Evaluation was so complicated in 10 printing samples with several observers that for several Days had to be scheduled.

There several different approaches initially promising and seemed suitable, they were parallel z.T. pursued intensively. Only after some time were problems and restrictions clearly, so that a process emerged that very well seemed to work. The extraction of structures and their Review was ultimately a way in which much visually checked and matched was to make sure that mathematical calculations and visible structures could be summarized in numbers. This is Now the procedure is roughly described and needs further steps be optimized.

10.1 Correlation size

The are calculated correlation quantities at this stage of the investigation very uncertain, because the measurements z.T. only on one sample. Only by multiple measurements will show whether the proposed methodology is feasible as listed. In further measurements then the limits of the process demonstrated.

10.2 Weaknesses and problems

Some Steps of the investigation have methodical weaknesses. The formation of rank orders was done with few observers carried out, which could be compensated by a high, professional competence. Nevertheless, it would be it makes sense, such rank order still eimal with at least 10 To perform observers.

Also the limits of visual appraisal by humans were added this investigation. Especially when it comes to complex Observers often judge ratings (such as area homogeneity) differently. Therefore, it is therefore acceptable that the mathematical Capturing can be more accurate than an appraisal of people. These Mathematics must then be adjusted according to the visual perception but probably can not accurately simulate human vision, because the processes taking place in the brain are very complex. Also the restrictions by the physical nature of the human optical system have to considered because it can be detected more metrologically than visually can be perceived.

10.3 Expectations for multiple measurements

Of the Procedure must be checked in a further step with multiple measurements. This will be for each paper sample measured and evaluated at least 6 samples. Furthermore is under consideration pull, if also further Tonwertfelder should be measured.

To These measurements and their evaluation are expected to have a Scattering within the paper patterns, but within certain quality Circle lie. When measuring other tone value fields becomes visible where waving is in a very critical area. This is probably different for all paper types Job.

To this multiple measurement can then process steps can be improved or changed to make it optimal To agree on the results.

11 Outlook

This survey on the Waving Index provides a foundation on which to build further research. Unfortunately, the timeframe did not allow all procedures and methods more intense and so the decision had to be made for the most promising method. The selected method (extraction of the structures and their evaluation) must now be tested and improved in further measurement methods. Only then will it become clear to what extent the chosen path was right and achieves useful measurement results. In the outlook some links to different aspects of the process are mentioned, which require further discussion and which need to be taken up in further investigations.

11.1 The visual assessment methods

The different rank orders were created with few observers. Although these observers were highly skilled in technical matters, but it would be beneficial to have these orders evaluated by other observers. This would give you an objectified rank order. Interesting would be at then consider whether the uncertainty in some papers continues. Depending on the samples to be assessed then possibly more Statements about visually dominant factors are made.

11.2 The procedure

For the procedure in total, there are several points of contact. In the first place would have a corresponding and possible full Automation will be elaborated if all process steps finalized after a detailed review were. This automation can be used as a program, plug-in, macro or similar be implemented.

In another step would be to think, How to reach suitable measuring fields during printing events. To could for example, in areas that are gone after further processing be cut, an area with different Tonwertfeldern be printed. Out of it could then samples are taken and measured, so for such a Investigation no trial pressure must be carried out.

There for this Examination only monochrome prints with black ink used would have to also investigations on other printing inks take place. Because the ink Black the highest contrast to a bright substrate is in this case the agglomeration visually most critical. It is to be expected that for brighter Printing inks agglomeration visually less critical. Also with regard to a compression of the different printing inks this aspect is interesting.

The Process has been used in gravure printing. It is also possible an extension to other printing techniques, e.g. the flexo print, which also works with low-viscosity colors. Even there it might be possible to change the procedure so that structures are extracted from the printed image and evaluated. Especially strongly absorbent substrates could be similar in flexographic printing Have problems like waving.

Also There is a need for investigation and improvement regarding the individual steps.

11.2.1 Selection and digitization of the measuring field

Microscope images show that the LWC and CS papers flow significantly differently (see 73 ). The SC papers flooded much more strongly in the 55% field than the LWC papers, so that more area is covered by ink and possibly the agglomeration behavior is different. This is to be found out.

In This context must also be examined, depending on the paper type Cup also gives an optimal measuring field. Presumably this would be for SC papers in a lower tone value field than for LWC papers.

For digitization can be an optimal scan resolution be found out. This has two advantages for the process: the scanning process shortened if not with the highest possible resolution is scanned and at the same time the amount of data shrinks, which must be processed. Complex calculations would thereby faster accomplished.

11.2.2 Extract the structures

• 1. Filter Median: Die genaue Wirkung des Filters muss überprüft werden. Dazu zählt sowohl der mathematische Hintergrund (der in Image-Pro Plus einsehbar ist) als auch die Möglichkeiten mit verschiedenen Parametern zu arbeiten. Eine kontrollierte Überprüfung kann man mit künstlich erzeugten Mustern durchführen (siehe Abschnitt »Verfahrenstest mit künstlich erzeugten Mustern«).
• 2. Schwellwerttrennung: Die genaue Wirkung der Schwellwerttrennung muss überprüft werden. Möglicherweise ist eine Normierung auf eine Helligkeit dadurch nicht vollständig möglich. In diesem Zusammenhang müsste die Schwelle für die Trennung diskutiert werden, die hier für alle Papiermuster identisch berechnet wurde. U.U. muss diese Trennung helligkeits- oder papierbedingt flexibel erfolgen.
• 3. Filter Close: Der Filter Close entfernt kleinste Strukturen, die nach der Schwellwerttrennung erhalten blieben sind. Für die Berechnung des Streu- und Schwerpunktmaßes sind diese weniger relevant, aber sie könnten für die Strukturmerkmale der einzelnen Objekte eine Rolle spielen.
• 4. Beschnitt: Um einen idealen Beschnitt durchführen zu können, müsste herausgefunden werden, wie weit sich der Randeffekt durch den Median-Filter auf das digitalisierte Messfeld auswirkt. So kann eine optimaler Beschnitt gewährleistet werden.

The extraction of the disturbing structures and thus the elimination of the grid characterizes the this procedure. There are also further considerations in the various steps that are necessary for this.

• 1. Filter Median: The exact effect of the filter must be checked. This includes the mathematical background (which is visible in Image-Pro Plus) as well as the possibilities to work with different parameters. A controlled check can be done with artificially created patterns (see section "Artificially generated patterns test").
• 2. Threshold Separation: The exact effect of thresholding must be checked. It may not be possible to normalize to a brightness. In this context, the threshold for the separation would have to be discussed, which was calculated here identically for all paper samples. This separation may have to be flexible due to brightness or paper.
• 3. Filter Close: The Filter Close removes the smallest structures that were left after the threshold separation. For the calculation of the scatter and centroid they are less relevant, but they could play a role in the structural features of the individual objects.
• 4. Trimming: In order to perform an ideal trimming, it would be necessary to find out how much the edge effect of the median filter has on the digitized measuring field. So an optimal trimming can be guaranteed.

11.2.3 Evaluation of structures

Out the order of rank showed which areas visually evenly or less act evenly. Nevertheless, not all connections could be clarified yet. For example, e.g. Not clearly which factors are visually dominant (see section »Visual Effective dominance ").

calculations to neighborhood relationships or local contexts not done or only secondary integrated in the calculation of the structure value.

• 1. Streumaß: Das Streumaß wurde bisher aufgrund der Einschränkungen in Excel anhand skalierter Strukturmuster durchgeführt. Eine geschicktere Lösung wäre eine Matrixfunktion, die, wie auch das Schwerpunktmaß, direkt die Werte aus Tif-Bildern generieren kann. Dieses wäre z.B. auch in Mathcad umsetzbar. Außerdem muss möglicherweise eine optimale Größe der Matrix gefunden werden.
• 2. Schwerpunktmaß: Das Schwerpunktmaß kann sehr einfach in Mathcad anhand vorliegender Tif-Bilder errechnet werden. Auch hier muss eine optimale Größe des Abtastfensters und der Schrittweite gefunden werden. Zu überprüfen ist, ob diese an die Papierhelligkeit oder andere Aspekte flexibel angepasst werden muss. Außerdem muss überprüft werden, ob eine Differenz aus den drei größten und kleinsten Werten ein besseres Ergebnis liefert, denn mit der bisherigen Methode ist es möglich, dass das Abtastfenster auf extreme Bildstellen trifft. Da momentan die Werte als Grauwert ausgegeben werden, könnte hier noch eine Umrechnung in Schwärzungsanteile bzw. Prozent erfolgen.
• 3. Strukturmerkmale: Durch eine Excel-Maske können beliebige Statistiken aus den Objektdaten der Strukturen gewonnen werden. Hier müssen in einem weiteren Schritt informative und relevante Daten herausgesucht werden, die viel Aussagekraft über Papier und Agglomeration bieten. Dadurch werden im ausgegebenen Report keine überflüssigen Informationen ausgegeben. Die Daten könnten visuell wirksam auch in Diagrammform dargestellt werden, um so eine schnelle Erfassung der Qualität zu ermöglichen.

Such relationships also include extreme cases that may indicate the limits of the algorithms (see section "Artificially generated pattern test").

• 1. Scattering Measurements: The Scattering Measurements have been performed based on scaled structure patterns due to limitations in Excel. A more clever solution would be a matrix function that, like the centroid, can directly generate the values from TIF images. This could also be implemented in Mathcad, for example. In addition, an optimal size of the matrix may need to be found.
• 2. Focus: The centroid can be easily calculated in Mathcad using existing Tif images. Again, an optimal size of the sample window and the step size must be found. It has to be checked if it has to be adapted flexibly to the paper brightness or other aspects. In addition, it must be checked whether a difference of the three largest and smallest values provides a better result, because with the previous method, it is possible that the sampling window encounters extreme image locations. Since the values are currently output as a gray value, a conversion into blackening percentages or percent could take place here.
• 3. Structural features: Through an Excel mask any statistics from the object data of the structures can be obtained. Here, in a further step, informative and relevant data have to be found out, which offer much information about paper and agglomeration. As a result, no unnecessary information is output in the output report. The data could also be displayed graphically in a visually effective manner to enable a quick recording of the quality.

11.2.4 Integration of brightness or material properties

The Integration of brightness or material properties must be more important connecting factor be further investigated. It is u.U. first an optimal measuring field for the to determine each papers and cups, then again try to integrate the brightness. Because LWC and SC papers have a different flooding behavior, this appears Approach sensible.

Also the weighting of the brightness in the calculation of the Wi must be in reviewed and examined this relationship again become. One guess is that this one does not matter so much must play, as initially assumed. As dominant she could to be synonymous only in conjunction with a strong Waving.

11.2.5 Output Wi / Report

When the number ranges in which the scattering measure and the centroid move are final set a scale for the structure value and the Wi. Based on a location on this scale, an observer must be able to assign a quality. This scale should also be visually verifiable.

As in the section above »Rating of the structures «already mentioned, it must be determined what kind of data in which form are displayed in the output report. If Multiple measurements, even over several Tonwertfeldern, made may even have a waving character of each Paper (see section »Characteristics of papers«).

11.3 Procedure test with artificial generated patterns

The Procedure should be artificial generated patterns are tested. These should be extreme cases as well also ideal cases map so that the algorithm and its individual steps are checked and the Working method can be understood. This may possibly also the limits of the algorithm are shown.

Around also individual steps of the procedure to their specific function to be able to test need for this Explicit patterns are generated. Thereby it is e.g. possible, the to document the exact mode of operation of the median filter. Also for the structural measurements would special patterns are developed.

11.4 Visual Impact Dominance

• • Welche Merkmale sind visuell dominat: Form, Lage, Größe, Umfang?
• • Was passiert, wenn große Weißflächen zwischen einzelnen Strukturen liegen?
• • Wie dominant ist die größte Struktur?
• • Welche Rolle spielt Symmetrie?
• • Gibt es eine Vorzugsrichtung des Sehens, z.B. eine höhere Empfindlichkeit für vertikale Strukturen?

Although it emerges from this study that different features of structures have a major impact on visual assessment, the exact interplay of several such factors has not been fully clarified. This could also be clarified by artificially generated structural patterns, so that then visual and statistical assessments could be compared and possibly improved. Particularly important would be the following examinations:

• • Which features are visually dominant: shape, position, size, size?
• • What happens when there are large white areas between individual structures?
• • How dominant is the largest structure?
• • What role does symmetry play?
• • Is there a preferred direction of seeing, eg a higher sensitivity for vertical structures?

There But never all observers will be judging a sample equal there is always a dispersion in certain areas.

There it was found that with bright papers bad printing is emphasized, there is a need for investigation for the addition of optical Brightening up in papermaking. Maybe it's for one Paper manufacturers under certain circumstances do not make sense to the paper Brightener mix, because then restless structures visually effective become.

11.5 Characteristics of papers

With multiple measurements over several cups and tonal fields a waving characteristic for a paper could be represented (see 74 ). Such a profile would be interesting for both paper manufacturers and the printing industry.

Perhaps let yourself from such a profile also read the optimum for a measuring field, that for one latter to serve standardized waving measurement.

Perhaps In this context, it is also possible to determine preferred directions. This would have to be metrologically be detected by a narrow in the printing direction scanning matrix on the Picture is running.

11.6 Causes of the Wavings

Also the causes of the Wavings need be examined more closely. Because a lot of parameters influence it can have, such an investigation will be very extensive.

The various possible parameters are described in the section "Causes for the occurrence of Waving". described. In particular, the interaction between capillary action and viscosity of the paint would be interesting to study.

List of Figures

0 : Source image and resulting SNW bitmap image

1 : Pressure tests with waving phenomena in the midrange

2 : Enlarged view of paper agglomeration 6

3 : Different shapes of large agglomerates (after processing and sorting taken from paper 8)

4 : Partial agglomeration along paper fibers on a SC paper

5 : The unprinted paper with clearly visible fibers (SC paper)

6 : Reduced sections of the printed image

7 : A printed pattern on a test strip

8th : The evaluation cabin

9 : A dominance matrix D [gij] of an observer

10 : A dominance matrix R [rij] of an observer

11 : The intransitive judgments

12 : Rank after PVi

13 : Rank to DV

14 : Correlation of PVi and DV

15 : Correlation of PVi and DV

16 : Correlation of the rank PVi with the brightness L * (all papers)

17 : Correlation of rank PVi with brightness L * (only LWC 50 papers)

18 : Correlation of rank PVi with medium gray value (LWC 50 papers only)

19 : Resolving power of the human eye

20 : Modulation transfer function for color and grayscale viewing

21 : Contrast sensitivity of the eye for spatial grids of different spatial frequencies at different illuminances on the retina

22 : Paper 7 (LWC) and Paper 10 (SC) with pressure points to be measured

23 : Choices when importing images

24 : Fluctuations in camera recording

25 Image taken with microscope and camera (reduced size): paper 7 (left) and paper 8 (right)

26 : The collected samples (paper 10-1, from left to right)

27 : An excerpt from the scanned area of paper 7

28 : Expected ideal distribution in a real grid

29 : Histograms of papers 1-10

30 : Histograms of papers 1-10 after 8⎕8 blur

31 : Correlation of the maxima with the rank PVi

32 : Synthesis of a non-sinusoidal signal by summing up three sinusoidal signals

33 : Amplitude representation after FFT for paper 6 (from Image-Pro Plus).

34 : Very good and very poor workable recording

35 : Isolated halftone dots with agglomerates

36 : The image after filtering out the regular patterns

37 : After a threshold separation

38 : The digitized section of paper 6

39 : After the application of the filter median

40 : Before (left) and after the application of the filter median (right)

41 : After performing the threshold separation

42 : Different processing levels from the input image (left) to the finished structure image (right) in chronological order

43 from left to right: paper 1-5, bottom row from left to right: paper 6-10

44 : Ranking order of the structure patterns

45 : Ranking order of the structure patterns

46 : Input mask for the calibration

47 : Width of the image in the previously defined unit (⎕m)

48 : The Count / Size Dialog Box

49 : Selectable information for Count / Size

50 : Options for Count / Size

51 : The image representation after using Count / Size

52 : Export options for the measured values from Count / Size

53 : Options for Object Sort

54 : Objects made of paper 6 sorted by size

55 : Correlation of Total Object Area vs. Structure Rank

56 : Correlation largest object against structure rank

57 : Correlation largest box against structural rank

58 : Correlation of the greatest extent against structural rank

59 : Buzzer of the k largest object areas in relation to the total area

60 : Automatic report output to object data in Excel

61 : Calculation of the mean value and standard deviation from 8⎕8 fields (dark background)

62 : Correlation of litter measure against structural rank

63 : Results edition in Mathcad

64 : Results for the centroid in Mathcad and Excel

65 : Correlation of centroid and structural rank

66 : Correlation Structural Value vs. Structural Ranking

67 : Correlation structure value versus rank PVi

68 : Correlation Wi (from structure value * Y / 100) versus rank PVi

69 : Correlation Structural Value vs. Rank PVi: LWC 50

70 : Correlation Structural Value vs. Rank PVi: LWC 50 and LWC 70

71 : Correlation Structural Value vs. Rank PVi: SC 50

72 : Correlation Y vs. Mean Gray Value (Digitized Measurement Field)

73 Image taken with microscope and camera (reduced): LWC paper 7 (left) and SC paper 8 (right)

74 : Imaginary Waving Characteristic of an imaginary paper

List of Tables

• Table 1: Gravitational phenomena associated with Printing ink and substrate
• Table 2: The paper patterns
• Table 3: The points system for the pairwise comparison
• Table 4: Quality Groups after PVi and DV
• Table 5: Description of the structure patterns

• a. LWC = leight-weight coated
• b. SC = supercalandered

Qualitätsaussage Punkte Element gij Probe j gleicht Probe i 1 Probe j ist besser als i 2 Probe j ist schlechter als i 0 Tabelle 3 Gruppe PVi DV 1 (gut) 3, 4, 8 3, 8, 4 2 (mittel) 2, 5, 10 5, 2, 10 3 (schlecht) 6, 1, 9, 7 6, 9, 7, 1 Tabelle 4 Strukturmuster Visuelle Beschreibung Papier 1, LWC 50 3 Bereiche mit größeren Punktwolken: links, mittig und am rechten unte-ren Rand. Viel Weißraum, viele größere aber auch viele kleine Agglome-rate. Wenige mittelgroße Agglomerate. Wirkt ungleichmäßig. Papier 2, LWC 50 Viele Strukturen, überwiegend kleine und mittlere Größe. Eine Punktwolke im linken unteren und rechten oberen Bereich. Dort sind auch die größeren Agglomerate enthalten. Wirkt etwas ruhiger als Papier 1, aber auch ungleichmäßig. Papier 3, LWC 50 Wenig Strukturen. Kleine bis mittlere Größe, keine sehr großen Agglome-rate und keine Punktwolken. Viel Freiraum im mittleren Bereich, ein Verteilungsschwerpunkt zum Rand hin. Wirkt sehr ruhig. Papier 4, LWC 50 Sehr viele ähnlich große, mittelgroße Strukturen, wenig kleine und keine sehr großen. Außerdem keine Punktwolken. Viel Dunkelfläche. Gleichmäßige Verteilung über die gesamte Fläche. Wirkt sehr ruhig und gleichmä-ßig. Papier 5, LWC 50 Wenig kleine, mehrere mittlere und große Strukturen mit viel Weißraum. Punktwolken links und mittig oben, rechts unten. Wirkt z.T. unruhig. Papier 6, LWC 70 Mehrere Bereiche mit deutlich großen Agglomeraten und mehrere ebenso große Weißflächen. Wirkt sehr unruhig und ungleichmäßig. Papier 7, LWC 70 Mehrere deutlich große Agglomerate und Punktwolken im rechten Rand-bereich sowie in der Bildmitte bis in die linke untere Ecke. Wirkt sehr unru-hig und ungleichmäßig. Papier 8, SC 50 Sehr wenig Strukturen, einige mitelgroße. Wirken regelmäßigverteilt. Große Weißräume. Wirkt insgesamt ruhig. Papier 9, SC 50 Viele größere Strukturen, die sich in Punktwolken auch über breite Bildbe-reiche erstrecken, wirkt sehr unruhig und ungleichmäßig. Papier 10, SC 50 Viele kleine bis mittelgroße Strukturen regelmäßig verteilt. Einige große Weißräume, am rechten Rand auch größere Strukturen. Tabelle 5 designation description Plucking, taking off Partial tearing of an already printed gravure ink Pulling off, lying down Overriding of color components on the opposite side of the substrate in the stack or in the roll Demolish, cancel Uneven, sometimes erratic transition from individual halftone dots to flooding printing Pulling, undressing, squeezing The ink prints irregularly beyond the outline dry out The printing ink loses binding with long runtime and low ink consumption, the print image loses color strength Repel Non-acceptance of the ink by the substrate or pre-printed gravure ink To shine through Visibility of the printed image on the back through the substrate through due to low opacity of the substrate Breakthrough, Print Through On the back of the substrate visible effect, which is based on the penetration of the gravure ink or its constituents by the substrate bleed out Coloring agents of a print are dissolved by solvents or other substances and color the environment or in contact substances Migrate (hiking) Migration tendency of constituents of the printing ink or of the printing material, adhesive or other materials into a different material which is in close contact with it for a long time crowsfeet Bird foot-like extraction of gravure ink from the printing plate due to electrostatic influence Doctor blade, smudge streaks With the squeegee moving strokes or strips in the printing direction, caused by damaged doctor blades or sticking of solids under the doctor blade cylinder strips Strip-shaped error in the pressure, caused by damage of the cylinder surface in the circumferential direction comets Comet tail-like error in pressure, which occurs temporarily, caused by short-term adhesion of solids on the surface of the plate cylinder pinholes Needle-size imperfections in the print caused by inadequate wetting Missing Dots (snowflake effect) Point-shaped imperfections in the printing, caused by unevenness of the substrate ghost Just visible image of the print without color transition on the non-printed back of the substrate after rewinding Restless lying Is caused by uneven bleeding of the ink in interaction with substrate and printing conditions. Different forms of appearance, also known as flow marks: Benard's cells, pearls, streaks, tree bark structure, frog eyes Uneveness Denotes an unevenness in the solid surface (see restless lying) donuts Donuts are a squeegee effect. Due to the ink layer in the cup, the edges print first, so that there is a large part of the color, while in the middle of the substrate remains unprinted by the lack of color. This creates a shape that resembles a donut. waving Mid-tone phenomenon. Irregular, partial flooding creates disturbing, visible agglomerates that are often perceived as cloud or waveform Table 1 paper LWC / SC grammage 1 LWCA 50 gsm 2 LWC 50 gsm 3 LWC 50 gsm 4 LWC 50 gsm 5 LWC 50 gsm 6 LWC 70 gsm 7 LWC 70 gsm 8th SCb 50 gsm 9 SC 50 gsm 10 SC 50 gsm Table 2

• a. LWC = light-weight coated
• b. SC = supercalendered

quality statement Points element gij Sample j is equal to sample i 1 Sample j is better than i 2 Sample j is worse than i 0 Table 3 group PVi DV 1 (good) 3, 4, 8 3, 8, 4 2 (medium) 2, 5, 10 5, 2, 10 3 (bad) 6, 1, 9, 7 6, 9, 7, 1 Table 4 structural patterns Visual description Paper 1, LWC 50 3 areas with larger point clouds: left, center and right lower edge. Lots of white space, many larger but also many small agglomerates rate. Few medium sized agglomerates. Works unevenly. Paper 2, LWC 50 Many structures, mostly small and medium size. A point cloud in the lower left and upper right area. There are also the larger agglomerates included. Works a bit quieter than paper 1, but also uneven. Paper 3, LWC 50 Little structures. Small to medium size, no very large agglomerate rate and no point clouds. Much free space in the middle area, a distribution center towards the edge. Works very calmly. Paper 4, LWC 50 A lot of similar sized, medium sized structures, little ones and not very big. In addition, no point clouds. A lot of dark space. Uniform distribution over the entire surface. Works very calmly and evenly. Paper 5, LWC 50 Little small, several medium and large structures with a lot of white space. Point clouds left and center top, bottom right. Partly restless. Paper 6, LWC 70 Several areas with large agglomerates and several equally large white areas. Acts very restless and uneven. Paper 7, LWC 70 Several distinctly large agglomerates and point clouds in the right border area as well as in the middle of the picture down to the lower left corner. Looks very unruly and uneven. Paper 8, SC 50 Very few structures, some medium sized. Act regularly distributed. Great white spaces. Acts altogether calm. Paper 9, SC 50 Many larger structures, which also extend over broad image areas in point clouds, appear very restless and uneven. Paper 10, SC 50 Many small to medium structures regularly distributed. Some large white spaces, on the right edge also larger structures. Table 5

Claims (6)

A method for the objective determination of the print quality of flat print images, characterized in that a printed image is digitized at least partially by means of an image acquisition system and the resulting digitized image is a bitmap image is formed and for this a scattering measure is determined from the standard deviation of a parameter, in particular percentage area coverage, within the grid window of a given grid. Method according to claim 1, characterized in that that on the image obtained after digitization before detection of the litter measure at least one digital filter is applied, in particular for conversion into a black and white bitmap image, which only the disturbing ones Structures shows. Method according to claim 2, characterized in that that the digital filter is a blur and / or a threshold separation and / or a close operation. Method according to one of the preceding claims, characterized characterized in that a centroid of the difference between two Image sections with the highest and lowest area coverage is formed. Method according to one of the preceding claims, characterized characterized in that the determination of the percentage area coverage a given image section as matrix / raster overlapping in predetermined steps, over the bitmap image is moved. Method according to one of the preceding claims, characterized characterized in that the product of litter mass and centroid, in particular divided by 1000, is formed to a dimensionless structure value to build.