Printing Quality Control--Subjective Evaluation (II)

To evaluate the print quality of different papers, samples should be printed on different papers with a constant amount of ink. If it is printed on a color press, the image should be cut off from the sample and affixed to a neutral gray card with a classification number. The assessors observed these samples against a neutral gray background under a diffuse fluorescent light source. The evaluation results were recorded in a book and asked the assessors to:

1 Use a symbol to mark the degree of discrepancy observed for each pair of prints (Figure 6-4).

2 Use a symbol on the other ruler to indicate which of each pair is better and better.

It should be pointed out that when making discriminatory and subjective psychological weighted judgments, do not indicate to the evaluator which parameters should be used for evaluation. Multidimensional scaling techniques can extract statistically significant parameters from the evaluation data, calibrate these parameters, and determine the relative importance of these parameters in the overall print quality assessment.

To avoid the fatigue of the evaluator, the number of samples to be evaluated should be as small as possible, because the more samples are used, the greater the degree of freedom in the analysis. It has been found that 8 samples are a good compromise. Furthermore, when the number of samples is small, it is convenient for the evaluator to make a few exercises in advance, practice them to familiarize them with these samples, and stabilize their calibrated values.

The 8 samples consisted of 28 pairs of comparisons and were divided into 25 groups according to differences and psychological weightings. Therefore, from the evaluation of each evaluator, a large amount of detailed information about the samples was obtained.

Figure 6-5 is an example of fitting a multidimensional scale to all differences data obtained by a group of evaluators. If the best fit is three-dimensional or four-dimensional, then a series of two-dimensional coordinates can be used to represent other dimensions.

Other information obtained from multidimensional scaling analysis includes the calculation of standard deviations and deviation maps. With these means, illogical evaluators can be excluded from the information group. Each of these evaluators should be distinguished from each person who made a consistent evaluation, and should be distinguished from everyone who does not agree with the team's fitted model. If the assessor's sampling test is to reflect the number of professional teams involved, then the latter must be included.

The final step in the multi-dimensional scale is to explain the arrangement model of the sample prints. The direction chosen for the main coordinate (first dimension) is the direction that has the largest variance in the sample distance, the second or other coordinate is perpendicular to the main coordinate, and this direction occupies the largest residual.

A simple, approximate explanation of the alignment model is to flag these coordinates based on the subjective evaluation parameters used. The specific approach is to arrange the prints on a table according to the model and ask the evaluator to describe the situation of each coordinate change of the print. If the parameters are of obvious statistical nature, it is usually easy to mark these parameters because only the dimension problem is selected.

The analysis of eight prints either produces a two-dimensional coordinate corresponding to the field or contrast (see Figure 6-5), or results in a three-dimensional coordinate. High-contrast, good-quality prints can be placed on the upper left of Figure 6-6 (position A in Figure 6-6), while prints with poor field and contrast may be placed on the upper right of Figure 6-6 (eg, position E) Prints with high contrast and different contrasts will be placed in the upper right and lower right.

After the markers of the coordinates have been determined, subjective parameters can be obtained. For example, field, contrast, and print-through values, and make them related to the corresponding physical properties. Relevance is a common method for selecting physical metrics. It is in good agreement with subjective perception. Analyzing the correlation between subjective print quality parameters and paper properties is a method of diagnosing the root cause of print quality problems.

A single subjective parameter does not necessarily have to do with a single physical property. For example, the complex nature of spot specks and ink-moisture may be caused by many potential physical properties.

Studies have pointed out that print quality parameters are rarely independent of each other. A first approximation of the speckles and contrasts can be depicted in Cartesian coordinates as shown in Figures 6-6, but a more precise process requires that the orientation of these contrast coordinates reflect the independence of these two parameters.

Third, psychological weighted vector analysis

From a production point of view, it is not enough to know what quality parameters are used to evaluate differences in print quality, but also indicators that relate each quality parameter to the user's psychological factors. According to the mental information matrix obtained in Figure 6-4, Used to provide such information.

As in the case of the difference, a model is created in which each evaluator's psychological weighted benefit index dominates, and then based on the bias of the predicted psychological weighted model, the standard deviation of each evaluator can be calculated. It is also necessary to convert the mentally weighted model value into a comprehensive rms value for each print:

In the formula:
Ui—the effective value of an i th print;
P*ij - a model for predicting the distance between the i-th and j-th prints (only the lower triangular matrix);
n - number of prints.
Then return the effective values ​​of 8 prints to the difference coordinates.

In the formula: B1, B2, B3 are the characteristics of each deviation coordinate.
A mentally weighted vector diagram is made with features B1, B2, B3, etc., which can be plotted on the psychological weighted difference model, and FIG. 6-6 represents the psychologically weighted vector of a group of assessors drawn on the corresponding two-dimensional difference model. The length of the vector is a measure of the weighted intensity. The short vector represents a relatively unstable evaluator. The relative importance of each print quality parameter in the psychological weighting of each evaluator is expressed in terms of sagittality. Parallel to a certain coordinate axis, it indicates that the corresponding parameter is the only parameter that affects the evaluator's weighting of the printed product.

Therefore, the weighted vector distributed in Figures 6-6 is a simple expression method for a group of evaluators' psychological weights.

As far as the weight of each print quality parameter is concerned, various types of evaluators, such as paper industry experts, printing workers, advertising staff, and readers, may be quite different. The multidimensional scale shown in Figure 6-7 provides a quantitative measure. The method of these differences. The figure shows the weight characteristics of the 4 parameters that have an impact on the quality of a set of printed products. Printers tend to put a greater weight on the quality parameters and contrasts that he controls. Papermakers and readers are more sensitive to the color of the paper. Stripes and prints are common concerns for the four groups. This analysis method can be used to match the nature of the paper based on the customer's psychological tendency.

To date, multidimensional scaling is a powerful tool for quantifying the subjective print quality with a blurring characteristic. Multidimensional scaling can identify print quality parameters that have a significant impact on the evaluation of print differences, assign numerical values ​​for each print quality parameter to the print, and use these numbers to relate the subjective print quality parameters to physical properties or paper properties of the print. .

The importance of each print quality parameter for each evaluator's psychological weight can be clearly expressed by a weighting vector.

The stability of each evaluator and his consistency with other members of the evaluation team can be tested by estimating the standard deviation and plotting the deviation.
(to be continued)