home

Plating the Ten Lepta Large Hermes Head Stamps of Greece

By Louis Basel

(This article was originally published in the Lighthouse of the Philatelist (Journal of the Philatelic Society of Patras), Vol. 1, No. 1, October-December 2007, p. 11, Greek/English. Plating of the large Hermes head stamps is useful in the classification of these issues because the grouping of several stamps from the same positions allows the comparison of their control numbers, ink spots and other flaws and provides valuable evidence for their arrangement in chronological order. An examination of the stamps' impressions, colors, papers, etc. then leads to their classification in the appropriate issues.

1. Introduction

I developed my first plating program for the 20 lepta large Hermes head stamps in 1984 and published an article describing it in The American Philatelist in 1985. This program used the very first IBM personal computer, the IBM-1 without a graphics monitor. All of the image comparisons were done in computer memory without any visual image display on the computer monitor. The computer program was written in the IBM Basic language and was used successfully to plate thousands of 20 lepta stamps. This is described in the article available on this web site: Computerized Plating of the 20 Lepta (1985)

A few years later, with the advent of graphic monitors and graphic display boards, a new program was developed using the Turbo Pascal programming language, a Tecmar monitor, a Chorus Data Systems digitizer board, the Halo graphics subroutines of Media Cybernetics and an Intel 80486 based computer. It was a DOS program and not adapted to the newer, at that time, Windows operating systems. This program was much faster than the previous one and was used to plate thousands of 10, 20 and 40 lepta stamps. It is still being used infrequently today, primarily to plate stamps of friends who send me image scans of their stamps. For many years, I feared that my old computer or one of the other auxiliary hardware units would fail. Since some of the manufacturers of these parts are no longer in business, it would be impossible to replace them and I would have to abandon my computer plating program.

Recently, Kosta Hatzis of Patras Greece started an e-mail correspondence with me and I helped him plate a few stamps. He asked how he could plate the early large Hermes head stamps which had no ink spots, characteristic control numbers or other identifiable varieties. I told him about my computer program and that the hardware it used was no longer available. I said that I had thought about developing a new program which would use modern windows based computers and software programs commonly available. With Kosta’s encouragement, I finally started work on this new program.

2. The Plating Method

It is well known that the numerals in the lower inscription block of these stamps were manually punched into the individual clichés which made up the printing plates. Because of this manual operation, the numerals are not exactly centered between the two white dots on either side. Both of my previous programs were based on the horizontal and vertical deviations of these numerals. To use the same technique, it was first necessary to find a method of measuring these deviations. I had been familiar with the CorelDraw software program which provides a means of displaying and manipulating images on a computer monitor. Most importantly, it provides a measurement system which is accurate to thousandths of a millimeter. Thus, it was possible to use this CorelDraw program to display an image of the stamp and to measure accurately the horizontal and vertical coordinates of the centers of the numerals and the white dots.

There is one additional piece of information that was useful in plating the large Hermes head stamps. That is the frame line spacing that was very different for many of the plate positions. This was measured only in one location, immediately below the left dot of the left numerals.

It was then necessary to calculate the numeral deviations and the frame line spacing for all 150 positions of the sheet of ten lepta stamps. This was done using the Microsoft Excel program where the necessary formulas were entered. Then the CorelDraw measurements were entered and the deviations were calculated automatically.

Once this table of 150 standard deviations was established, it was possible to enter the CorelDraw measurements for an unknown stamp and compare them with the 150 standard deviations to find the closest match. A mathematical technique called the “least squares method” was used which provided a comparison number indicating how close the deviations of the unknown stamp were to each of the 150 positions. The lower this number, the closer the unknown is to that position. By sorting these positions and their comparison numbers with the smallest numbers at the top, one could see the closest positions to that of the unknown stamp.

3. Measuring the Coordinates

The calculations used to determine the deviations of the numerals are performed using simple arithmetic functions. However, to explain them we need to use some sketches of the numerals and white dots in the lower inscription block of the stamp. We first create a mask for the numeral “10” by using the CorelDraw tools and adjusting the numeral to fit the numerals of the stamp. An enlarged image is shown below:

Figure 1. Mask of numeral “10”.

A rectangle is drawn around the figure “10” with diagonals connecting the corners of the rectangle. The intersection of the two diagonals is used as the “center” of the numeral. It is marked with a plus sign in red called a cursor.

It is convenient to use various terms to describe the horizontal and vertical coordinates of the centers of the dots and the numeral “10”. These are shown below in the sketch of Figure 2 and in Table 1, where the letter “x” denotes the horizontal coordinate and the letter “y” the vertical.

Figure 2. Mask of numeral “10”, white dots and frame line.

The first step in the program is to find these coordinates for stamps from each of the 150 positions of the sheet. This was done by scanning the stamps and importing the scanned images into the CorelDraw program. In Figure 3 is displayed a photograph of the CorelDraw computer screen with a ten lepta stamp image and the mask of the “10”, dots and frame line.

Figure 3. CorelDraw screen with ten lepta stamp and mask below.

An enlarged view of the stamp on the same screen is shown in Figure 4 with the mask of the “10’s”, dots and frame line superimposed on the corresponding parts of the stamp in the lower inscription block.

Figure 4. Lower portion of ten lepta stamp with numeral and dot mask superimposed.

Note that the zero mask appears that it is not centered over the zero of the stamp. This is caused by variations in the figures “10” which were manually punched into the copper clichés of the printing plate. When adjusting the “10” mask over the stamp’s “10”, the inner oval of the zero was used because this seemed to give more consistent results. Figure 5 shows an enlarged view of the left numerals and the frame line with the measured coordinates. These measurements were repeated for the right numerals except that the frame line spacing was not measured there.

Figure 5. Left Numerals with coordinates of the various elements.

In Table 2 are listed the coordinates of the various elements which compose the numerals and white dots on both sides of the lower inscription block for the example stamp of Figure 5.

4. Calculating the Deviations of the Numerals and Frame Line Spacing

As mentioned above, the Microsoft Excel program was used to calculate the numeral deviations and frame line spacing of a stamp whose position was unknown and then to compare these deviations with the standard deviations of the 150 positions of the sheet to determine the position with the closest match.

   a. Data Input

The first step in the program is to enter the data from the CorelData screen into the assigned area of the Excel spreadsheet as shown below for an example ten lepta stamp:

The next step is to calculate the deviations of the numerals “10” from the center point between the two white dots. The formulas for these calculations are presented below:

  b. Calculate the Coordinates of the Center Between the Two Dots of the Left Numerals

• ldotx = the x coordinate of the left dot
• ldoty = the y coordinate of the left dot
• rdotx = the x coordinate of the right dot
• rdoty = the y coordinate of the right dot
• dxdot = rdotx - ldotx = the horizontal distance between the dots
• dydot = rdoty - ldoty = the vertical distance between the dots
• dotctrx = the x coordinate of the horizontal center between the dots = ldotx + 0.5 * dxdot
• dotctry = the y coordinate of the vertical center between the dots = ldoty + 0.5 * dydot

  c. Calculate the Horizontal and Vertical Deviations of the Left “Ten”

• dxl = tenctrx – dotctrx = x deviation
• dyl = tenctry – dotctry = y deviation

  d. Repeat the Calculations in b. above for the Right Numerals

• dxr = tenctrx – dotctrx = x deviation
• dyr = tenctry – dotctry = y deviation

  e. Calculate the Frame Line Spacing

fs = f2 – f1 = ldoty – f1

  f. Correct for Variations in the Size of the Stamps

If dimensions of several stamps are measured, it will be observed that there are slight differences in size. This could be caused by differences in the humidity of the air at the time of making the measurements or when the stamps were printed. To eliminate errors caused by these size differences, the five “deviations”, dxl, dyl, dxr, dyr, fs are all divided by the horizontal distance between the two white dots on either side of the left numerals “10”. ” These dimensionless variables are shown below:

• dxl/dxdot
• dyl/dxdot
• dxr/dxdot
• dyr/dxdot
• fs/dxdot

The above formulas are entered only once into the appropriate cells of the program spreadsheet and the calculations are made automatically. The formulas do not have to be entered again. Only the coordinate information for a new unknown has to be entered into the designated cells.

An example is shown below:

  g. Compare Unknown with Standard Deviations of 150 Positions.

The five deviation values dxl, dyl, dxr, dyr, fs are then automatically entered into cells where they are compared against standard deviations for all 150 positions. An example is shown below in Table 5 for only thirteen of the 150 positions (because of space limitations):

In order to determine which stamp positions have deviations which are closest to those of the unknown, a mathematical method is used which is called the least squares method. In this method, the differences between the deviations of the unknown and those of the standard 150 positions are calculated. These differences are then squared and their sums calculated. These “sums of the differences squared” are then sorted with the lowest values listed first. The positions which have the “sums” with the lowest values are the ones closest to the unknown.

In Table 6, the first five columns show the deviations for the unknown and the standard deviations for the first 13 positions (because of space limitations, only the first 13 positions are shown). The next five columns show the differences between the unknown and the standard position squared, multiplied by 1000 (to make the numbers easier to read). The last column shows their sums.

In the Excel program, the “Data” choice in the top menu has sub-menu with “sort” as one of the selections. We first select the block of cells containing the13 columns for all 150 positions and then “sort” the list based on the last column “the sum of the differences squared”. The sums with the lowest values will be listed at the top and their position numbers in the first column at the left will tell us which positions are closest to the unknown. Table 6 shows these data for the first 13 positions before they have been sorted. Table 7 shows the sorted list of sums for those positions with the closest sums at the top.

The position with the lowest “sum” is at the top of the column at the extreme right and the position corresponding to this sum is at the top of the column at the extreme left. This position with the closest match to the unknown is position 20. The positions with the next closest matches are listed further down in the column at the left. In this case they are positions 116, 136, 93, 38, 126, etc.

In some cases, the first position is not the correct one for the specific unknown. The researcher can check each of the closest positions by comparing the unknown with other stamps from those positions. Often, there are other clues which are helpful in confirming which of these closest positions is the correct one. Such clues include ink spots, scratches, white spots, other plate flaws, inner and outer frame line spacing, etc. The important advantage of this plating program is that it narrows the number of possible correct positions to just a few. In the case given above, position 20, which was the first choice, was the correct position.

Marginal Stamps

Stamps belonging to positions at the edges of the sheet (marginal stamps) are identified by having heavy frame lines at the side facing these edges. If a stamp has a heavy frame line at its right side, for example, we know that it comes from the right edge of the sheet: positions 20, 30, 40, etc. Corner stamps have heavy frame lines at two sides and their positions are immediately revealed.

When using this plating program to identify a marginal stamp, it is not necessary to compare the deviations of the unknown with the deviations of all 150 positions. For a right marginal stamp, for example, the comparison could be made against only 13 positions: 20 to 140. In the plating program described here, this is accomplished by having different sections of the Excel spreadsheet devoted to each of the four types of marginal stamps: top, bottom, left and right. In each such section, the data input area is located at the top and the deviations for the marginal positions immediately below. Table 8 below shows the section of the spreadsheet for the right marginal stamps. The differences of the deviations from the unknown, their squares and the sums are not shown here, but in the actual spreadsheet, these values are automatically calculated and values sorted with the closest match to the unknown at the top, as described above for the entire 150 positions.

Examples and Results Using the Plating Program

During the development of this program, there was constant interaction between this author and Kostas Hatzis who was instrumental in encouraging me to work on it. His efforts in successfully plating several hundred ten lepta stamps was very useful in confirming that independent results could be obtained by another researcher. In a companion article, Kostas explains his use of the program and presents his results in plating these 10 lepta stamps.

Summary

1. Although a computerized plating program for the large Hermes head stamps has been used for over 20 years, it uses obsolete hardware and software which could not be converted to modern Windows-based computer systems. A new program was successfully developed for use on modern computers.
2. The program requires a scanner with at least a 600 dot per inch resolution, the Microsoft Excel spreadsheet program and the CorelDraw paint program. Alternate programs for Excel and CorelDraw can be used. However, the paint program must have the ability to locate points on the stamp image in horizontal and vertical coordinates accurate to three decimal places.
3. The program is based on the horizontal and vertical deviations of the numerals “10” from the center point between the two white dots on either side of the numerals. It also measures the distance between the left dot of the left “10” and the frame line immediately below. There are thus five measurements made: the horizontal and vertical deviations of the left numerals, the same deviations for the right numerals and the frame line spacing.
4. Standard deviations are measured for stamps whose positions have been identified by ink spot patterns for all 150 positions. These standard deviations are arranged in cells of the Excel spreadsheet program.
5. The measurements of the stamp coordinates from the CorelDraw program are entered into allocated cells of the spreadsheet program and the deviations automatically computed using formulas previously entered.
6. The deviations of a stamp whose position is unknown are compared with the standard deviations of the 150 positions using the “least squares method”. The table listing these positions is then sorted according to the sums of the differences squared and those positions with lowest “sums” are the closest matches to the unknown. A visual comparison of the first few closest positions will in most cases immediately identify the correct position of the unknown.
7. A copy of the spreadsheet program will be made available to anyone interested in plating the 10 lepta large Hermes head stamps, free of charge.

Conclusion

Kostas Hatzis, who has plated hundreds of ten lepta stamps using this program, has indicated that he will soon start measuring the standard deviations for the 40 lepta stamps. After that project is completed, he will work on establishing standard deviations for the other values. I am indebted to Kostas for his enthusiasm in pursuing this project and proving the successful use of the program.

go to top       home