Don, You make great points and in the past I have read about these problems with color analysis with scanners, computer display, file formats, and software. Comparing one image from one scanner to another from another scanner, each run on different computers and different software creates potential problems. However, my objective was not so much to identify a specific real life color, but to identify which stamps have the same color and which were of a different color, regardless of what one call that color. As long as I used one single scan to scan all the images and then used the same file formats, same software program, same computer to analyze each image, etc, on that basis, then the relative differences in color should be measurable. In fact, this is what my analysis is indicating. I have 24 of these stamps all scanned in one pass. I have now analyzed seven of the 24 stamps. The desired results should be such that there are three statistically different colors found, as there are three color varieties. If I come up with four colors, either my approach is faulty or I have found a new color variety (Chances of that are slim to none). If I find less than three relative colors, then I probably have only two of the three stamp varieties in my group of 24 stamps. So far, with seven stamps done, I have three clearly defined colors identified. The "common" color stamps are almost exactly matched by RGB values, certainly by any statistically significant measure. The difference between relative colors I found is meaningful, even though these color differences are a matter of shades. To further test my approach, I measured the color of all four margins on each stamp (top, bottom, right and left). This served to test the robustness of my approach as each border of each stamp should have the same RGB values and those values should be a good match between stamp to stamp. I found that the margins (which are not perfectly white as you can see). I found that the margins all have RGB values of R 240-242, G 224-227, and G 189-195. The other three (reddish/rose shades) variety colors are also closely defined in their respective RGB values. The grouping of the stamps into these three color RGB levels indicates that my analysis is accurate. All of my stamps so far fall into one of these three types.
Color Type 1 R 180 G 72 B83
Color Type 2 R217 G99 B115
Color Type 3 R196 G118 B120
There is absolutely no question that these color types are random or just dispersion. Every stamp so far falls into one of these types. There has been no fourth type.
I have to finish the final 17 stamps, but I strongly expect to see them all fall into the above groupings.
The final exercise will be to determine which of my color types is the #38, #43, and #43a stamp.
My approach solves one of the great color analysis problems in stamp collecting. However, it requires that the collector acquire a statistically significant number of appropriate stamps, scan them all at once on one scanner in one pass, and then analyze them with the same software on the same computer. To further enhance the accuracy of my analysis, I took 7 separate RGB color measurement on each stamp all from various, different place on the stamp and I did this for identical positions on each stamp. Those seven measurements do not include the four stamp margin color measurements I made.
Had I scanned the images separately on different scanners, downloaded images from different web sources, used different software programs, my analysis would be subject to some color analysis issues as you point out. This then would manifest itself in a massive dispersion and range of colors I measure for what should be a set of stamps with only three relative colors. This dispersion did not happen with my approach. I have found only three different colors, as should be the case.
More to follow. |
Posted 12/04/2018 3:20 pm
