The Original Test Image Showing the Piece Selected for Magnification
The images below show the results of magnifying that tiny piece of the image. On the left is a piece from the original “Save 0” image, which was the first time the image had been compressed by the JPEG process from 13+ million bytes (one byte for each of the three primaries, red, green, and blue, per pixel multiplied by the width of 1817 pixels and the height of 2400 pixels, or 13,082,400 bytes) to a mere 197 kilobytes, or 202,390 bytes.
A number of very faint JPEG artifacts may be seen in the white area near the edges of the black lines. The hardest thing for the JPEG process to portray is an abrupt change from black to white, which is what we have here.
On the right is the same precise piece of the image, also at 1200×. The JPEG artifacts are slightly more noticeable in this 30th consecutive save than they are in the original save. They are still very faint, and in full-size displays—whether on screen or in a print-out—they are still nearly invisible. It would take an extremely sharp—almost preternaturally sharp—eye to detect any of these artifacts in a full-size display of this image.
Keep in mind that in the magnified views, each of those little squares is a pixel. Each of them has sides that are precisely 1/2400th of the vertical dimension of the overall image. If you printed the image in a format twenty-four inches high, each pixel would measure exactly one-hundredth of an inch on each side and would be one ten-thousandth of a square inch in area.
As Wikipedia puts it, “The mathematical relationships that define these color spaces are essential tools for color management. They allow one to translate different physical responses to visible radiation [from color prints], illuminated displays, and recording devices such as digital cameras into a universal human color vision response.”
Color would never be the same again. It had been reduced to numbers: Specify the x and y value and you’ve defined the color. This eventually wreaked havoc with what painters thought they had been doing for centuries.
And therein lies a tale. Fast forward to the 1970s, when I first became intrigued by all of this. By 1975 I had become interested in a Dutch painter named Piet (Peter) Mondrian (1872–1944). As Wikipedia states it, “He evolved a non-representational form which he termed neoplasticism. This consisted of white ground, upon which he painted a grid of vertical and horizontal black lines and the three primary colors.”
Ah . . . and there’s the rub. Like most painters then—as well as many nowadays—Mondrian mistakenly believed that the primary colors were red, yellow, and blue. This was because paints are a subtractive medium and thus operate in the CMYK color space. The CMYK primaries are magenta, yellow, and cyan—magenta being (roughly) red and cyan being (roughly) blue. Here is a genuine Mondrian painting, titled “Composition in red, blue, and yellow” (1930):
Of course, human eyes see in red, green, and blue (RGB), and our monitors and TVs display the same primary colors, which are additive. These Venn diagrams illustrate the difference:
Note that for each of these, the primary color has a complimentary color—its exact opposite—which shows as a smaller two-color overlay directly opposite the primary circle itself. This is true for both sets, the compliments in the RGB color space being the primaries in the CMYK gamut, and the CMYK compliments being the primaries in the RGB gamut.
To top this off, the CIE Color Space pictured above is not even accurate. It cannot possibly be accurate, because some of the colors near the outer edges of the diagram can’t be reproduced by paints, inks, or colored light sources. This little CIE color space diagram shows why:
Colors that lie outside the RGB (often called “sRGB”) triangle cannot be reproduced by your computer monitor—or your TV, either. Colors that lie outside the CMYK polygon (in yellow above) cannot be reproduced by paints or printing inks.
But wait—there’s more. Our human visual system operates in a four-primary color space, where the primaries are red, green, yellow, and blue. These four are the only colors that cannot be described in terms of a combination of any other colors—proving that they are primaries in our visual system. True, monitors and TVs produce yellow by mixing red and green, but no human would ever describe yellow as “greenish-red” or “reddish-green”—it just doesn’t work.
Yet, the other two CMYK primaries (the “K” stands for black) can be described in terms of other colors: cyan as blue-green or greenish blue, and magenta as bluish-red.
So in 1975 I boldly decided to create a new Mondrian-type painting that would include all four of the human primary colors. I titled it “The Greening of Mondrian.” It’s painted on 12 by 16 canvas board, framed, and hangs in my family room. I recently created a digital version of this painting in Photoshop. I made the black line elements out of blocks (using the Rectangular Marquee Tool), which I filled in with black. When the pattern was complete, I filled in the appropriate enclosed spaces with pure colors from the sRGB gamut available in Photoshop.
By the way, should you imagine that creating a picture like this would be an easy matter, perish the thought. That’s what I thought before I started. It was anything but easy. I must have made about 25 studies of this painting before I got an arrangement that looked pleasing to my eye. Perhaps I’m too fussy, but I don’t think so.
After doing all this, I did some Googling on Mondrian’s paintings and discovered that I was truly innovating when I created “The Greening of Mondrian”: Because Mondrian never used diagonal elements in any of his paintings. His lines were all strictly horizontal or vertical. Perhaps this was why it proved so difficult to get a pleasing arrangement, although it might have been just as hard to make a horizontal/vertical picture. By using diagonal elements, I was trying to ensure that I was not copying anything Mondrian had actually done. But I didn’t know he had never used any diagonal elements. Back in 1975 there was no Google to search with. I had only library reproductions (in books) to go by.
For more than you probably will want to know about color models for printing, see THIS.
Degradation in Successive Saves of a JPEG FileThe second view of “The Greening of Mondrian,” on the right labeled “30th Save,” deserves some explanation. You will find “experts” on image preservation who will tell you that you should always work with an original (RAW or bitmapped) image, and after editing it, save the final version as a JPEG (Joint Photographic Experts Group), which has a jpg filename extension. For example, check out this website.
Some of the text there reads: “Most of you probably know that JPEG is [a] lossy compression method, meaning compression permanently throws out data and detail. Luckily, a typical compression can save 10 times the space of an uncompressed image without sacrificing much noticeable quality. However, if the image is repeatedly compressed and saved, artifacts introduced during compression become more and more obvious.”
This may be true, but the losses are seldom obvious to the eye or in the printed image. To test the theory that saving and re-saving a picture in JPEG format degrades the image, I made an extra copy of the “Greening of Mondrian” file and saved, reloaded, and saved it again--until I had done this thirty times. The result is the right-hand image.
There is no discernible difference between the two versions, the original and the 30th Save. The hues are all identical, and the saturation and brightness are all either 99% or 100% when measured in Photoshop. It could be that the simplicity of this image is the reason almost no degradation occurs. The JPEGs above were made with a Quality 10 setting. The original image is 12.4 MB (3 × 1817 × 2400 pixels); the JPEG image is 197 KB (202,390 bytes), a savings of more than 98%. In a typical photograph, the savings at level 10 would be about 90%; the 98% here is the result of the very simple image used.
On the PetaPixel website (at the link given above) a fellow by the name of “Grundle” saved a picture 500 times and made a video out of the results. Here is a picture of the original picture and the 500th saved version:
Only a fool would edit the same picture 500 times, saving each edit as a JPEG file, and the resulting degradation is very noticeable. But wait—notice that the young woman’s face and hair are mostly just as good looking in the 500th save as they were in the original. In fact, the original picture Grundle used has noticeable pixelation in the background. This is especially noticeable in the right-hand portion of the picture, but even the odd pixelation on the woman’s right cheek is present in the original, albeit as an almost invisible artifact. These strange pixelations, which would never be produced by a decent digital camera, are the source of the awful-looking pixelations in the 500th save.
The lesson here is that if you start with a slightly pixelated image and edit it numerous times, saving each edit as a JPEG file, you are going to end up with a badly pixelated image. Not good.