Walk through the computer department at Sears, and every computer on display has an animation in progress, full of colorful pictures that invite you to buy the computer. How does the computer make those pictures?
Most computer displays are constructed around a standard color television tube, which we will call the video display. Give it a quick review.
Technically, it is a CRT--a vacuum tube with a flat face and a phosphor coating on the inside of the face. At the back end of the tube, a beam of electrons is created by a cathode. Magnets 'n stuff cause the electrons to form a beam that scans over the phosphors, causing some of them to glow. The scanning pattern is a set of horizontal lines, called the raster. That makes the dots, and the pattern formed by the dots is seen as the picture.
For color pictures, there are three kinds of phosphors: red, green, and blue, and each screen dot consists of a red, green, and blue clumps, as shown above.
All computer displays are raster displays, which means...
And that's not a problem, that's a feature! Every sixtieth of a second, the screen display is created and then allowed to fade away. The computer then refers to a screen database, often called a bitmap, and creates a completely new display. The fast fad allows the computer to display amazing animations--60 new images every second.
Of course, we often use the computer to store our word processing and spreadsheets. As we stare at the images of these documents, the computer is renewing the image 60 times a second, just as though we were in the middle of a rousing video game.
Any picture you see on the computer screen is recorded in the screen bitmap--a computer database used to create the mosaic of screen dots. This database contains brightness values. For a grayscale picture, there is one number for each dot in the picture, specifying how bright to make the dot. The brightness range is 0 to 255. You can store any number in that range in a single byte of memory, so the datafile contains one byte for each dot.
For color images, we must store the brightness values for the red, green and blue phosphors of each dot. Again, the brightness range is 0 to 255, so we need three bytes for each dot, for a complete range of color.
The bitmap lives in a special set of chips called
VRAM: video random access memory.
The bigger your computer screen, and the more colors you want to see on
the screen, the more VRAM you need. Since people are buying larger monitors,
the amount of VRAM installed in typical desktop computers has increased
from one megabyte to two to four megabytes over the past few years.
Back in the 1970's, the days of the first desktop computers, VRAM was prohibitively expensive: "a buck a byte" or more. Images were stored in the most compact way possible:
One pixel = one bit
Thus each pixel was either black or white. The data for eight pixels could be stored in one byte, and the entire bitmap for a 14" display (640 by 480 pixels) could be stored in 64 KB of VRAM. Black and white bitmaps work fine for word processing and diagrams, blueprints, and some types of charts. Collectively these types of images are known as line art.
But this type of image was terrible for pictures:
|
Here's a typical color picture... |
 |
Here's the same picture, stored in a one-pixel = one bit bitmap. All parts of the picture with a brightness of less than 50% are black. All other dots are white. |
Suppose we want to show shades of gray on the display. To do so we must enlarge the bitmap.
|
Here the bitmap has two bits per pixel, so each dot can be: 00, 01, 10, or 11. That gives us four shades of gray. |
|
Here the bitmap has four bits per pixel, so the range is 16 shades of gray. You might almost think this picture is "normal" until you see the next picture... |
|
This bitmap has eight bits of information per pixel, for 256 shades of gray. Look carefully, and you will see more folds in the nurse's sleeve, and less graininess in the drape behind her, and less blotchyness in the wall. Still, you might like the upper picture, because it has more contrast. |
Virtually all monitors are limited to 256 levels of gray, so for grayscale:
One pixel = one byte
Grayscale monitors, or color monitors set to grayscale, are still very useful, especially for people who print newspapers and magazines. If you are printing color pictures on a grayscale printer, you can use a grayscale setting to accurately preview how your printed material will appear.
 |
Recall that in a color monitor, each screen dots is really a trio of red, green, and blue dots, as shown in the top row. You can produce many colors by simply turning on or off different combinations of the pixels, as shown in the second row. But the video monitor can also vary the intensity of each subpixel, and their relative intensities determine the pixel color. The color range for each subpixel is 0 to 255. If all three colors are 255, we see the color white. If all three colors have any value less than 255, we see gray. By assigning different values to the three colors, one dot can be any of 16 million colors (256x256x256). |
Most of the images we put on a video monitor come from other media, such as photographs, printed illustrations, and slides. Each media has its own color system. Color create additive color, but they do so by absorbing color in the emulsion dyes of the 35 mm film. As we shall see later, photographic prints and printed illustrations create color by subtracting colors, not adding them together. Consequently, digitizing a picture involves re-creating a picture in a new color system.
Recall that we must store all display images in VRAM. How much VRAM does it take to store a color bitmap?It all depends on how you do it. There are three currently accepted ways:
Under the Windows 95 and Mac operating systems, you have the option of setting your computer for "hundreds", "thousands" or "millions" of colors. If you select hundreds, you'lll see a little picture like the one shown below. It is the standard Macintosh color table, or palette. It consists of 256 colors. There is a completely analogous palette for Windows. Everything on the Mac (or Windows) desktop is created using the standard palette.
Since there are only 256 colors in the palette, we can number them and store the numbers in bytes. So for color table color our bitmap will have:
One pixel = one byte
Just like the grayscale, bitmap, but we have color! Simple!
Although we can number the colors from 1 to 256, we have not told the computer how to make the colors. Each color in the table represents three brightness values, for the red, green, and blue phosphors. In the diagram below, the bitmap is on the left. Only the first 16 entries are shown. At lower right is the color table. For each color shown there are really three brightness settings. Again, we have only shown part of the table. Finally, there is the screen, at upper right. Thus drawing each pixel is a two-step process: get the color table entry from the bitmap, and get the corresponding brightness settings from the color table. Color tables are often called Color Look Up Tables, or CLUTs. Ugly acronym for beautiful colors, but there you are.
Remember how great it was to get a big box of crayons, with lots of different colors? And remember how even then you couldn't find the exact color you wanted? Even 256 colors is not enough for all pictures. Below is a picture of a nurse, created using the sytem color table, the 256 colors used in all displays of menus, dialog boxes, icons, and other system stuff:
The grainy appearance of the picture is due to a process called dithering. Dithering occurs when the original image contains colors than are not in the color table. The display program approximates the missing colors by creating dot clusters of other colors. If you move back from the screen, the graininess will disappear and the image will resemble one in true color.
Compare the above to the same picture below, created with a custom color table, containing the best possible set of colors for the picture. A graphics program can automatically create a custom color table for a 24-bit image.
Although a color table contains only 256 colors, there are 16 million possible choices to choose from. So a color table can contain whatever colors are necessary. Here's a chameleon, with his mostly green color table:
...And here's a girl and her grandfather, with their mostly red and pink table:
Picture composed of 255 colors or less can be perfectly represented with a color table, and pictures with more than 256 colors can be well handled with inobtrusive dithering.
But the color table method fails with multiple pictures.
Suppose we wish to put both the chameleon and the girl and grandfather pictures on the screen, in two separate windows. The operating will set the color table to match the active window (the last one clicked on with the mouse) and display both pictures with that table. So the inactive window will look like this:
Or this:
If you combine the two pictures into one graphic, in one window, you have to create a color table that serves both pictures well. For pictures like those above, there are simply too many colors; the result will be a mishmash.
Digital video obviously consists of multiple images, and is normally created in 24- or 16-bit color. You can process digital video to give the entire video one color table, but the resulting video will look awful If you are limited to 8-bit color, stick to black-and-white digital video!
There is generally no relationship between the colors in a CLUT; they are simply the most commonly used hues in the picture. But in graphics programs, many editing techniques derive from color models that assume that the picture color is in a continuous spectrum. Consequently, you can't brighten, enhance the contrast, arbitrarily or sharpen CLUT pictures. Other editing techniques, such as changing the picture size, produce poorer results in color table pictures. The computer programs for these methods all assume use of the color channel model, discussed below.
If you select a monitor setting of "thousands" or "millions" on your computer, you not only get more colors, you also get a new color model. We'll call it the channel model, because this term is used in Photoshop to refer to the way color is being processed. The Mac represents thousands of colors with the following picture:
This is a spectrum, not a palette. It represents the fact that the red, green, and blue phosphors can be varied continuously.
The following diagram represents how the bitmap in VRAM looks with 16-bit color:
The bitmap is on the left. The entry for each pixel is 16 bits wide, divided up into three five-bit components. In five bits, we can store values from 0 to 31, so each pixel can be set to 32 different brightness values. For example:
There are (32x32x32), or 32,768 possible combinations. Thus 16-bit color is usually labeled "Thousands of colors." Note that the brightness represented by the 32 values is normalized over a brightness range of 256. Thus the actual brightness for 32 value would be 256 times the brightness for the lowest value.
Perhaps you are thinking,
I thought the brightness range was 256?
Correct; the phosphors can be set to about 256 distinct values. But with 16-bit color, you use one-eighth of the possible values, evenly spread through the spectrum: 0, 8, 16, 24, ... 256. This is done to keep the bitmap smaller.
Most pictures are handled well in 16-bit color, but very pure color color gradients sometimes break into bands. This is seen in the blue sky over this mountain village:
You might wonder why one would limit the color brightness range to 32 values. This allowed computer manufacturers to offer pretty good color on a standard (14") monitor with one megabyte of VRAM. Also, the previous generation of microprocessors didn't have to work so hard sustaining and refreshing the image. Sixteen-bit color simply represented a cost-effective compromise.
With 16-bit color, it is possible to display multiple images in their true colors--no pink chameleons or green children. It is also possible to display imges which actually have a color range of millions of colors, usually with few or no artifacts, banding being a rare exception. The channel processing filters and features of advanced graphics programs, such as PhotoShop, work well on a 16-bit display, even though they are designed for a 24-bit color display. Digital video looks fine in 16-bit color. Finally, anti-aliasing, a technique which eliminates harsh "jaggy" effects in text and charts, works fine in 16-bit color, and doesn't work at all in a color table system.
The standard SVGA display for Windows95 has a resolution of 800 by 600, which just happens to be exactly how many pixels can be represented in one megabyte of VRAM, in sixteen-bit color.
For people in the graphic business, only 24-bit color will do. For each pixel, the bitmap contains three bytes for each color, allowing each color phosphor to be set to any possible brightness value. Since there are 256 x 256 x 256 (about 16 million) possible colors, this type of display is called "millions of colors." Actually the bitmap will contain 32 (four bytes) for each pixel, because the computer can process data in multiples of four much faster than multiples of three. The fourth byte is often used to create special visual effects; as a fourth channel (see below) Often this channel is used to specify transparency, which can be used to create clouds, fog, and other visual effects.
At one time, the processing needed to sustain and change a large 24-bit display was a big job for desktop computers; artists had to wait seconds or even minutes for their display to "come back" after making an edit. Today's Pentium and PowerPC processors can handle these displays with little or no redraw lag.
A way to visualize what's going on is to visualize the screen display as superimposed red, green and blue pictures. Usually each of these is shown in grayscale, representing the color intensity. In many graphics programs, these three components of the image are called channels. These are the red, green, and blue channels for the picture we saw earlier:
We have seen how the computer image is stored in a bitmap in the VRAM chips. The limited VRAM capacity of early computers limited them to black-and-white images. When color monitors became popular, VRAM and processing limitations resulted in the use of color table driven displays. Today most computers have enough VRAM to use 16-bit or even 32-bit color, even with large displays.
Here are three common types of commercial monitors. The inch figure measures the length of the diagonal between top left and bottom right corners. The actual image size is usually about an inch less. Larger monitors often have a multiscan capacity, allowing the user to pick the resolution.
The amount of VRAM is given to the nearest 0.5 Meg for small amounts and to the nearest 1.0 Meg for larger amounts. I didn't put up the numbers for 21" monitors, but you can imagine that such a monitor would have limited color depth unless installed with a large amount of VRAM.
| Diagonal Measure | Pixel Resolution (horizontal x vertical) |
VRAM needed for 8-bit color |
VRAM needed for 16-bit color |
VRAM needed for 32-bit color |
| 14 inch | 640 by 480 |
0.5 Meg |
1 Meg |
1.5 Meg |
| 15 inch | 640 by 480 800 by 600 |
0.5 Meg 0.5 Meg |
1 Meg 1 Meg |
1.5 Meg 2 Meg |
| 17 inch | 640 by 480 800 by 600 1024 by 768 |
0.5 Meg 0.5 Meg 1.0 Meg |
1 Meg 1 Meg 2 Meg |
1.5 Meg 2 Meg 6 Meg |
Clearly if you have a limited amount of VRAM, you have to trade off resolution. Some users always opt for very high resolution, and they are apt to have limited color depth. Occasionally you see the other extreme--a computer user that doesn't realize he can set his monitor to a resolution higher than 640 by 480. You can read fine print on their computers from across the room...
A computer picked at random may have 8-bit, 16-bit, or 24-bit color capacity. Unless you check the appropriate control panel in the operating system, you can't tell for sure. Moreover, some computers can be upgraded by simply adding a chip or two of VRAM. Since VRAM costs about $13 per meg, this is a cheap way to rejuvenate a computer. But other computers have no place to plug in extra VRAM. They can only be upgraded by adding a color card--an electronic circuit board that plugs into an expansion slot inside the computer. Color boards are more expensive ($100 and up) than VRAM chips and they may have limited compatability.
Yes. Most digital pictures really only have about 200 to 300 distinct hues. So long as a custom color table is used, most pictures look fine in 8-bit color. But since dithering will create color artifacts, scientific illustrations should be reproduced with as broad a spectrum as possible.
NO. Expect slight differences even when only one picture is displayed. And when more than one picture appears on the screen, the 8-bit monitor will misrepresent some or all of the pictures. Always avoid using an 8-bit monitor if possible.
Usually not. Professional graphic artists can do this by selecting pictures which can easily share a common color table. But if you must present the results of a study (or pictures of your family) the results are usually terrible.
It is most critical in creative work such as commercial illustration, animation and 3-D modeling. The brightness variations in these pictures are arbitrary and thus more extreme, and this may lead to banding in 16-bit color. Most pictures of medical and scientific material will look acceptable in 16-bit color.
Usually not. You will see dithered versions of your 16-bit pictures, making them look grainy and blurry. Even with good dithering, you will not be able to judge the true colors of your pictures.
Yes. You should edit your pictures in 32-bit color, so the editing program can take full advantage of three wide channels of color, plus a fourth channel for special effects. After editing, you can create 16-bit versions of the pictures for display or distribution.
6. I prepared a presentation in PowerPoint on my laptop. It looked great on my laptop. But when I hooked up my laptop to a screen projection system, it looked awful. Other presenters used the projector with great success. What went wrong?
A limited "monitor out" connection. Your laptop sent your images to the projector via a "monitor out" connection. Your laptop may be able to display 16-bit color on its display, but only output 8-bit color to the projector.
