RGBA versus Color-Index Mode

In either color-index or RGBA mode, a certain amount of color data is stored at each pixel. This amount is determined by the number of bitplanes in the framebuffer. A bitplane contains 1 bit of data for each pixel. If there are 8color bitplanes, there are 8 color bits per pixel, and hence 28 = 256 different values or colors that can be stored at the pixel.

Bitplanes are often divided evenly into storage for R, G, and B components (that is, a 24-bitplane system devotes 8 bits each to red, green, and blue), but this isn't always true. To find out the number of bitplanes available on your system for red, green, blue, alpha, or color-index values, use glGetIntegerv() with GL_RED_BITS, GL_GREEN_BITS, GL_BLUE_BITS, GL_ALPHA_BITS, and GL_INDEX_BITS.

Note: Color intensities on most computer screens aren't perceived as linear by the human eye. Consider colors consisting of just a red component, with green and blue set to zero. As the intensity varies from 0.0 (off) to 1.0 (full on), the number of electrons striking the pixels increases, but the question is, does 0.5 look like halfway between 0.0 and 1.0? To test this, write a program that draws alternate pixels in a checkerboard pattern to intensities 0.0 and 1.0, and compare it with a region drawn solidly in color 0.5. From a reasonable distance from the screen, the two regions should appear to have the same intensity. If they look noticeably different, you need to use whatever correction mechanism is provided on your particular system. For example, many systems have a table to adjust intensities so that 0.5 appears to be halfway between 0.0 and 1.0. The mapping generally used is an exponential one, with the exponent referred to as gamma (hence the term gamma correction). Using the same gamma for the red, green, and blue components gives pretty good results, but three different gamma values might give slightly better results. (For more details on this topic, see Foley, van Dam, et al. Computer Graphics: Principles and Practice. Reading, MA: Addison-Wesley Developers Press, 1990.)

RGBA Display Mode

In RGBA mode, the hardware sets aside a certain number of bitplanes for each of the R, G, B, and A components (not necessarily the same number for each component) as shown in Figure 4-2. The R, G, and B values are typically stored as integers rather than floating-point numbers, and they're scaled to the number of available bits for storage and retrieval. For example, if a system has 8 bits available for the R component, integers between 0 and 255 can be stored; thus, 0, 1, 2, …, 255 in the bitplanes would correspond to R values of 0/255 = 0.0, 1/255, 2/255, …, 255/255 = 1.0. Regardless of the number of bitplanes, 0.0 specifies the minimum intensity, and 1.0 specifies the maximum intensity.

Figure 4-2 : RGB Values from the Bitplanes

Note: The alpha value (the A in RGBA) has no direct effect on the color displayed on the screen. It can be used for many things, including blending and transparency, and it can have an effect on the values of R, G, and B that are written. (See "Blending" in Chapter 6 for more information about alpha values.)

The number of distinct colors that can be displayed at a single pixel depends on the number of bitplanes and the capacity of the hardware to interpret those bitplanes. The number of distinct colors can't exceed 2n, where n is the number of bitplanes. Thus, a machine with 24 bitplanes for RGB can display up to 16.77 million distinct colors.

Dithering

Advanced

Some graphics hardware uses dithering to increase the number of apparent colors. Dithering is the technique of using combinations of some colors to create the effect of other colors. To illustrate how dithering works, suppose your system has only 1 bit each for R, G, and B and thus can display only eight colors: black, white, red, blue, green, yellow, cyan, and magenta. To display a pink region, the hardware can fill the region in a checkerboard manner, alternating red and white pixels. If your eye is far enough away from the screen that it can't distinguish individual pixels, the region appears pink - the average of red and white. Redder pinks can be achieved by filling a higher proportion of the pixels with red, whiter pinks would use more white pixels, and so on.

With this technique, there are no pink pixels. The only way to achieve the effect of "pinkness" is to cover a region consisting of multiple pixels - you can't dither a single pixel. If you specify an RGB value for an unavailable color and fill a polygon, the hardware fills the pixels in the interior of the polygon with a mixture of nearby colors whose average appears to your eye to be the color you want. (Remember, though, that if you're reading pixel information out of the framebuffer, you get the actual red and white pixel values, since there aren't any pink ones. See Chapter 8 for more information about reading pixel values.)

Figure 4-3 illustrates some simple dithering of black and white pixels to make shades of gray. From left to right, the 4 × 4 patterns at the top represent dithering patterns for 50 percent, 19 percent, and 69 percent gray. Under each pattern, you can see repeated reduced copies of each pattern, but these black and white squares are still bigger than most pixels. If you look at them from across the room, you can see that they blur together and appear as three levels of gray.

Figure 4-3 : Dithering Black and White to Create Gray

With about 8 bits each of R, G, and B, you can get a fairly high-quality image without dithering. Just because your machine has 24 color bitplanes, however, doesn't mean that dithering won't be desirable. For example, if you are running in double-buffer mode, the bitplanes might be divided into two sets of twelve, so there are really only 4 bits each per R, G, and B component. Without dithering, 4-bit-per-component color can give less than satisfactory results in many situations.

You enable or disable dithering by passing GL_DITHER to glEnable() or glDisable(). Note that dithering, unlike many other features, is enabled by default.

Color-Index Display Mode

With color-index mode, OpenGL uses a color map (or lookup table), which is similar to using a palette to mix paints to prepare for a paint-by-number scene. A painter's palette provides spaces to mix paints together; similarly, a computer's color map provides indices where the primary red, green, and blue values can be mixed, as shown in Figure 4-4.

Figure 4-4 : A Color Map

A painter filling in a paint-by-number scene chooses a color from the color palette and fills the corresponding numbered regions with that color. A computer stores the color index in the bitplanes for each pixel. Then those bitplane values reference the color map, and the screen is painted with the corresponding red, green, and blue values from the color map, as shown in Figure 4-5.

Figure 4-5 : Using a Color Map to Paint a Picture

In color-index mode, the number of simultaneously available colors is limited by the size of the color map and the number of bitplanes available. The size of the color map is determined by the amount of hardware dedicated to it. The size of the color map is always a power of 2, and typical sizes range from 256 (28) to 4096 (212), where the exponent is the number of bitplanes being used. If there are 2n indices in the color map and m available bitplanes, the number of usable entries is the smaller of 2n and 2m.

With RGBA mode, each pixel's color is independent of other pixels. However, in color-index mode, each pixel with the same index stored in its bitplanes shares the same color-map location. If the contents of an entry in the color map change, then all pixels of that color index change their color.

Choosing between RGBA and Color-Index Mode

You should base your decision to use RGBA or color-index mode on what hardware is available and on what your application needs. For most systems, more colors can be simultaneously represented with RGBA mode than with color-index mode. Also, for several effects, such as shading, lighting, texture mapping, and fog, RGBA provides more flexibility than color-index mode.

You might prefer to use color-index mode in the following cases:

• If you're porting an existing application that makes significant use of color-index mode, it might be easier to not change to RGBA mode.
• If you have a small number of bitplanes available, RGBA mode may produce noticeably coarse shades of colors. For example, if you have only 8 bitplanes, in RGBA mode, you may have only 3 bits for red, 3 bits for green, and 2 bits for blue. You'd only have 8 (23) shades of red and green, and only 4 shades of blue. The gradients between color shades are likely to be very obvious.

In this situation, if you have limited shading requirements, you can use the color lookup table to load more shades of colors. For example, if you need only shades of blue, you can use color-index mode and store up to 256 (28) shades of blue in the color-lookup table, which is much better than the 4 shades you would have in RGBA mode. Of course, this example would use up your entire color-lookup table, so you would have no shades of red, green, or other combined colors.

• Color-index mode can be useful for various tricks, such as color-map animation and drawing in layers. (See Chapter 14 for more information.)

In general, use RGBA mode wherever possible. It works with texture mapping and works better with lighting, shading, fog, antialiasing, and blending.

Changing between Display Modes

In the best of all possible worlds, you might want to avoid making a choice between RGBA and color-index display mode. For example, you may want to use color-index mode for a color-map animation effect and then, when needed, immediately change the scene to RGBA mode for texture mapping.

Or similarly, you may desire to switch between single and double buffering. For example, you may have very few bitplanes; let's say 8 bitplanes. In single-buffer mode, you'll have 256 (28) colors, but if you are using double-buffer mode to eliminate flickering from your animated program, you may only have 16 (24) colors. Perhaps you want to draw a moving object without flicker and are willing to sacrifice colors for using double-buffer mode (maybe the object is moving so fast that the viewer won't notice the details). But when the object comes to rest, you will want to draw it in single-buffer mode so that you can use more colors.

Unfortunately, most window systems won't allow an easy switch. For example, with the X Window System, the color-display mode is an attribute of the X Visual. An X Visual must be specified before the window is created. Once it is specified, it cannot be changed for the life of the window. After you create a window with a double-buffered, RGBA display mode, you're stuck with it.

A tricky solution to this problem is to create more than one window, each with a different display mode. Then you must control the visibility of the windows (for example, mapping or unmapping an X Window, or managing or unmanaging a Motif or Athena widget) and draw the object into the appropriate, visible window.