Nearly all computer displays today are capable of 16 and 32 bit color. "Bit," in this case, describes the number of settings that are available to store various shades of color.

B&W

Here, in its most simple fashion, a single "bit" can be toggled on and off to reproduce a black and white image. Unlike a black and white television set, here, there is only black and white. Any shade of grey that is closer to white than black is set as white while conversely any shade closer to black than white is set to black.

To find how many colors can be represented simultaneously at a current bit depth, one must simply understand that the bit depth is simply a function of:

2^{# of bits} = number of colors

Hence, this image with 4 bit color would have 16 different colors to choose from
(2 x 2 x 2 x 2).

4 bit color

Notice that only certain areas of the image preserve their anti-aliasing (such as the right side of the cube and the top of the cone). This has to do with the color palate. If an image is to be represented with only 16 colors--which colors should be selected out of the many possibilities? As this image was "sampled" down in color by Photoshop, priority was given first to true black and to true white, followed by the shades in the image which tended to be the most frequent, with the rest divided to maintain the transitions as smoothly as possible. In order to anti-alias an image, new colors have to be created which are halfway between the colors of the objects that are bordering on each other. This tends to generate more colors than the palate can handle so only colors that are already prevalent in the picture can be used in anti-aliasing.

8 bit color

8 bit color allows for 256 distinct shades (2 x 2 x 2 x 2 x 2 x 2 x 2 x 2) and this image puts them to good effect. Perhaps too good an effect, as the image had a specialized palate created for it that uses the best possible 256 colors, creating a favorable condition. Imagine adding a fourth object to the scene that uses a different color or even a pattern on its surface. If a new palate was selected, the gradations between the colors on the current objects would grow more pronounced--but what if a new palate could not be selected? Then only current colors could be used to represent it, perhaps resulting in an image that does not accurately represent the new object.

16 bit color

16 bit color is an enormous jump to 65,536 individual colors. Even so, it is still often not enough to convince the human eye. It is interesting to note, however, that the previous "problem area" of the concentric rings on the sphere has been rectified as many those rings have grown smaller in diameter than a single pixel. On the other hand, the infinitely subtle gradation of the color change across the flat surfaces of the cube now stand out like a sore thumb--the top seemingly represented by only two shades of green!

32 bit color

24 bit color and 32 bit color both contain 16.7 million colors. Why are they the same? 32 bit color essentially is 24 bit color with the addition of an 8-bit "alpha channel" which is used for blending images to create various special effects. Both create colors by blending 256 shades of red, blue, and green. Black would be represented by [0,0,0] while white would be represented by [255, 255, 255]. Still, to many eyes--and this is subjective--the green faces of the cube still contain several abrupt transitions meaning that 16.7 million colors are not enough.

Some models of consumer level flat-bed image scanners have already made the transition to using 48 bit color to gain more accurate images even if the results are not immediately apparent on-screen (as their video cards are not capable of displaying that bit depth). By creating a better initial palate of colors, even if the image is eventually sampled down to a lower bit depth, the result will be more accurate. This is why many software engineers are pushing for a move to 64 or even 128 bit graphic displays as in complex 3D games, objects are often rendered using a variety of overlapping effects such as textures, bump-maps, light-maps, shadow-maps, and even more advanced "pixel-shader" effects. The accuracy of a higher bit depth allows for them not to interfere with each other for a more accurate image when it is eventually displayed onscreen.

B&W Dithered

However, there is a smarter way to use your color palate that while less accurate mathematically (generating noise in the image) is often more appealing perceptually. This is called "dithering." There are many ways to accomplish dithering. Some use static patterns to represent colors while others use random seeds to create noise. This image still uses only black and white yet it is able to largely represent the lighting of the original

4 bit color Dithered

A dithered 4 bit image (16 colors) is much more three-dimensional than its original counterpart. However, the dithering has produced artifacts where the algorithm defining the dither pattern has decided that blue pixels would be appropriate to use in the middle of the red sphere. This might not seem objectionable mathematically to the computer--but it most likely is to its user. This is one of the many draw-backs of dithering images.

8 bit color Dithered

An 8 bit image (256) color is particularly convincing when dithered with a minimum of artifacts.

Please click here to continue to the second page.

by Richard Leader

[back to Bargain Bin Review]