The Bit Depth of an image refers to how much computer memory is devoted to each pixel in that image. One bit in computer memory represents a single switch that can be in the on or off position. If an image was only represented by one bit, each pixel could only have one of two states for its red, green, and blue channels: on or off. This source image would display like this as a one bit image. With two bits, there are 4 possible combinations of the switch positions in memory, giving as many as 4 different steps in brightness for red, green, and blue values in a pixel from dark to light. With three bits, that number goes up to 8, and so on. Eight bits of data provides 256 steps in brightness from dark to light for each of the red, green, and blue values in a pixel. The combination of all three channels can produce over 16.7 million colors, and so 8 bit images have long become the standard for computer graphics and computer displays.
With 256 steps in brightness for red, green, and blue, the human eye can barely notice the shift from one shade to the next, creating the illusion that there is a smooth shift from one color to the next across an image. However, a problem can occur in digital video when moving between different color spaces, such as moving from the RGB color model, where pixel values are stored as red, green, and blue values, and the various video color spaces, where pixel values are stored as luminance and chroma subcarrier components. When translating between these color spaces, rounding errors can occur producing banding in fine gradients, a phenomenon where the human eye can visibly detect the jump in color values from one part of the image to the next. To solve this problem, many video codecs offer a 10 bit color space. The extra two bits increase the number of brightness steps for each color channel from 256 in an 8 bit image, to 1,024 in a 10 bit image. These extra steps of brightness data ensure enough precision that when an image is shifted to and from RGB space to one of the video color spaces, no banding occurs in the result.
There are several other common bit depths used in video and film work, including 10 bit log, 16 bit linear, and 16 & 32 bit float. Most important in modern video pipelines is this last one, float space, which allows a massive increase in the dynamic range of stored image data. See the entries on log space and float space for more information. Finally, there’s a common confusion when describing bit depth between people who think in terms of bits per pixel (bpp) and bits per channel (bpc). Check out the entry on bits per channel for disambiguation.