A video stream, or video sequence, is generated by sampling fixed
images of a scene at certain time intervals -- temporal
sampling. If the sampling
frequency is high enough, typically between 20 and 30 images per
second, a playback at the same speed will make the eye and the brain
see continuous motion pictures.
Each digital image, also called frame, is generated by
spatial sampling. Using a camera or a scanner, the continuous,
real-life image is converted to a grid of pixels, each having a discrete value,
or a set of discrete values, giving a measure for the intensity or
color of the small square it represents. For grayscale images, the
pixel value is typically represented using eight bits, giving possible
values between 0 and 255 inclusively. The value usually represents the
amount of light within the pixel; 0 is black and 255 is white, while
the values between give various shades of gray.
Pixels of color images normally consist of three values, describing a color in a certain color model. Well known color models include RGB, where the three values represent the red, green and blue color components, and YCbCr (the digital version of YUV [8]), where one value is used for intensity, while the two others are used to represent chrominance. RGB representation is used by most (if not all) color monitors, while YCbCr and other schemes, separating luminance and chrominance, are used in several image representation schemes containing irreversible compression, along with television sets. For more on color models, see for instance [9, chapter 13,] or [10, chapter 3,].
To sum up, a raw video stream is a sequence of bytes in which a single or a triple of bytes represent a pixel. A sequence of pixels represent a single image, and a sequence of images make up the entire movie.