A spatial resolution smaller than that described by one pixel. Although digital images are composed of a matrix of pixels it can be very useful to resolve image detail to smaller than pixel size or position, i.e. sub-pixel. For example, the data for generating a smooth curve on the screen needs to be created to a finer accuracy than the pixel grid itself, otherwise the curve will look jagged. Again, when tracking an object in a scene, executing a DVE move, or calculating how a macroblock in MPEG-4 AVC coding moves from one picture to another, the size and position of the manipulated picture or element must be calculated, and the object resolved, to a far finer accuracy than the that of whole pixels, otherwise the move will appear jerky or wrong.
Moving an image with sub-pixel accuracy requires picture interpolation as its detail, that was originally placed on lines and pixels, now has to appear to be where none may have existed, e.g. between lines. The original picture has to be effectively rendered onto an intermediate pixel/line position. The example of moving a picture down a whole line is achieved relatively easily by re-addressing the lines of the output. But to move it by half a line requires both an address change and interpolation of the picture to take information from the adjacent lines and calculate new pixel values. Good DVEs and standards converters work to a grid many times finer than the line/pixel structure.