pixel

A pixel — short for picture element — is the smallest piece of a digital image you can address and assign a value to. Imagine laying a fine grid of tiny cells over a scene and recording one number (or a few numbers) per cell: how bright it is, or how much red, green and blue it carries. Each cell is a pixel. Crucially it is a sample of the light, not a literal little tile of the world: the camera integrated the light falling on a small sensor region and condensed it into a value at a single grid location.

Formally, a pixel is a sample of the image function at a discrete spatial location, addressed by integer coordinates (commonly row and column, or column x and row y). In a grayscale image a pixel stores a single scalar, its intensity; in a color image it stores a vector, typically three components for red, green and blue. The numeric range is set by the bit depth: an 8-bit pixel takes integer values 0 to 255. A pixel has no intrinsic physical size — its real-world footprint depends on the lens and the distance to the scene — so a pixel is best thought of as an array index plus a stored value, not a measurement of area.

The familiar mental image of a pixel as a small colored square is a display convention, not what the sensor recorded. The underlying truth is a point sample, which is precisely why reconstruction, interpolation and anti-aliasing matter when you resize or render an image. Confusing the rendering square with the sampled point leads to errors in resampling and sub-pixel geometry.