A way of mathematically reducing a data set to a subset of its original components. For example, if you have captured a 512-by-512 image, but want to encode it at a resolution of 256-by-256, you can subsample the data before encoding it by discarding every other pixel in both and and directions. On systems that work with YCbCr data, it is common to subsample the Cb and Cr chrominance information so that each pixel has a unique luminance value, but shares chrominance information with one or more nearby pixels.